Love it when those that are on top are making it harder for those in the bottom to innovate and move up.
GDPR is doing great at keeping creative-less* companies in check.
----------------------------------------
Creative-less because they can't have a vision and rather gather data and try to find something customers like to keep milking until the product inevitably dies.
I’m not a fan of comparisons to the Hippocratic oath. The greatest risk to AI ethics is not the ethics of software engineers but the ethics of the software engineering process. By the time tasks are handed to engineers, most of the ethical decisions have been made by product managers, designers, and business stakeholders who are focused on their own goals. Software engineers are accountable to their bosses before their users, no matter how high minded we like to pretend to be. To say it’s on the engineer to do no harm puts them in the tenuous position of doing the job or being replaced by someone who will. That isn’t setting us up for success.
The Hippocratic oath sounds more altruistic than the alternatives, but good legislation, including business audits and incentives, will have far more impact than a software engineer swearing they won’t be evil.
Being expected to behave ethically is part of being an Engineer regardless of what your boss expects. Now we increasingly see real world negative consequences of the work of software engineers I don't see why they should be any different.
And, what's more, isn't behaving ethically not part of being human?
Aristotle and a while host of philosophers certainly thought so.
Extending this, why don't we expect everyone in society to behave ethically? But then we get into arguments about what is ethical, because people disagree on that, and people naturally will disagree on what is ethical in software development, which isn't always a problem. The is a diversity of opinion, though since are clearly extreme.
One of the issues with a "Hippocratic oath" for software developers is that software spans the while spectrum of human activity and thought.
Constraining software is nearly equivalent to constraining human thought in more ways than one.
Yeah, the problem with this idea is that it assumes there's one fixed set of right ethical values everyone should agree on. Take for example the actual, original Hippocratic Oath - it required, amongst other things, doctors not to provide abortions to women. There's still a large number of Americans who hold this ethical belief, but I'd venture a guess that almost all the people pushing for ethics in software development hold the exact diametrically opposite view.
A lot of the demands for software ethics get into even fuzzier and more complicated territory, like whether the companies which control our mass communications should use that control to decide which political views people should be able to share and what they should target, whether this is good or bad for democracy, etc. Also, many of them seem to be things that would've been incredibly niche viewpoints even a decade ago. I've seen people in another thread thinking that Google promised not to use e-mail contents to sell ads originally and then snuck that in, but in reality basing ads on the contents of e-mails was part of their business model all along and it's just that no-one cared when they launched because it was so much less obnoxious than the ads on other providers. Somewhere along the way, we got this meme about big tech selling our personal information, and everyone seems to project it backwards in time onto how people felt before the meme.
That's why Smith also talks about a new DIGITAL GENEVA CONVENTION - not just a code for devs - so there is a defined framework for everyone in the chain.
Pls excuse the caps, but too many people seem to ignore this very important part of his argument here.
I thought everyone ignored it because that is what would happen in practice. Like the actual Geneva convention only applying in an unconditional surrender or complete defeat.
I think the point is that we can't just rely on individual ethics to enact change. People have bills to pay and kids to feed, if it's all on the man at the bottom to say no then a lot of bad software is still going to be created.
I think making a statement as an industry helps avoid some of this stuff even being asked. With research you can easily find doctors who are brought in to assist with torture or other heinous crimes, but in general I would expect it to be harder to find a doctor to help with my project of creating a more addictive cigarette versus finding a web developer to help me market them.
There's some (not a lot, but some) power in a profession just claiming to have ethics, even if there is no enforcement and change in the hearts of practitioners.
Plus, at some point, we all just have to own what we do with our lives. It's great to advocate for systemic change but doing your own small part is at least as important.
I think we should license some software developers - if you're developing software to be used in industries where the people using your software need to be licensed (i.e medical equipment, structural engineering, industrial control systems) then perhaps the software engineers should be similarly rigorously qualified.
But if you're working on video games, or pizza delivery, or really probably like 90% of software then no.
Medical software systems already have legal obligations that fall on (I believe) the CEO of a firm. In the CRO space for clinical trials for example you're often audited by the FDA, MHRA, and others to ensure your processes are documented and followed, your issues traced and corrected, etc. You also have a requirement to fulfil things like 21 CFR Part 11 for records and signatures tracking/management.
Failures of a company responsible for this kind of thing that lead to injuries or deaths can result in imprisonment of (at least) the CEO and possibly others. I'm not sure licensing is necessary here, that's what documented process is supposed to manage (change tracking, development process, etc.)
I think the way Uncle Bob said it was reasonable. We have a choice: either we commit to doing good ourselves, and get out of licensing. Or we fuck up repeatedly, and eventually someone else will force a licensing process upon us. And it will not be a pretty one, because it's unlikely the person designing it will know the first thing about how we work.
The consequences of the decisions made by Microsoft Presidents and their ilk, since they are the ones who actually decide what software will look like.
But still it would be good when MS gets found to do unethical things that a programmer could be scapegoated instead of their president, at least from the presidents perspective.
Well there is one major assumption error there - that negative real world consequences are only linked to negative ethics.
That is so wrong it isn't even funny. If the car was invented as powered by a Mr. Fusion the buggy whip makers going out of business would be a negative real world consequence.
In some Engineering societies maybe, in general you may be liable if you construct a building you shouldn't have.
Software is fundamentally different because until you run it it has no consequences, and even if you run it, it can be contained. I can write a worm and not release it on world. In that regard, it is more like engineering _plans_. I can draw up plans for a building that is designed to collapse with X number of persons inside -- in fact I can imagine either of the two assignments given as an exercise in University.
No reason to make more laws: it should be immaterial whether I chop down the Christmas tree at the local town square or program a robot to do it.
I agree, however not all ethics can be legislated.
This sort of thing works in other professions like medicine because malpractice can cause doctors to lose their license. Same with civil engineers. This changes things because the choice is now quitting or possibly never being able to work in the field again.
Perhaps principal software engineers in charge of life or death software should be licensed for accountability, “engineer on record”.
But that's not an independent fact. That's one of the major goals of medical oaths; they establish that doctors must be able to determine the right course of action on their own, that ethical doctors won't work for an organization that won't give them that agency.
I don’t think this is true in America. Doctors are limited on what will benefit the system monetarily. Healthcare administrators will decline suggested procedures from doctors due to expense.
> Software engineers are accountable to their bosses before their users, no matter how high minded we like to pretend to be.
I agree to a very limited extent about the hierarchical nature of a typical corporation, but I also disagree. Software engineers at a certain level of their career and with relatively uncommon skills can pick and choose what companies they want to work at. In my opinion people of good moral character and conscience need to be prepared to refuse to accept a position at companies known to engage in activities against their principles. And further need to be prepared to resign if they are asked to do something clearly unethical.
From my particular specialization in network engineering, I would never accept a role at an ISP in an environment where I had to implement something like the GFW in China, or further walled-garden/censorship of the global Internet. It's directly contradictory to my principles. I sincerely hope that the best and the brightest of my colleagues would never choose to aid and abet internet-fuckery by autocratic regimes. If people from my field look at a project and could reasonably say "Vint Cerf would be really disappointed if he saw me implementing this...", I hope they will choose to walk away.
But ultimately there's always a software engineer involved in the creation of software - and that's not true of any of the other roles you mentioned. Since software engineers are necessary and sufficient to produce software, they should always be held responsible, and any oath should fall on engineers.
> To say it’s on the engineer to do no harm puts them in the tenuous position of doing the job or being replaced by someone who will.
Well, yes - if there were no tradeoffs there would be no point in having an oath to begin with. But there are software engineers today, including some on HN, who do things more harmful and unethical than medical malpractice, and they are personally culpable for the decision to do so - just as their replacements would be if they refused. I would also like to see laws criminalizing those individual engineers' conduct - maybe you're alluding to the same thing? - but an oath is a good start.
How about a compromise then? Software engineers are responsible for the negative consequences of the software. Any and all responsibility they take is then also shared with every person that is above them in the hierarchy.
Eg a developer does something and society finds this unethical and punishes them. The developer's boss, the boss's boss, the boss's boss's boss etc up to the CEO all get punished in the same way. Furthermore, to avoid companies trying to shield themselves from this by putting their developers into a different company, it will apply to software that you get from someone else too.
Suddenly this doesn't sound very appealing anymore, does it?
> Any and all responsibility they take is then also shared with every person that is above them in the hierarchy.
Currently, if I write software that performs illegal actions -- let's say software that allows me to use unlicensed Adobe products -- at the request of my boss and their boss, all three of us would be legally liable.
That makes no sense. Management just needs to decide on an ethics standard for the entire company and install an audit process that maintains that standard. That's the entire problem. Don't ask a handful of employees to do some charity.
Not sure hou understand how software is made. A programmer doesn't decide what to write, when to write it and they are lucky to be included in how it's made.
Programmers get specs and write programs to match those.
At no point is it the programmers responsibility to talk about the moral compass of the project and where it fits into society.
An oath to do no harm? You first need to give programmers the power to decide the fate of projects on their own the way only a doctor can decide medicine or treatment.
Although it would be optimal for the management/leadership to not be pursuing unethical developments, a software engineer having the fallback of "I can't implement this in good faith" is another layer of defence (to society).
It would probably also allow for legal push back against being terminated for refusing to implementing the unethical thing.
The programmer still decides whether that code gets written, since they’re the one writing it! If you write or review a piece of software, even if the spec was written by the PM/business, you’re endorsing whatever that spec says and all of its ethical implications. “Just following orders” is a famously poor defense at this point.
Exactly. I can’t believe all the blame-shifting I’m reading in this thread! It’s as if software engineers are suddenly these powerless victims, lacking agency over their work, only capable of saying “yes, boss, whatever you say, boss!”
If a civil engineer’s manager told them to design an unsafe building or bridge, they’re not going to just say, “Sure thing manager! One death trap coming right up!” It is their ethical duty to build it safely.
Eh, this isn't a great analogy either. If I'm an engineer that develops a single beam in a bridge, is it my fault if someone assembles those beams together in such a way that is dangerous? At which point does a function become unethical? Do you now have to only use software made in your country by ethical developers?
Software isn't a bridge and comparisons fall apart quickly.
Wow, I’m realizing what an unpopular opinion this is on HN! Yes, as a software developer you should absolutely be accountable for the ethical concerns around what that protobuf you’re moving around from one API layer to the other is used for. You’re not a code monkey. Ask, and refuse if it’s unacceptable. I have quit jobs where the ultimate purpose of what I was building was evil.
EDIT:
Forum the original OP:
> Software engineers are accountable to their bosses before their users, no matter how high minded we like to pretend to be.
They are accountable to themselves and their own conscience before both their bosses and their users. I understand this is an uncomfortable line of thinking if your employer asks for ethically questionable project work, but I’d argue that if this is the case for you, it warrants career introspection.
Sure, same as the LLVM example someone else pointed out. Good points. I’ll qualify my opinion then. To the extent that the engineer can know the ultimate application of their work, he or she should be responsible for ensuring it is being used ethically.
So, the engineer writing a binary search, knowingly working on “Project Orbital Death Ray” or “Voter Suppression 2.1” should know better. I hope we can at least agree on that one.
The engineer writing a linked list or moving around Protobufs for their some open source toolset gets a pass because their project as they understand it is ethically neutral. BUT there will be that engineer who then takes those tools and integrates it into “Project Orbital Death Ray”. That’s maybe where accountability should begin.
Everyone’s talking about the managers taking the blame and yes they’re culpable too. But at the end of the day an actual software developer’s fingers type the code in. If that developer knows what he is working on, he needs to bear the responsibility, too.
There is a gigantic difference between building an unethical software product and abusing an ethical software product for unethical purposes. The developer is not the user. Do you not understand that?
Engineers at the beam companies just certify that the beam meets its specs.
I'm not sure why software would be any different. Bridges are complicated and made up of versatile submodules, just like software. Some other software engineer eventually designs the "bridge" and selects "beams" for the structure. If those beams fail to meet their specs, then the engineers who stood by them are at fault. If the bridge fails because the beams weren't used in accordance with their spec, or didn't have a spec at all, then the engineers who approved their use in the bridge is at fault.
A bridge is limited to a single purpose, like an appliance. If you insist on veto power over every outcome of what you build, that means you can only ever build sealed appliances for hapless consumers, not unfettered tools that empower clever human beings who will use them in unanticipated ways. Having sworn a Hippocratic oath, are you allowed to work on LLVM, which half of all evil apps probably depend on?
I could get behind a requirement that code be reliable and fit for purpose, though very few of us have any experience with the formal methods that might get us there, and most don't want to work that way.
Imagine if your manager copied the safe bridge you designed with a magic replicator and now uses that exact same design somewhere else. You tell him that the bridge was not designed for this location and that the bridge will collapse in 5 years. Your boss fires you but you are still responsible for the collapse of the bridge.
Let's go further into absurdity. The engineer is kidnapping the daughter of the manager and blackmailing the manager to take the bridge down. Is it ethical to force someone else to be ethical even if its only possible through unethical means? What if there is a hero saves the daughter? Will the hero be liable for the collapsed bridge?
This argument is more akin to saying its the builders responsibility to decide whether a bad architectural decision should be built or not. They might bring it up, but it's not really expected that they get to decide.
This is a great point, and reasonable people can disagree about when an application's abuse or potential for abuse crosses the line. The same goes for pivots or for general-purpose code that's used elsewhere. (Is it ethical to contribute to internal tools at Facebook? What if those tools make other engineers way more effective at doing things that ultimately undermine democratic systems?)
My point here isn't to dictate what software is or isn't ethical, but to argue that if a program is unethical, its ethical implications are the responsibility of the engineer(s) who wrote it.
> You write a tool for let's say recognizing faces. Will it be used for login onto computer? Tracking dissidents? IDing corpses? Who knows.
I mentioned this in another comment, but I'll say it again:
Irrespective of any legal/ethical concerns, yes, I would like to know! If my boss just came to me and said "build a facial recognition system" I certainly will ask how it is going to be used. Not because I care about ethics, but it's a basic aspect of the job. You can replace "facial recognition" with "CMS" and I'd still ask.
If they tell me the facial recognition is for logging into computers, and then later decide to use it to track dissidents, that is a different concern. But I'll at least ask!
> What if you start as something 100% ethical. But your company pivots to unethical application?
If they pivot after my work is done, I won't feel responsible. If they never used it for the original application and pivoted to this, I may get upset and quit, but my conscience would be clear.
> If they pivot after my work is done, I won't feel responsible.
So if you invented dynamite you wouldn't feel responsible for its use?
But, let's change it a bit more personal. You write an awesome OSS yaml parser. It's so good, that GFW of China uses it as a main component, and this gets published in the news.
What would you do? Nothing you did changed, but suddenly your work is powering an unethical component.
Dynamite is a good example of why the philosophy is complete bullshit. How about you blame the stupid evil fuckers who started all of the futile wars to try to get rich in the quagmire that was European geopolitics instead of tbe person who made them safer?
It is the same sort of stupid blame shifting involved with the hippocratic oath for x nonsense. Oaths are majorly outmoded in the zeitgeist anyway because everyone recognizes lies are commonplace.
> How about you blame the stupid evil fuckers who started all of the futile wars
And I fully agree. Expecting people to individually bear the burden of "some oath", is a fool's errand.
My point was software on its own, much like a fridge, is amoral. You can use it to store your groceries, or you can use it to store corpses.
That said, there are some extreme cases (like a gun), that have very limited non-violent uses. And IMO, that should be regulated, instead of depending on people Doing The Right Thing™.
> So if you invented dynamite you wouldn't feel responsible for its use?
I would if I were inventing dynamite, but that's not what this scenario is.
A person working for a knife manufacturer need not worry about it being used for murder, as that's not what the primary use. And facial recognition is a lot less harmful than even that.
Trust me: I work for a company that produces certain goods used for all kinds of good and nefarious purposes depending on who buys it. My conscience is clear.
> But, let's change it a bit more personal. You write an awesome OSS yaml parser. It's so good, that GFW of China uses it as a main component, and this gets published in the news.
> What would you do? Nothing you did changed, but suddenly your work is powering an unethical component.
I wouldn't do anything:
1. This is milder than the knife scenario above. Of course I don't care if people use it in a poor way - unless there is a straightforward technical mitigation I could do. In your example, given that the source code is available, that is not an option.
2. There's a certain hypocrisy in releasing something as open source and then complaining about how it is used. If it bothers you, then modify your license!
That's hypocritical. Nobel didn't invent dynamite because he wanted people to blow themselves up. He invented dynamite because nitro-glycerin was a horrid mess used in mining.
He definitely didn't have an easy technical solution to problem of people misusing dynamite.
You can either say in both case do nothing, or in both case do something.
Right now your boss has no reason not to tell you- people at GitHub knew their software was being used to support ICE during the time in which families are being separated. People at Microsoft knew that Microsoft was having contracts with the military. Google engineers knew about Project Dragonfly.
Right now bosses don’t even have incentive to lie about it because no engineer is obligated to give a shit about the society they live in broadly.
This is a good time to point out that a significant chunk of the population isn't opposed to working for the military, for ICE, or for defence contractors who make weapons; they don't view that work as unethical. Moreover, the origin of Silicon Valley, and indeed the entire internet, is DARPA contracts and weapons manufacturing.
Any oath would either not be taken by those people, would be watered down so far as to be meaningless, or would require the entire industry to refuse to make weaponry. The first and second are ineffective and the third is ludicrous.
So management bears no blame for requiring illegal work be done, on pain of termination? Said another way, engineers now need to be technical and legal experts in the business domain?
(Remember employees in the US depend on the company for health insurance. Saying 'no' could cost a lot more than just ones position.)
Most software engineers are not like doctors. We have little autonomy over what is created. Our responsibility is primarily the how. And with devops sometimes the actual deployment and maintenance itself.
> Said another way, engineers now need to be technical and legal experts in the business domain?
Consider something like the 737 MAX debacle – did the programmers writing the MCAS code actually have enough aviation domain knowledge and understanding of where the component fit in the overall system to realise it was a threat to people's lives?
I don't know, but my guess is the most likely answer is "No".
If a doctor says no, it's because of legal liability and risk to licence.
For the same reason, it's harder to hire some rando budget doctor because the field is gate-keeped by the requirement of a licence, and liability.
You can't magically make Engineering the same without the same conditions. Add barriers top entry that see my pay rise, or have cheap programmers with no liability.
Engineering already has these barriers, they just aren’t required or enforced.
I’ve never been on a project that requires a software product stamped by a licensed engineer. NCEES dropped the software license because there was so little demand, compared to, say, civil engineers who consider a license a rite of passage to career growth.
You need a licence to practice medicine, if there is something called a "license" for engineering but isn't required for practise, then it's not the same. If we have the same barriers, but not enforced, then we don't have the same barriers.
That’s not quite right. You actually do need a state license in the US to practice engineering for the public except for a few basic instances:
1) you work for the federal government
2) you work under a licensed engineer
3) you work under an industrial exemption
There’s differences depending on state. There has been a more concerted effort to remove #3 recently due to both political reasons and the technical issues in this thread. Most people performing engineering work under an industrial exemption don’t actually realize it. Again, this is different state to state. For example, in some states you can’t start a business with “engineering” in the name unless you have a certain percentage of owners/principals with an engineering license.[1]
What actually happens is conflating terms in common parlance. “Engineer” and “engineer” are not necessarily the same. For example, a computer engineer may work under an industrial exemption (due to working in a manufacturing service) while a software engineer does not. Legally, an “Engineer” claims an explicit responsibility to public safety.[2] Apropos to the headline article, there is a distinction with this difference.
A doctor refusing to do a procedure because he worries for his patient is seen as a good guy doing what is right, he is in his opinion saving a life. In addition he is trying to avoid the massive cost of a medical malpractice suit.
A software developer refusing a job because it does not meet his ethical parameters is just an unemployed software engineer.
I think one of the issues is the domain of is vast. One developer may be working on a basic CRUD app while the next is working safety critical code on a vessel going to Mars.
There are definitely areas where the prudent thing for a developer to do is raise a dissenting opinion, if not halting work. What seems lacking is clear industry consensus standards to back up that decision.
From my experience, software engineers don't rotate nearly as often in traditional cyclical and defense businesses (auto, aerospace) as they do in the FAANG technology sector. There are a lot of grey beards who weren't so grey when they started.
That's completely the wrong comparison and context.
The reason MCAS came about was because management wanted to try to fudge a larger engine into an outdated design created for a different purpose rather than do the engineering and certification necessary for the new requirements and to update the system.
Management wanted to save money. Of course the engineering leadership did not want to fudge something -- they wanted to do proper engineering. But the people in charge just wanted to save money, and the engineering leadership could not do anything otherwise, even they knew that just making the engine larger and compensating did not make sense from an engineering standpoint.
By the time it got to the MCAS, that was far down the line of the decision to not do proper engineering.
Which demonstrates the irrelevancy of a hippocratic oath for engineers. Because the complexity of the system is such that no single engineer understands it all, and thus no single engineer is (or feels) responsible.
Blaming it on management is also irrelevant, since management merely takes the advice of engineers, and do financial/business trade offs to maximize profit. If the engineers cannot tell that MCAS system could fail this way (due to the complexity), management will not question it.
Hippocratic oath for engineers is irrelevant, correct. But management does not take the advice of engineers. As I said, the engineers wanted to do proper engineering, but management wanted to save a buck, so they instructed engineering to fudge a cheap solution.
if the engineering leadership said, this is not a good idea. Management said, it saves money.
Engineering failed to convince management. Management didn't have the understanding that it was a bad idea.
It is now, no one's fault?
If management merely takes the advice of engineers ( and other who specialize in the things that they do not ), and they choose to ignore it because they do not understand the things they do not specialize in. I believe it's a reasonable to assume that management is more at fault than engineering ( I'm not sure they're is a situation here where any party is fault less )
I'm not saying engineers taking an oath helps, its about the executives.. maybe something about the thread structure implies that but I actually only read the comment above.
From my limited information the MCAS code was primarily causing problems in association of incorrect readings by damaged sensors. Of course one could argue that this is an engineering failure because the MCAS failed to account for wrong sensor input but when you consider the legal implications of a MCAS fallback there is actually not much that can be done on the software side.
The MCAS is an optional component that reduces certification and training costs. It is definitively possible to fly the plane without accidents even with a disabled MCAS. So why can't the MCAS be turned off automatically when sensors fail? Because that changes the classification of the plane and therefore requires pilots to be certified for a new machine and receive new flight training for both MCAS and no MCAS modes.
If the software engineer was under a hippocratic oath then he would have to refuse to build the MCAS entirely but not because the idea of an MCAS is inherently unethical, no, he would have to refuse because the company he works at wants to use the MCAS for a non ethical purpose (namely operate and hide the existence of MCAS even when it is unsafe to do so).
This is basically a reverse audit but the software engineer has no authority conduct such an audit and even if he was allowed to, the business has no obligation to give him the necessary information to determine whether the MCAS will be used unethically.
> he works at wants to use the MCAS for a non ethical purpose (namely operate and hide the existence of MCAS even when it is unsafe to do so).
You think a programmer, handed a spec and asked to implement it, can be expected to know that their employer (or the employer's customer) wants to use it for a "non ethical purpose"?
Again, I can't know for sure, but I doubt the programmers who wrote MCAS (who most likely didn't even work for Boeing, but rather some subcontractor) actually knew, or could have known, how the code fit into Boeing's larger purposes
With all of that being said, were these software engineers (probably subcontractors) even given access to actual MCAS readouts or, more likely, virtualizations of expected readouts. These people probably didn't account for this type of faulty readout because the virtual machine never put out that type of fault.
Most types of development in these large companies is so compartmentalized that it's next to impossible to see the whole structure from a software engineers prospective. You need to be at a management level to understand how most of the pieces really come together, which is the only place where one of these "oaths" might make an influence. At that point, however, the selection is so goal oriented, I have a doubt as to whether or not people would take that oath.
Generally, there is somebody (typically in software assurance or systems engineering) who is supposed to ensure the fidelity of the simulator. Additionally, the hazard analysis or failure modes effects analysis should trace to specific test cases.
Of course, there’s all kinds of pressures that make these fall through the cracks. I vaguely remember an article stating some of these documents in the case of MCAS were not up to date
> You think a programmer, handed a spec and asked to implement it, can be expected to know that their employer (or the employer's customer) wants to use it for a "non ethical purpose"?
No imtringued does not.
imtringued wrote:
> This is basically a reverse audit but the software engineer has no authority conduct such an audit and even if he was allowed to, the business has no obligation to give him the necessary information to determine whether the MCAS will be used unethically.
imtringued is saying that it would be impossible for a software engineer to determine whether what they were asked to do was ethical or not.
What the actual hazard analysis showed is that Boeing did not have the technical insight at the right level.
The HA listed MCAS as "hazardous" rather than "catastrophic". Meaning those in charge of that process document did not realize MCAS had the ability to down the airplane. I know it's tempting to arm-chair quarterback this, but let's assume they should have realized this hazard.
To your point, maybe the programmer doesn't have the systems knowledge to make those calls, but the process is predicated on somebody having both the technical acumen and the responsibility) for those decisions. This process broke down though.
>actually not much that can be done on the software side.
This isn’t exactly true. There are mitigations (both software and non-software) that are expected to be done depending on hazard analysis. One of the items discovered is Boeing mischaracterized the MCAS hazard (it should have had a “catastrophic” hazard class). In addition, they didn’t appear to follow their own process for dual inputs required even for the lower severity class assigned. The “optional” part of MCAS was the secondary sensor reading into the software
>The MCAS is an optional component that reduces certification and training costs
No. The MCAS was a "necessary" component for pitch stability - without it, a 737 MAX in a pitch-up attitude would, in the absence of correcting inputs, pitch up further and further until it stalled. Without it, the airframe is uncertifiable, full stop.
I'm certain that's not correct, everything I've read on it has said MCAS was specifically a software modifier put in place to allow the plane to respond substantially the same as a regular 737 without the larger engines, in order to avoid having to have additional training for all 737 pilots worldwide.
Most aircraft, in a "pitch up attitude" will increase their angle of attack as thrust is applied. The issue was that the MAX would do so in a more radical way than the regular 737 did, and so the software was put in place to limit that so it flew like a regular 737 as far as the pilots could see.
Conceptually, MCAS wasn't a bad idea. The execution and using it as a replacement for training and not informing pilots of the flight characteristics changes between the models was stupid.
Although to be fair my summary wasn't entirely accurate - it wasn't that a MAX was outright dynamically unstable with no control input, as I described, but rather not sufficiently stable as to cause a monotonic increase in stick force as AOA increases, which can cause the combined system of pilot + flight dynamics to be unstable since the pilot relies on stick force as an indicator.
> Most aircraft, in a "pitch up attitude" will increase their angle of attack as thrust is applied
This is both incorrect and irrelevant. Most aircraft will climb when power is applied, but will not change their AOA unless the thrust axis is off-center. To a first approximation, power controls climb rate, and stick input changes AOA. Change in behavior under different power settings has little to do with the problem with the MAX. The problem with the MAX is that at high angles of attack - i.e., when the stick is held back, causing the air to meet the wing (and the engines) at a steeper angle - the engines, which are flung forward, start producing lift of their own and produce a pitch-up moment. This means that the further the pilot pulls the stick back, the less hard they have to pull. This is a dangerous inversion that increases the control order of the system, as it breaks the usual assumption that a given stick force will result in a given AOA, more or less.
Right -- it didn't give the exact same feedback to the pilot that the regular 737 did, which was why MCAS was created. The aircraft is no more or less unstable than a regular 737.
The original 737 does exactly the same thing the Max does with respect to producing a pitch-up moment -- as does nearly every other aircraft. It's just not nearly as pronounced as the Max is.
I'm sorry, none of that is correct. Did you read my link? The aircraft doesn't meet FAA regulations without the MCAS.
>The Boeing 737 MAX MCAS system is there ONLY to meet the FAA longitudinal stability requirements as specified in FAR Section 25.173, and in particular part (c) which mandates "stick force vs speed curve:, and also FAR Section 25.203 — "Stall characteristics".
> If the software engineer was under a hippocratic oath then he would have to refuse to build the MCAS entirely but not because the idea of an MCAS is inherently unethical, no, he would have to refuse because the company he works at wants to use the MCAS for a non ethical purpose (namely operate and hide the existence of MCAS even when it is unsafe to do so).
One, it's not going to be clear from the request that the MCAS would want to be used in unethical ways.
Since the Hippocratic oath is the argument here, how many software developers want to work in a system closer to physicians? A national cartel controlling membership and licensure - tough luck if you want to hire more developers because there's an artificially limited supply. Mandatory academic training - goodbye self-taught developers. Follow-on training with pay 1/5th or less of your attending physicians - I know residents in specialties where attendings are paid $500k a year to start, and they are making $60k a year. Brutal shift work - residents work 70-80 hours a week easily. Toxic leadership - I've heard horror stories of residents being forced to lie on ACGME forms regarding their hours under penalty of being outright fired from their residency slot, which would make it nearly impossible to get a job as a physician (mainly because you'd have to apply to a different residency program and explain your termination).
I know they're not suggesting bringing the entire medical education & training structure over to tech workers, and everyone here likes to think that they're brilliant and changing lives every day but most of us are just throwing shitcode JS into a computer for 3-4 hours a day for an ad tech company and not much more. The comparison falls apart pretty quickly.
Edit to add: If you are being asked to perform illegal or unethical acts as part of your employment, then perhaps termination is an ideal course of action? Unless of course your personal enrichment outweighs legalities or ethics in your worldview?
And I wasn't implying engineers should be entirely blameless. Everyone has a limited understanding of legal systems too complex for one person to fully grasp. And workers far below the level of decision makers should be judged according to evidence of their knowledge and responsibility. Likewise those who give orders should bear more responsibly.
All these "companies take on a life of their own" arguments sound a lot like executives priming the pump of potential jurors with excuses. If decision makers cannot bear responsibility because of a company size or organizational structure then we can make some sizes and structures illegal before they stumble/march into devastating incompetence.
It may be fine in some countries, but saying that you’ll make some organization sizes and structures illegal, barring other criminal activity, smells like a violation of the freedom of association.
Because it allows people to associate with whom they choose to. Remove that and you’ve opened the gate to legal racism, legally institutionalized homophobia, banning of religion; the list is endless. The five freedoms are the pillars of our Constitution. Without them, we are no better than China or Russia or even any third-world hellhole you care to mention.
I don't see that even a little bit, your cause effect isn't explained.
I should phrase it differently. Why is an absolute freedom of association more important then the freedom from being harmed by large associations with amoral machinations. The original argument asks that if large corporations inherently obscure moral outcomes, maybe they are immoral, which is an argument that puts these two moral axioms in conflict. Simply stating that one side wins is thought terminating; its important to argue for why its better.
Morality is highly variable, depending on the observers beif system. Legality is the only framework that we can establish in common. Ethics comes in second as it can be established by a group and does not bind those outside the group.
It doesn't matter if it's a good freedom, the chances of the US repealing the 1st Amendment any time soon are basically nil. You'd have a better chance of getting Apple/Amazon/Google to voluntarily split up their own companies out of the goodness of their own hearts -- it just isn't going to happen.
The only argument that actually matters here is whether or not restrictions on corporate structure actually do violate freedom of association or not.
I'm reasonably skeptical that they do, given that the 1st Amendment hasn't stopped us from enforcing antitrust and monopoly legislation in the past. Yeah yeah, Citizens United and all that, but we regulate companies all the time.
But I'd still want an actual lawyer to weigh in on that, I wouldn't feel confident saying that there aren't limits on how far we can go in that direction.
Antitrust doesn't violate the first amendment, so clearly limits on corporate scale aren't unconstitutional, so the legal defense is insufficient and the moral question stands.
> I'm reasonably skeptical that they do, given that the 1st Amendment hasn't stopped us from enforcing antitrust and monopoly legislation in the past. Yeah yeah, Citizens United and all that, but we regulate companies all the time.
> But I'd still want an actual lawyer to weigh in on that, I wouldn't feel confident saying that there aren't limits on how far we can go in that direction.
It doesn't necessarily hold that because one thing is legal, everything is legal. For example, we have 1st Amendment restrictions on threats and libel, but in the US hate speech is still protected speech. 1st Amendment exceptions are generally pretty narrow and specific in the US.
In the same way, clearly some corporate regulation is OK. It does not follow that there's literally no limit on what the government can dictate about how a company can operate. I would prefer to get input from a lawyer before asserting that so confidently.
> ...then we can make some sizes and structures illegal before they stumble/march into devastating incompetence.
Was with you until this part. Just hold them personally liable if someone gets hurt should they create an uncontrollable system and predictably fail to control it.
Right. My point was in response to excuses being made elsewhere that the nature of large companies mean these executives cannot be personally liable. So if we accept that the nature of huge companies is no one can be liable (I'm not convinced yet) then it would be time for capping sizes or outlawing structures.
Keep in mind the US already has laws around corporate structures and conflicts of interest. (Even if they're selectively applied.)
The nature of any size corporation is to have one person in charge. In terms of assigning responsibility I'd think that works better than the alternative you'd get by breaking it up. Namely a bunch of cooperating smaller firms only doing part of the job each, and able to point the blame at each other.
We heard the "too complex to understand" excuse a lot regarding the pricing of subprime debt. Except a lot of people did understand it was a problem. It's basically the "I'm too stupid to know what I was doing" defense. If we accept that defense and try to make regulation to protect them from failing (as was done in finance back then), we basically allow stupid people to continue to be in charge rather than being replaced as they need to.
> So management bears no blame for requiring illegal work be done, on pain of termination?
No, I wouldn't say that. In many cases, management and engineering share the blame jointly and severally since they both have an opportunity to stop it.
> Said another way, engineers now need to be technical and legal experts in the business domain?
Engineers should know enough about their business domains to understand the ethical impacts of their work. Ethics and law are orthogonal, so thankfully this is generally much easier than being a legal expert.
> (Remember employees in the US depend on the company for health insurance. Saying 'no' could cost a lot more than just ones position.)
Thankfully, the healthcare safety net in much of the US is far better than it gets credit for, and the pay and availability of opportunities for software engineers in the US has generally been quite good. I'm sympathetic to this argument in general, which is one reason I don't think there should be an oath for, like, Amazon warehouse workers, but I'm far less sympathetic for anyone making 5+ times the median US income.
> Most software engineers are not like doctors. We have little autonomy over what is created. Our responsibility is primarily the how. And with devops sometimes the actual deployment and maintenance itself.
Your responsibility as framed to you by the business is the how, but upstream of the how is the question of whether or not to do it at all. If you contribute to a piece of software, you've tacitly answered yes to that question.
> Engineers should know enough about their business domains to understand the ethical impacts of their work.
This would be more reasonable in the era of 10-20 years in the same company or industry. Needing to job hop every 2-3 years for a decent raise, and software skills applying to a vast array of industries, makes it less reasonable IMO.
> ...but I'm far less sympathetic for anyone making 5+ times the median US income.
Not everyone here or in software makes that kind of money. Some of us in the Midwest--or who aren't as skilled at negotiating--don't pull down nearly that much.
> This would be more reasonable in the era of 10-20 years in the same company or industry. Needing to job hop every 2-3 years for a decent raise, and software skills applying to a vast array of industries, makes it less reasonable IMO.
Yeah, to be clear, I don't think software engineers have an infinite level of responsibility for understanding the ethical implications of their work. If you were a software engineer at a credit rating agency in 2006, and you didn't see the ethical dilemma because you didn't anticipate that contagion would be exacerbated by the shadow banking system to bring down the global economy, you get a pass. But if your prospective employer is, like, locking children in cages, or spreading disinformation on political candidates, you should probably find that out during the interview process.
> Not everyone here or in software makes that kind of money. Some of us in the Midwest--or who aren't as skilled at negotiating--don't pull down nearly that much.
Good point - I'm also in the Midwest and make less than that, for what it's worth. I've naturally had FAANG in mind as I type these comments, and more generally I think salaries for the more unethical roles tend to skew higher.
> Yeah, to be clear, I don't think software engineers have an infinite level of responsibility for understanding the ethical implications of their work.
Yep. That's more the responsibility of product managers, upper managements and chief architects/engineers.
Ethics and law are not orthogonal. Look at definition from Wikipedia: "Ethics or moral philosophy is a branch of philosophy that "involves systematizing, defending, and recommending concepts of right and wrong behavior"."
Ethics forms basis on which law is built.
And, given that, it is not simpler to make ethical decision, it is harder. The decision should be worth being basis for a law, how's that simpler than following law?
Even if you take definition of ethical decision from Wikipedia: "An ethical decision is one that engenders trust, and thus indicates responsibility, fairness and caring to an individual." These words can bear negative connotations - if I beat people, I should be trusted that I will beat people, I should beat people fairly and I should care to beat an individual thoroughly.
Yes, I fully see you talk about principles. It can be seen that it is easier to make decisions from principles. But, you can misguide yourself about application of these principles.
> So management bears no blame for requiring illegal work be done, on pain of termination? Said another way, engineers now need to be technical and legal experts in the business domain?
This is a red herring. The oath is not needed merely for illegal work. In fact, the more common use cases will likely be legal. It's a common sentiment, but: Don't conflate ethics with legal.
> Most software engineers are not like doctors. We have little autonomy over what is created. Our responsibility is primarily the how. And with devops sometimes the actual deployment and maintenance itself.
This is not a dichotomy - there can be a spectrum. You can restrict it to those who do know what the product is used for, or at least have good guesses for them.
And while not everyone is this way, I wouldn't really want to work in a job for long if I'm not told what the code I'm writing is for. It's not even an ethical concern for me - it just makes for a boring job. Ideally I want people to tell me the problem they are solving and give me some leeway in crafting a solution. Don't come to me with a solution and ask me to implement it.
> Most software engineers are not like doctors. We have little autonomy over what is created. Our responsibility is primarily the how. And with devops sometimes the actual deployment and maintenance itself.
If you're an actual engineer of any kind, you always have some choice on this. You make architectural decisions every day, and you generally work for places that do, in fact, take your input into consideration. If you don't work for any of those kinds of places, then you are still responsible because you wrote the code to enable it. You can always say "no". There are consequences for that, for sure. You can always quit as well. And it may still get made. But it won't be by you.
And sometimes, that's still better than the alternative.
You are missing the point, the idea is that a average developer at a major bank has very poor understanding exactly what impact of his code will be. He generally has neiglther the information (secrecy, need to knwo basis) nor the understanding of the financial system.
On the contrary, a doctor's desision affects the life of an individual patient in a very clear and understandable manner.
Just check out cases of VW software devs who added code for emission test cheating. Check out the case of dev who was copy-pasting code at Toyota.
Illegal stuff is illegal because not knowing the law is no excuse. For your own safety and good, you better have some grasp on legal stuff in your domain. Don't have to be an expert.
Very few software engineers in the US, despite receiving health insurance as compensation, do not have sufficient cash comp to purchase health insurance on their own should they need to.
I believe that this is a red herring.
Yes, the drone operator that pushes the button is a murderer, even if it was the POTUS that gave the order for it to happen.
I don't think the drone operator is a good example. There are situations in which unethical things must happen, no matter what. If wars could be fought without killing we'd be doing that already.
How much autonomy do you think doctors have? They're not in the operating room inventing new procedures. They are following careful scripts, adapted for the intricacies of one particular human body. There are times where they need to qualitatively improvise, yes, but that's generally only when something has gone horribly wrong.
Also, the Hippocratic Oath is not terribly complicated. It basically says, I will not furtively and maliciously hurt people in an abuse of my authority, and I will try to heal them when they are sick. I don't think it's a lot to ask that software engineers to agree not to create knowingly malicious software. It actually addresses exactly the problem you describe. If everyone has taken this oath, and adheres to it, there is no "someone else" to do that evil work.
Last point, there is a wide variety of software engineering work out there. Some of it may be mindless of the bigger picture of what is actually happening, but for any sufficiently advanced behavior to emerge out of a complex software product, some engineer at some level has to have some idea of the path they are going down to create or allow that behavior. And every engineer has the ultimate autonomy over how and what is created because it is our hands on the code. If you don't understand that, you don't understand the power of the profession.
We can take lessons from other engineering domains. Disregarding exemptions, engineers who design and build for “the public good” have someone who is ultimately responsible. A newly minted civil engineer, for example, may be low on the hierarchy but has to work under the direct supervision of a licensed Professional Engineer. That licensed PE is the one responsible, legally and ethically, for now just the “what” but also the “how”. They may work for a project manager but ultimately it’s their stamp that allows the build.
As someone who works in safety critical code nothing irks me more than when people absolve themselves by saying “I’m just a programmer, that’s not my job/problem”. We need to hold ourselves to a higher professional standard
The problem is, software can be really complex it's possible to make it so no one programmer can understand the whole picture. The tasks can be divided so the individual programmers are given order like "do X in condition of Y" which itself seems harmless and lawful, but combining them lead to malicious behaviours.
this is precisely the problem. few physicists knew they were all collectively building a nuclear bomb because the big picture was well compartmentalized. feynman only found out because he picked the locks of his colleagues to piece it all together.
Can you point to the part of the book where it is stated that only few physicist (not including Feynman) knew this? I remember it very differently. From my memory:
- lock picking was for fun and didnt gain info
- some warehouse workers didnt know about critical mass of uranium and stored the stuff to closely together
Maybe you meant the part where it is clarified that the weapon was not developed against Germany but strategically against the UDSSR. (if I remember that part correctly)
Just finished “Atomic” by Tim Baggot. A history of the development of the bomb. From my reading, lots of the physicists knew. They pushed forward as they didn’t know the extent of the german project (who had stalled due to lack of resources and a belief that while possible, a bomb wasn’t practical due to a need for a large amount of material). The safe-cracking was for amusement.
> Since software engineers are necessary and sufficient to produce software, they should always be held responsible, and any oath should fall on engineers.
That's not true in any way. Lots of software is written by people who don't even have a degree, others by some who have a computer science degree, but not an engineering one, etc.
The other issue is that software is rarely unethical. The unethical bit often comes from the way it is used.
And I'd have to agree with OP. In an idealist world you could assume software engineers would all be ready to quit their job at any sight of unethical affair, even say, launching something to production with a known vulnerability, or without anything but the most rigorous security review process having passed. But in practice, you're not going to achieve this result, unless you put a framework to incentivize software engineers towards being ethical. If you allowed them to sue their employer, and made it that they more often win the lawsuit, for asking them to build something unethical, or insisting that they do so even after the SE said it was unethical, or to retaliate in any way to an SE refusing to build something on ground of ethics, then you'd maybe start to see results. Otherwise, won't happen, and you've only created a scape goat to make it even easier for companies to push for unethical software, since they can now just blame SE they coerce into building it anyways.
I think it something should already kick in if you create tracking pixels to read canvas data to identify users or generally work on fingerprinting. Especially if it is for a benign purpose as advertising, an industry that is notoriously toxic and would have no problems selling ever kind of data they get their hands on. It is fine to generalize it that way in my opinion. The directly conflict with any spirit of the law in most countries regarding privacy.
Aside from that, the quantification of attributes/properties of people can have negative implications for many people. Oversharing is a problem on the net, but at least here people just endanger themselves.
> But there are software engineers today, including some on HN, who do things more harmful and unethical than medical malpractice...
I'm having a hard time trying to find examples of this, outside the field of armament development.
And in those fields where a software failure may result in death, e.g. aircraft development, proof of a software engineer willingly causing it, would likely result in jail time already.
The big question is: who has ultimate visibility on the consequences of a particular project? Very frequently software engineers are asked to work on projects where they only know one side of the picture. The executives in the company are the ones who know the ultimate context of what they're doing.
1) New regulation forces some branches of software engineering to have some type of oath.
2) Now some software jobs will require only oath takers to do.
3) A well payed and powerful new cast of software engineers is born.
4) They are highly paid and have a powerful lobby working for them.
5) The oath takers become very frisky and only work on jobs with minimal risk. The ones that do screw up have an armada of lawyers, because of course they have a new association with deep pockets.
6) Innovation stalls for a while.
7) Big corps start outsourcing some of the oath-taking jobs. These engineers are not bound by the same regulation. Screw ups happen, people die at some point.
8) Maybe we should have the outsourced engineers also take an oath? Back to square 1
This is exactly what I found happened for medical Doctors in Canada (don't know for US). Not saying doctors are not doing a good job, and I can't imagine the stress and pressure they operate from. But suing for malpractice in Canada can be challenging to say the least. I have personal account of a Family member who was grossly mistreated, and all the Doctor did was changed hospitals, nothing more than a slap on the wrist.
> But ultimately there's always a software engineer involved in the creation of software - and that's not true of any of the other roles you mentioned. Since software engineers are necessary and sufficient to produce software, they should always be held responsible, and any oath should fall on engineers.
Nah, it's not the same at all. The fundamental difference between creating a program and medicine is creating a program only has to be done once, or at least a only by a few.
Medicine on the other hand: it has to be redone with each new patient. If the Hippocratic Oath works to prevent 99.9% doctors from doing a harmful procedure then you've hit a home run. Sure, you will never completely stop some bad egg removing a perfectly good limb because a patient suffering from Xenomelia offered enough money. But who wouldn't call a thousand fold reduction a huge win.
We demonstrably have the 0.1% of programmers who are willing to break any oath. They make malware, and willingly take out Sony as mercenaries because Kim Jong-Un got pissed off at a movie. All that 0.1% has to do is write the program once. Thereafter you are not trying to discourage hoards of high skilled professional from doing it again, you are trying to stop a legion of dark net operators copying the thing and selling it to anyone. An oath is a waste of time under those circumstances.
With a combination of local engineers, remote engineers in other countries, mechanical turk and some sleight of hand, I wonder if you could craft a nefarious project where nobody knows the whole picture.
Corporate strategists blaming software engineers for the consequences of corporate strategy is a fairly brazen kind of blame-shifting.
A system of ethics within the health system are necessary for customers to retain trust in the health industry. It's also strongly aligned with the selfish interests of workers who must enact that system of ethics. These properties do not neatly translate to software engineering—mostly because the most difficult ethical dilemmas in technology are rarely obvious when looking at source code. The problems with Facebook (for example) are not always inherent in code; many are only revealed after being deployed at scale and external groups begin exploiting the system.
> Software engineers are accountable to their bosses before their users, no matter how high minded we like to pretend to be
I've quit jobs in the past because of ethical concerns about the way in which those above me have been acting. In one case this involved bribery of senior government officials to push through a project that put at risk the privacy of hundreds of thousands of people.
If you go along with shit like that, you're an accomplice and share partial responsibility. As professionals we have a responsibility to stand up for what is right. It's not good enough to fall back to the lazy excuse of "just doing my job".
And I mean, that is good of you to have a moral backbone, unfortunately there are many people behind you that will do the same job in software and they can be located anywhere in the world.
In the current state of things, that is factually true.
However the same argument could be applied to labor abuses in the textile/garment and shoe manufacturing industry. Most people who follow news are probably aware that about 20 years ago Nike, Adidas and other brands went through a period of terrible public relations disasters, after the working conditions in some of their shoe factories in developing nations were exposed by journalists.
The argument that could have been made at that time would also have been "well but if we don't employ these people, somebody else will just do the same thing with equally terrible labor/human rights violations somewhere else in the world, with even cheaper factories".
The situation today is not great, but it is significantly improved from how it was twenty years ago. There are third party neutral inspection/oversight agencies. Companies in the garment industry are forced to make public commitments to labor rights and reasonable working conditions, and to allow external auditing. They can't just hand wave away the problem and say "but if we don't do it , someone else will.."
The more abstract the problem the less people care. Police suffocating someone to death > 1,000,000 people being subtly spied on. "I got nothing to hide" after all.
The farther away and more abstract the problem, it's absolutely more difficult to get people to care. Trying to get Americans to care about what's going on with prison camps in Xinjiang province right now is difficult, if not impossible, outside of the subset of people who deeply care about the issues involved.
And again, the analogs in software are much more, messy.
A shoe is a physical artifact. It must be made somewhere out of something. Software is much more flexible. If I tell you to build an 'ethical' piece of software with an extensible API, it will only take a tiny amount of work to make it do something unethical.
And much like the shoe companies abusing labour's, that had nothing to do with the engineers that made a a product line, but the management that is driven to ever lower costs.
Also unlike a shoe, the management of a company with ethically produced/designed software could fire all of the original developers and sell the company and its IP. To an organization that would make API changes/data storage changes and it would suddenly become something much more harmful.
I don't want other people making decisions on morality for me. There are people who don't believe that software developers should develop software for the U.S. government because they don't like the way immigration law is being enforced. I'll make my own decisions about what is and isn't "moral", thanks.
Where did this come from? The parent comment simply commended the grandparent comment for not bribing a government official to push a project with poor privacy for users. I don’t see why anyone would see that as a challenge to their moral code.
The Hippocratic oath is similarly local in scope. Individual doctors try hard not to cause harm to individual patients, but the medical establishment causes massive amounts of harm, by:
- developing treatments for chronic symptoms instead of curing diseases
- being unprepared for pandemics
- making health care unaffordable except through employer plans
- promoting wrong nutrition guidelines for decades after the evidence was in
and more.
To have good outcomes you need ethics at both individual and system-wide levels.
Agreed, the vast majority of software engineering disasters can be laid at the feet of bad management, not software engineers. Let's clean house in management first, just like Deming did when he straightened out Ford.
> To say it’s on the engineer to do no harm puts them in the tenuous position of doing the job or being replaced by someone who will. That isn’t setting us up for success.
This is where a strict licensing requirement, like Canada's P. Eng, can empower the engineer. If you think what you're being asked to do would violate your professional ethics, not only can you decline to do it, but you have a system to ensure that you won't just get replaced by someone who will do it.
a licensing system is just a form of regulation as alluded to in the comment you're replying to.
And in the end, if software engineers are to conduct themselves in moral and ethical ways, they must be empowered to do so without having to sacrifice their personal wellbeing or livelihood. Regulation, it seems, is the only way to achieve that end.
Indeed it is the same with Chartered Accountants. Management can make whatever decisions they think they can make, but an accountant won't inact things that are unethical at the risk of their charter and professional standing.
You really believe many people with won't violate their oaths and associations and licenses for enough money? Accountants audited Bernie Madoff and Wirecard for years and "saw nothing". Michael Jackson hired a surgeon to pump him full of drugs for $150K a month. Money > Ethics for a lot of people.
To me this argument sounds a bit like a "only following orders" defence. I'm sure smart engineers ,I.e., all of them, can figure out what the application of their work will be.
Agreed. Let's look at it from a simple emperical standpoint: where does the problem mostly arise, and who has the power to fix it?
Organizations that went through true iterative process to reduce failure rate like NASA figured out that they needed to allow true authority to specific domain experts to blow the whistle and not face reprisal or suffer for it. Oaths fix nothing, you need organizational change, and if someone is going to do that, its the management in charge.
not to mention that software is reusable: a technology might be invented with all the good intentions and ideals... and later be used for evil purposes.
> To say it’s on the engineer to do no harm puts them in the tenuous position of doing the job or being replaced by someone who will. That isn’t setting us up for success.
If we were to introduce an oath we would have to take further inspiration from doctors, e.g having a certification required to do the job, or, failing that at least having an industry-wide union/guild protecting the position.
> Software engineers are accountable to their bosses before their users, no matter how high minded we like to pretend to be.
Isn't that what a hipppocratic oath would solve. They'd be accountable to the oath before their bosses, and that would give them reasonable grounds to refuse unethical work.
How do you guarantee that the boss will cooperate with the software engineers and tell them about all the unethical business practices that are currently happening?
Eventually it would have to be told to a CTO, president or Vice President it some technological product, software architect, etc. Software Engineers generally need to understand the use case of a product to be able to develop something effective. It would be exceptionally difficult to convince someone to code something that creates fake profiles of people who don’t use your site and no one have any idea that that’s what’s happening.
Well if it's something that the software engineers are expected to implement, then they're going to have to be told about it. If it's not, then it seems out of the realm of software engineering.
The Hippocratic oath for software engineers should come along with legislation that makes it illegal to fire a software engineer for refusing to break the oath.
Exactly. An oath means nothing: what you need is skin in the game.
Some health practitioners are literally bought by Big Pharma, by their hospital accountant, etc. How would an oath fix that? Same with engineers or any other discipline.
You need to make sure that everyone in the process has skin in the game. For me it's less about control (legislation) than about responsibility and accountability (assessments, eating your own dog food).
Having personal exposure, taking accountability for your actions. From Taleb of course.
The case of Snowden shows how bad it can turn when not everybody in the loop has skin in the game (asymmetry). Ironically, his behavior tells us he has been faithful to some kind of oath, but apparently none of his coworkers or supervisors.
>To say it’s on the engineer to do no harm puts them in the tenuous position of doing the job or being replaced by someone who will. That isn’t setting us up for success.
obviously for a hippocratic type oath to work you need the same kind of system in place for qualifying engineers that you do for doctors and not allowing anyone to work as an engineer who failed the ethics board.
I think people assume the Hippocratic oath actually makes a difference. It's culturally important to many doctors of course, but how often has that hindered the military or the CIA from hiring enough doctors to facilitate a new interrogation program? Of the illegal human experiments that happened in the US many had doctors working for them as well. It's a good guiding principle but the idea that it would actually make an impact of any kind is debateable. It doesn't matter if 90% of coders follow the oath if 10% or even 5% are enough to handle the demand for oath breaking.
This assumes all of the vast majority of engineers are easily replaceable. Given that several tech companies prefer the chance of not hiring a good fit over the chance of needing to fire someone, I think the cost to a company if a significant subset of its engineers declined unethical work (especially engineers who at this point are high enough to have a good idea of the broad scope of the software) would make it difficult to perform what you suggest.
Here's a prospective from someone who became a software dev in my late 20s after other jobs. Relatively speaking, if you can write working code and show up on time, you have a lot of leverage in getting and keeping employment, more than most other middle to upper middle class professions. This means org politics has less effect on you, and less incentive to get involved.
As a dev you can, but don't have to, think as much about the politics and operations of your org for all kinds of reasons. You are relatively harder to replace so internal politics tends to matter less, and if the org makes decisions you don't like, you can be confident that you can leave and find something else versus the long and often unfruitful process of trying to change an org from within.
Politics can and often is messy, how often have you heard something like "I just want to build things" (it's how I feel for sure), if you can get paid well to do that, why get involved with a messy decision making process?
To extend upon this (though personally I feel a little differently than yourself, but inclusive of how you feel as well):
My personal experience has seen individual contributing software developers have little voice in the matter regardless. Outside of "tech"-forward companies they are of little consequence in the larger political structure of a company.
I've seen it and been it—speak up about a desired direction, voice concerns about a decided direction, concerns about faulty legacy software, etc. Those voices, unless amplified by political clout mean little to nothing to anyone else.
In a lot of organizations, title is everything when it comes to moving discussion.
So my thinking, if we're continuing the comparison to the medical profession, is hierarchies must follow similarly.
A head of Surgery in a hospital ward is going to be a doctor. Hospital directors are going to be doctors. Sure, the CEO may not be, but they rely on the expertise of their directors. If the directors don't follow the same code as the rest of the professionals under them, then they can theoretically impose any ethics they choose and the onus falls to the IC/surgeon/etc which for the intended purposes of the oath/license/regulation at all is tenuous.
The oath should be for everybody involved in the process.
After the 2008 mortgage crisis, Netherland required everybody working at banks to take the banker's oath, which is mostly about balancing the interests of the 4 main stakeholders of the bank: shareholders, customers, employee, and society. It's pretty broad, it doesn't magically fix everything, but it does make everybody more aware of their responsibilities. Maybe software companies should require something similar, where everybody needs to be aware of their responsibilities towards, well, primarily user data, I guess. And that goes for not just software engineers themselves, but for everybody involved in the process.
Do all those businesses have a problem with unethical behaviour causing serious problems in society?
Doctors, bankers and apparently software engineers have a pretty big impact on society, and often in ways that aren't very transparent to most other people. It's quite possible there are other professions that have a similar impact, but I'm pretty sure it's not all.
I think registered accountants also have some sort of oath, again because various stakeholders including society as a whole has to be able to trust them.
There is the problem that everything can be subdivided in a bunch of menial task so every cog in the system isn't aware of their impact. It already happens a lot in software.
the idea of Hippocratic Oath reminds me of Asimov's Three Laws of Robotics in "The Naked sun" (SPOILERS ahead): the detective realises that the normally quoted First Law of Robotics ("A robot may not injure a human being or, through inaction, allow a human being to come to harm.") is actually just an approximation, he argues that the real Law is "A robot may do nothing that, TO ITS KNOWLEDGE, will harm a human being; nor, through inaction, KNOWINGLY allow a human being to come to harm."
This is important because even though robots really try their best, different robots could perform sub-tasks that look very harmless by themselves, but combined kill a human being:
- A robot is instructed to pour this bottle of poison into a caraffe of water and then leave the room
- Another robot is instructed to enter the room, take the caraffe of water and give it to a human to drink
The human is poisoned, but none of the robots are directly responsible (in the first law sense). Is the act of connecting the two dots the evil deed.
> There is the problem that everything can be subdivided in a bunch of menial task so every cog in the system isn't aware of their impact. It already happens a lot in software.
Precisely the issue with the original proposal. Would it have mattered whatsoever if PhD's took a Hippocratic oath when developing the Manhattan project?
I feel like this MSFT executive may already know that swearing engineer to "do no harm" is fruitless after reading the article, but it's still unfortunate that statements like his diverts attention from more meaningful proposals.
Kind of funny to hear this from a company known for harvesting tons and tons of telemetry from customers, with no true way to fully opt-out which is pretty damn unethical and probably feeds their AI.
> Software engineers are accountable to their bosses before their users, no matter how high minded we like to pretend to be. To say it’s on the engineer to do no harm puts them in the tenuous position of doing the job or being replaced by someone who will. That isn’t setting us up for success.
The nazi officers who committed most of the atrocities used similar arguments. "I was just following orders!"
I expect better from a software engineer on hacker news. You've single handedly convinced most here - through your weak logic - that such an oath is necessary.
A software engineer is more like a chemist working for the pharmaceutical industry than a doctor treating patients. And chemists typically don't have an Hippocratic Oath. Pharmacists sometimes have their own version, but it is mostly about giving good advise to patients and respecting them as human beings.
But it doesn't stop the pharmaceutical industry from being heavily regulated, and while their business practices are often criticized, the drugs that come out of it are generally safe and effective. Many countries also have regulations making important drugs (ex: vaccines) accessible to everyone.
Well, it sure is tough to know the outcomes of what you create, however many oaths you take. Long ago, someone invented a pencil; people have written a lot of terrible things with pencils. More recently, people invented neural networks to recognize handwritten digits; now people are using networks like that to do terrible things. An oath won't solve the problem; we need laws and governance and more to ensure society uses technology safely and for good.
> There are no common ethics codes to determine how lethal autonomous weapons and systems that are developed for the military should be used once they end up in the hands of civilians.
It's interesting to me that this just presumes developing these autonomous weapons systems in the first place is ethical. I understand there is a difference of opinion on this ethical point, but it immediately frames the discussion pretty far away from the Hippocratic oath's requirement to abstain from causing harm.
> pretty far away from the Hippocratic oath's requirement to abstain from causing harm
So does abortion and euthanasia, and probably plenty of other practices as well. Both of those are without doubt harm-causing practices, with their related points of controversy primarily revolving around whether the harm that is caused is worthwhile in the context of the alternative being a potentially greater harm.
Putting aside the fact that the Hippocratic oath is not actually a relevant part of modern medicine (modern doctors are accountable to comprehensive, codified sets of ethics), the fact that there is no such thing as a set of common ethics by which people choose to live their lives kinda points out the futility of this idea.
One person could say developing weapons is bad because they cause harm, another could say it’s good because they can be used to reduce harm that would have otherwise been caused. Who’s right? Neither of them. That’s just two people with different opinions. I would personally suggest that establishing moral authorities like can often be harmful, because lacking any objective truths, it’s a topic people should generally be left to make up their own minds about.
Am I right or wrong? Who’s to say? I’m just a person with an opinion, and so is anybody who would want to agree or disagree with me.
All 3 of those examples are actually covered by the “original” Hippocratic Oath (which probably wasn’t written by Hippocrates, incidentally).
> Neither will I administer a poison to anybody when asked to do so, nor will I suggest such a course. Similarly I will not give to a woman a pessary to cause abortion.
Chemo is obviously a bit different though, because the potential harm caused by denying abortion or euthanasia is (generally speaking) the potential to deny somebody the right to exercise a form of personal agency over their body/life. The controversy isn’t really a medical one.
The Oath also doesn’t really address treatments that have potentially harmful side effects, and it’s debated whether the oath allows doctors to perform surgery. It’s basically not fit for purpose in 2020. If you wanted to suggest that software engineers adopt a code of ethics similar to that of doctors, what you’d be really suggesting is something like “We need a AMA Code of Medical Ethics for software engineers”. Which obviously doesn’t have the same broad appeal and simplicity of an oath.
That was essentially my point. With euthanasia and abortion you will find plenty of people who would call them unambiguously harmful and think they should be banned outright, regardless of context.
You'd be hard pressed to find people who want to abolish the entire field of oncology on the grounds that the treatments are horrible.
It's absurd to take the above-stated "requirement to abstain from causing harm" as a hard restriction out of context without taking into account the main point of the profession which is to help the sick.
The logical conclusion of considering "do no harm" as inviolable above all else is that doctors would have to restrict their treatments to homeopathy and compassionate smiles.
"It should be noted that no ethically-trained software engineer would ever consent to write a DestroyBaghdad procedure. Basic professional ethics would instead require him to write a DestroyCity procedure, to which Baghdad could be given as a parameter."
I like the way you challenge the framing, and I agree, the right first question is "should we develop lethal autonomous weapons at all, and if so what kind is ok, what's the limit on that".
The way its asked looks like an attempt to shift the Overton window until autonomous weapons of all kinds are treated as a mundane inevitability not worth worrying about, with just the niggling details subject to ethical questioning.
But big shifts like that are exactly the sort of thing serious ethical codes should be used to watch out for. Not the niggling details afterwards.
Well for ine autonomous weapons were already there. Landmines for one. Back in the stone age even with snares even meaning /rope/ is an autonomous weapon.
There is no human in the loop (no pun intended for snares). It decides when to strike using physics and the answer is always "yes" if it is triggered.
What makes the new "autonomous" weapons different is that they attempt target differentiation. Mobility becomes useful then when weapons systems can say "no" when presented a target. Since even the Military Industrial Complex, purveyor of unneeded bullshit which wantonly takes lives would find it impossible to sell a drone that goes around shooting missles at all targets after launch.
The ACM Code of Ethics and Professional Conduct has been around for a while and surely is a good start. But if you read through it, you'll quickly come to the conclusion that unless leadership buys into it your only real option is to quit your job if asked to do something you shouldn't.
They don't want to alienate Chinese members so they go to great lengths to prevent the code of conduct from outright prohibiting participation in the creation of things like the systems used to control and oppress the Uyghurs, etc.
That's the whole point of a professional code of ethics. If you want to protect people for behaving ethically, that's government's job (c.f. whistleblower laws et. al.).
This sits a layer down in the defense-in-depth stack. And the idea is that if there's a recognized code, and consensus on what constitutes a violation, that employers will conform because if they don't they'll risk not just one "activist" employee leaving but most of them, out of a shared sense of communal ethics.
Would it work? No idea. My experience is that software people tend to be pretty squishy on matters of personal ethics.
I am little worried this will sound glib, but how about we have an oath/pledge ( w/e you want to call it ) for decision makers at the company. Otherwise putting onus on engineers is, at best, silly.
Translation: we need to increase barriers of entry to working in this field by spreading the social cancer that is vocational licensing to software engineering.
That’s a bad faith question asked in an attempt to setup a straw man argument and you know it, and I know it, so cut the crap. Worry about the guy writing software for all the nuclear power plants we’re not building, not the guy who makes webshit, word processors and compilers for a living.
Architects, Doctors and Lawyers are closer to one extreme, you can make as strong a case for licensing them as you can for not licensing them. Software engineers are vocationally with the florists, where licensing is less a matter of ethics or workmanship or safety as it is an artificial government sanctioned means of preventing new competition from entering the market, or only allowing it in a controlled and revocable manner.
Writers of open source software are winding up to one day spin in their graves reading the comments of the authoritarians demanding software licensing in here.
Agreed. There is no professional class so enlightened that it won’t use whatever means available, governmental or otherwise, to insulate itself from competitive pressures.
Pay a huge amount of money. And, perhaps unsurprisingly, we have an incredibly expensive healthcare system, a legal system where deeper pockets have a large advantage, and a housing crisis.
All I need to make a website is to know how to do it. To build a house I need to comply with an endless list of regulations that are essentially intended to make me unable to do so. To buy an asthma inhaler (insert medicine here) I must pay a doctor to write me a prescription. It doesn't matter whether I know what I need and how much. I must pay for it. Lawyering I can do on my own for myself though.
Edit: I'm not saying that all of this is necessarily bad, but I just couldn't resist when you mentioned those 3. They're not great examples, because the ethics there are about protecting the person receiving the service. In software development you rarely build something for a specific human individual.
This is different. Anybody can create a website and thus the next Facebook/Twitter/TikTok. As far as I know, I don't need a permit to build a website. Requiring one would stifle freedom of speech.
On the other hand you bet that if you work on missile guidance/plane software you'll need some sort of certification. Your comparison just doesn't work. And there are already laws in place that any website dealing with user data or any sort of ecommerce must follow.
Software engineering isn't medicine. It isn't that difficult to learn and you can wield many of the advantages without much training. If software folks refuse to assist with morally questionable tasks, those tasks will be compartmentalized and the unethical components will be handed off to people willing to do the job.
Sadly, just about everyone in ML/AI is contributing to an easily weaponized technology. Every time you make it easier to train a network, every time you make it faster to train a network, every time you improve an image recognition algorithm, every time you improve the latency/jitter of inference... you're contributing to the pile of knowledge which will be leveraged to control populations or enable military action. Most people stay unaware of it and just focus on the benefits. Some of us just grow to accept it.
The parent is right. Software engineering is advantaged by a much more simplistic feedback loop. Software's "Hello, World" is validated in milliseconds. Medicine's "Hello, World" could take 30 years of clinical trials to ensure that you haven't killed anyone. It is easy because it is quick, allowing a greater understanding within an equal amount of time.
In fact, the adage of years of experience comes from learning that actually take years to encounter different circumstances and see the results play out in order to fully understand what you're dealing with. This is almost never a problem for software engineers, especially in the learning phase. Software engineers can gain "years of experience" each day by seeing the results of what they are doing in practically real-time.
I think we need to go well beyond a Hippocratic oath.
Most Software Engineering Degrees require an ethics course. Yet - we still have the issues we have today.
Asking for good intentions isn't going to change outcomes, because Software Engineers already have good intentions. Nothing will change.
How about Software Engineering Licensure? You have been found to contribute glaring security issues in widespread code? Lose your license and employment. Use dark patterns to defraud people? Lose your license and employment. Leave "debugging" endpoints open or collect unminimized telemetry? Lose your license and employment.
This seems stronger, and much more likely that outcomes will change.
This doesn't generalize very well, since a lot of software is written by volunteers making open source software. Those people aren't employed, and exempt from getting fired. People who own their own software business also can't be fired. Overall, what you propose doesn't make sense.
What we need is more ethics courses and courses on historical business frauds/schemes as a graduation prerequisite for getting an MBA. For the persons leading the groups of software engineers.
I am honestly much more worried about terrible directives coming from c-suite persons, that result in ethically questionable apps and software.
In addition to ethics training and commitments from the persons doing the actual engineering work.
this is absurd. We should not be putting the onus on the engineers to practice their trade ethically when they often have no control over the decisions being made at their company. Additionally, those software engineers do not profit from unethical practice nearly as much as the decision makers and executives.
This is especially rich coming from a faboulsy wealthy executive of a company engaged in many unethical activities
I think we also should finally abandon the idea that companies are only here to make a profit and should ignore all other factors when doing business. That doesn't mean they should make no profit at all but it means that they also should consider external costs as if they were internal ones. This is true for the environment, privacy, data protection, employees, and many other things.
I think something like this would be more easily 'enforced' if the current punishments weren't so low they got factored into the "cost of doing business."
It's hard to see this as anything but a way for executives to foist responsibility upon software engineers when things go wrong (and of course, claim credit and profit when they go right) as other commenters have pointed out.
That said, this might actually work! If a software engineer can suffer personal harm by working for a business with iffy ethics, then they are incentivized to play it safe by avoiding working for those types of businesses -- thus correcting the market by internalizing the externalities. I doubt anyone would work for Facebook in a world with a Hippocratic Oath for Software Engineers that has real teeth.
Put another way: pointing to decision makers instead of individual engineers is a simple rephrasing of the Nuremberg defense, "I was just following orders!" It is obvious that we should hold leaders accountable. The question here is whether we hold individual software engineers accountable too (they're not mutually exclusive) and the answer is probably yes.
Wouldn't offshoring lower liability? If you can blame the developer why not offsource that or better outsource and remove any responsibility from the company.
543 comments
[ 1.7 ms ] story [ 265 ms ] threadGDPR is doing great at keeping creative-less* companies in check. ---------------------------------------- Creative-less because they can't have a vision and rather gather data and try to find something customers like to keep milking until the product inevitably dies.
The Hippocratic oath sounds more altruistic than the alternatives, but good legislation, including business audits and incentives, will have far more impact than a software engineer swearing they won’t be evil.
Aristotle and a while host of philosophers certainly thought so.
Extending this, why don't we expect everyone in society to behave ethically? But then we get into arguments about what is ethical, because people disagree on that, and people naturally will disagree on what is ethical in software development, which isn't always a problem. The is a diversity of opinion, though since are clearly extreme.
One of the issues with a "Hippocratic oath" for software developers is that software spans the while spectrum of human activity and thought.
Constraining software is nearly equivalent to constraining human thought in more ways than one.
Want to see an example - Ask Aristotle are women same or lower than men?
A lot of the demands for software ethics get into even fuzzier and more complicated territory, like whether the companies which control our mass communications should use that control to decide which political views people should be able to share and what they should target, whether this is good or bad for democracy, etc. Also, many of them seem to be things that would've been incredibly niche viewpoints even a decade ago. I've seen people in another thread thinking that Google promised not to use e-mail contents to sell ads originally and then snuck that in, but in reality basing ads on the contents of e-mails was part of their business model all along and it's just that no-one cared when they launched because it was so much less obnoxious than the ads on other providers. Somewhere along the way, we got this meme about big tech selling our personal information, and everyone seems to project it backwards in time onto how people felt before the meme.
Pls excuse the caps, but too many people seem to ignore this very important part of his argument here.
There's some (not a lot, but some) power in a profession just claiming to have ethics, even if there is no enforcement and change in the hearts of practitioners.
Plus, at some point, we all just have to own what we do with our lives. It's great to advocate for systemic change but doing your own small part is at least as important.
Do you want to license software developers?
But if you're working on video games, or pizza delivery, or really probably like 90% of software then no.
Failures of a company responsible for this kind of thing that lead to injuries or deaths can result in imprisonment of (at least) the CEO and possibly others. I'm not sure licensing is necessary here, that's what documented process is supposed to manage (change tracking, development process, etc.)
That is so wrong it isn't even funny. If the car was invented as powered by a Mr. Fusion the buggy whip makers going out of business would be a negative real world consequence.
Software is fundamentally different because until you run it it has no consequences, and even if you run it, it can be contained. I can write a worm and not release it on world. In that regard, it is more like engineering _plans_. I can draw up plans for a building that is designed to collapse with X number of persons inside -- in fact I can imagine either of the two assignments given as an exercise in University.
No reason to make more laws: it should be immaterial whether I chop down the Christmas tree at the local town square or program a robot to do it.
The programming of said robot isn't the bad act here, it's the act the robot actually performs.
The danger of being castigated for having written something that wasn't used is that we then get into the area of thought-crimes.
This sort of thing works in other professions like medicine because malpractice can cause doctors to lose their license. Same with civil engineers. This changes things because the choice is now quitting or possibly never being able to work in the field again.
Perhaps principal software engineers in charge of life or death software should be licensed for accountability, “engineer on record”.
I agree to a very limited extent about the hierarchical nature of a typical corporation, but I also disagree. Software engineers at a certain level of their career and with relatively uncommon skills can pick and choose what companies they want to work at. In my opinion people of good moral character and conscience need to be prepared to refuse to accept a position at companies known to engage in activities against their principles. And further need to be prepared to resign if they are asked to do something clearly unethical.
From my particular specialization in network engineering, I would never accept a role at an ISP in an environment where I had to implement something like the GFW in China, or further walled-garden/censorship of the global Internet. It's directly contradictory to my principles. I sincerely hope that the best and the brightest of my colleagues would never choose to aid and abet internet-fuckery by autocratic regimes. If people from my field look at a project and could reasonably say "Vint Cerf would be really disappointed if he saw me implementing this...", I hope they will choose to walk away.
What percent of software engineers does this statement apply to?
But ultimately there's always a software engineer involved in the creation of software - and that's not true of any of the other roles you mentioned. Since software engineers are necessary and sufficient to produce software, they should always be held responsible, and any oath should fall on engineers.
> To say it’s on the engineer to do no harm puts them in the tenuous position of doing the job or being replaced by someone who will.
Well, yes - if there were no tradeoffs there would be no point in having an oath to begin with. But there are software engineers today, including some on HN, who do things more harmful and unethical than medical malpractice, and they are personally culpable for the decision to do so - just as their replacements would be if they refused. I would also like to see laws criminalizing those individual engineers' conduct - maybe you're alluding to the same thing? - but an oath is a good start.
Eg a developer does something and society finds this unethical and punishes them. The developer's boss, the boss's boss, the boss's boss's boss etc up to the CEO all get punished in the same way. Furthermore, to avoid companies trying to shield themselves from this by putting their developers into a different company, it will apply to software that you get from someone else too.
Suddenly this doesn't sound very appealing anymore, does it?
Currently, if I write software that performs illegal actions -- let's say software that allows me to use unlicensed Adobe products -- at the request of my boss and their boss, all three of us would be legally liable.
Programmers get specs and write programs to match those.
At no point is it the programmers responsibility to talk about the moral compass of the project and where it fits into society.
An oath to do no harm? You first need to give programmers the power to decide the fate of projects on their own the way only a doctor can decide medicine or treatment.
Although it would be optimal for the management/leadership to not be pursuing unethical developments, a software engineer having the fallback of "I can't implement this in good faith" is another layer of defence (to society).
It would probably also allow for legal push back against being terminated for refusing to implementing the unethical thing.
If a civil engineer’s manager told them to design an unsafe building or bridge, they’re not going to just say, “Sure thing manager! One death trap coming right up!” It is their ethical duty to build it safely.
Software isn't a bridge and comparisons fall apart quickly.
EDIT:
Forum the original OP:
> Software engineers are accountable to their bosses before their users, no matter how high minded we like to pretend to be.
They are accountable to themselves and their own conscience before both their bosses and their users. I understand this is an uncomfortable line of thinking if your employer asks for ethically questionable project work, but I’d argue that if this is the case for you, it warrants career introspection.
So, the engineer writing a binary search, knowingly working on “Project Orbital Death Ray” or “Voter Suppression 2.1” should know better. I hope we can at least agree on that one.
The engineer writing a linked list or moving around Protobufs for their some open source toolset gets a pass because their project as they understand it is ethically neutral. BUT there will be that engineer who then takes those tools and integrates it into “Project Orbital Death Ray”. That’s maybe where accountability should begin.
Everyone’s talking about the managers taking the blame and yes they’re culpable too. But at the end of the day an actual software developer’s fingers type the code in. If that developer knows what he is working on, he needs to bear the responsibility, too.
I'm not sure why software would be any different. Bridges are complicated and made up of versatile submodules, just like software. Some other software engineer eventually designs the "bridge" and selects "beams" for the structure. If those beams fail to meet their specs, then the engineers who stood by them are at fault. If the bridge fails because the beams weren't used in accordance with their spec, or didn't have a spec at all, then the engineers who approved their use in the bridge is at fault.
I could get behind a requirement that code be reliable and fit for purpose, though very few of us have any experience with the formal methods that might get us there, and most don't want to work that way.
Let's go further into absurdity. The engineer is kidnapping the daughter of the manager and blackmailing the manager to take the bridge down. Is it ethical to force someone else to be ethical even if its only possible through unethical means? What if there is a hero saves the daughter? Will the hero be liable for the collapsed bridge?
You write a tool for let's say recognizing faces. Will it be used for login onto computer? Tracking dissidents? IDing corpses? Who knows.
What if you start as something 100% ethical. But your company pivots to unethical application?
My point here isn't to dictate what software is or isn't ethical, but to argue that if a program is unethical, its ethical implications are the responsibility of the engineer(s) who wrote it.
I mentioned this in another comment, but I'll say it again:
Irrespective of any legal/ethical concerns, yes, I would like to know! If my boss just came to me and said "build a facial recognition system" I certainly will ask how it is going to be used. Not because I care about ethics, but it's a basic aspect of the job. You can replace "facial recognition" with "CMS" and I'd still ask.
If they tell me the facial recognition is for logging into computers, and then later decide to use it to track dissidents, that is a different concern. But I'll at least ask!
> What if you start as something 100% ethical. But your company pivots to unethical application?
If they pivot after my work is done, I won't feel responsible. If they never used it for the original application and pivoted to this, I may get upset and quit, but my conscience would be clear.
So if you invented dynamite you wouldn't feel responsible for its use?
But, let's change it a bit more personal. You write an awesome OSS yaml parser. It's so good, that GFW of China uses it as a main component, and this gets published in the news.
What would you do? Nothing you did changed, but suddenly your work is powering an unethical component.
It is the same sort of stupid blame shifting involved with the hippocratic oath for x nonsense. Oaths are majorly outmoded in the zeitgeist anyway because everyone recognizes lies are commonplace.
And I fully agree. Expecting people to individually bear the burden of "some oath", is a fool's errand.
My point was software on its own, much like a fridge, is amoral. You can use it to store your groceries, or you can use it to store corpses.
That said, there are some extreme cases (like a gun), that have very limited non-violent uses. And IMO, that should be regulated, instead of depending on people Doing The Right Thing™.
I would if I were inventing dynamite, but that's not what this scenario is.
A person working for a knife manufacturer need not worry about it being used for murder, as that's not what the primary use. And facial recognition is a lot less harmful than even that.
Trust me: I work for a company that produces certain goods used for all kinds of good and nefarious purposes depending on who buys it. My conscience is clear.
> But, let's change it a bit more personal. You write an awesome OSS yaml parser. It's so good, that GFW of China uses it as a main component, and this gets published in the news.
> What would you do? Nothing you did changed, but suddenly your work is powering an unethical component.
I wouldn't do anything:
1. This is milder than the knife scenario above. Of course I don't care if people use it in a poor way - unless there is a straightforward technical mitigation I could do. In your example, given that the source code is available, that is not an option.
2. There's a certain hypocrisy in releasing something as open source and then complaining about how it is used. If it bothers you, then modify your license!
> I wouldn't do anything:
That's hypocritical. Nobel didn't invent dynamite because he wanted people to blow themselves up. He invented dynamite because nitro-glycerin was a horrid mess used in mining.
He definitely didn't have an easy technical solution to problem of people misusing dynamite.
You can either say in both case do nothing, or in both case do something.
If I lived in a world lacking in nuance, I would agree. I do not live in that world.
Right now bosses don’t even have incentive to lie about it because no engineer is obligated to give a shit about the society they live in broadly.
Any oath would either not be taken by those people, would be watered down so far as to be meaningless, or would require the entire industry to refuse to make weaponry. The first and second are ineffective and the third is ludicrous.
Btw I have tried this, and I was just replaced by someone who would just follow orders. Then I get to come in after and clean up the mess ..
(Remember employees in the US depend on the company for health insurance. Saying 'no' could cost a lot more than just ones position.)
Most software engineers are not like doctors. We have little autonomy over what is created. Our responsibility is primarily the how. And with devops sometimes the actual deployment and maintenance itself.
Consider something like the 737 MAX debacle – did the programmers writing the MCAS code actually have enough aviation domain knowledge and understanding of where the component fit in the overall system to realise it was a threat to people's lives?
I don't know, but my guess is the most likely answer is "No".
For this to work, software engineer compensation would need to change radically.
For the same reason, it's harder to hire some rando budget doctor because the field is gate-keeped by the requirement of a licence, and liability.
You can't magically make Engineering the same without the same conditions. Add barriers top entry that see my pay rise, or have cheap programmers with no liability.
I’ve never been on a project that requires a software product stamped by a licensed engineer. NCEES dropped the software license because there was so little demand, compared to, say, civil engineers who consider a license a rite of passage to career growth.
1) you work for the federal government
2) you work under a licensed engineer
3) you work under an industrial exemption
There’s differences depending on state. There has been a more concerted effort to remove #3 recently due to both political reasons and the technical issues in this thread. Most people performing engineering work under an industrial exemption don’t actually realize it. Again, this is different state to state. For example, in some states you can’t start a business with “engineering” in the name unless you have a certain percentage of owners/principals with an engineering license.[1]
What actually happens is conflating terms in common parlance. “Engineer” and “engineer” are not necessarily the same. For example, a computer engineer may work under an industrial exemption (due to working in a manufacturing service) while a software engineer does not. Legally, an “Engineer” claims an explicit responsibility to public safety.[2] Apropos to the headline article, there is a distinction with this difference.
[1] https://fxbinc.com/wp-content/uploads/2016/10/state-by-state...
[2] https://www.theatlantic.com/technology/archive/2015/11/progr...
A software developer refusing a job because it does not meet his ethical parameters is just an unemployed software engineer.
There are definitely areas where the prudent thing for a developer to do is raise a dissenting opinion, if not halting work. What seems lacking is clear industry consensus standards to back up that decision.
The reason MCAS came about was because management wanted to try to fudge a larger engine into an outdated design created for a different purpose rather than do the engineering and certification necessary for the new requirements and to update the system.
Management wanted to save money. Of course the engineering leadership did not want to fudge something -- they wanted to do proper engineering. But the people in charge just wanted to save money, and the engineering leadership could not do anything otherwise, even they knew that just making the engine larger and compensating did not make sense from an engineering standpoint.
By the time it got to the MCAS, that was far down the line of the decision to not do proper engineering.
Blaming it on management is also irrelevant, since management merely takes the advice of engineers, and do financial/business trade offs to maximize profit. If the engineers cannot tell that MCAS system could fail this way (due to the complexity), management will not question it.
Its very common.
if the engineering leadership said, this is not a good idea. Management said, it saves money.
Engineering failed to convince management. Management didn't have the understanding that it was a bad idea.
It is now, no one's fault?
If management merely takes the advice of engineers ( and other who specialize in the things that they do not ), and they choose to ignore it because they do not understand the things they do not specialize in. I believe it's a reasonable to assume that management is more at fault than engineering ( I'm not sure they're is a situation here where any party is fault less )
Management likes hearing things they like, and simply don't hear things they don't like. Then act surprised about it when it becomes public.
By the rest of your comment, it looks as though it's an excellent comparison. How would engineers taking an oath help the situation?
The MCAS is an optional component that reduces certification and training costs. It is definitively possible to fly the plane without accidents even with a disabled MCAS. So why can't the MCAS be turned off automatically when sensors fail? Because that changes the classification of the plane and therefore requires pilots to be certified for a new machine and receive new flight training for both MCAS and no MCAS modes.
If the software engineer was under a hippocratic oath then he would have to refuse to build the MCAS entirely but not because the idea of an MCAS is inherently unethical, no, he would have to refuse because the company he works at wants to use the MCAS for a non ethical purpose (namely operate and hide the existence of MCAS even when it is unsafe to do so).
This is basically a reverse audit but the software engineer has no authority conduct such an audit and even if he was allowed to, the business has no obligation to give him the necessary information to determine whether the MCAS will be used unethically.
You think a programmer, handed a spec and asked to implement it, can be expected to know that their employer (or the employer's customer) wants to use it for a "non ethical purpose"?
Again, I can't know for sure, but I doubt the programmers who wrote MCAS (who most likely didn't even work for Boeing, but rather some subcontractor) actually knew, or could have known, how the code fit into Boeing's larger purposes
Most types of development in these large companies is so compartmentalized that it's next to impossible to see the whole structure from a software engineers prospective. You need to be at a management level to understand how most of the pieces really come together, which is the only place where one of these "oaths" might make an influence. At that point, however, the selection is so goal oriented, I have a doubt as to whether or not people would take that oath.
Of course, there’s all kinds of pressures that make these fall through the cracks. I vaguely remember an article stating some of these documents in the case of MCAS were not up to date
No imtringued does not.
imtringued wrote:
> This is basically a reverse audit but the software engineer has no authority conduct such an audit and even if he was allowed to, the business has no obligation to give him the necessary information to determine whether the MCAS will be used unethically.
imtringued is saying that it would be impossible for a software engineer to determine whether what they were asked to do was ethical or not.
The HA listed MCAS as "hazardous" rather than "catastrophic". Meaning those in charge of that process document did not realize MCAS had the ability to down the airplane. I know it's tempting to arm-chair quarterback this, but let's assume they should have realized this hazard.
To your point, maybe the programmer doesn't have the systems knowledge to make those calls, but the process is predicated on somebody having both the technical acumen and the responsibility) for those decisions. This process broke down though.
This isn’t exactly true. There are mitigations (both software and non-software) that are expected to be done depending on hazard analysis. One of the items discovered is Boeing mischaracterized the MCAS hazard (it should have had a “catastrophic” hazard class). In addition, they didn’t appear to follow their own process for dual inputs required even for the lower severity class assigned. The “optional” part of MCAS was the secondary sensor reading into the software
No. The MCAS was a "necessary" component for pitch stability - without it, a 737 MAX in a pitch-up attitude would, in the absence of correcting inputs, pitch up further and further until it stalled. Without it, the airframe is uncertifiable, full stop.
I'm certain that's not correct, everything I've read on it has said MCAS was specifically a software modifier put in place to allow the plane to respond substantially the same as a regular 737 without the larger engines, in order to avoid having to have additional training for all 737 pilots worldwide.
Most aircraft, in a "pitch up attitude" will increase their angle of attack as thrust is applied. The issue was that the MAX would do so in a more radical way than the regular 737 did, and so the software was put in place to limit that so it flew like a regular 737 as far as the pilots could see.
Conceptually, MCAS wasn't a bad idea. The execution and using it as a replacement for training and not informing pilots of the flight characteristics changes between the models was stupid.
Although to be fair my summary wasn't entirely accurate - it wasn't that a MAX was outright dynamically unstable with no control input, as I described, but rather not sufficiently stable as to cause a monotonic increase in stick force as AOA increases, which can cause the combined system of pilot + flight dynamics to be unstable since the pilot relies on stick force as an indicator.
> Most aircraft, in a "pitch up attitude" will increase their angle of attack as thrust is applied
This is both incorrect and irrelevant. Most aircraft will climb when power is applied, but will not change their AOA unless the thrust axis is off-center. To a first approximation, power controls climb rate, and stick input changes AOA. Change in behavior under different power settings has little to do with the problem with the MAX. The problem with the MAX is that at high angles of attack - i.e., when the stick is held back, causing the air to meet the wing (and the engines) at a steeper angle - the engines, which are flung forward, start producing lift of their own and produce a pitch-up moment. This means that the further the pilot pulls the stick back, the less hard they have to pull. This is a dangerous inversion that increases the control order of the system, as it breaks the usual assumption that a given stick force will result in a given AOA, more or less.
The original 737 does exactly the same thing the Max does with respect to producing a pitch-up moment -- as does nearly every other aircraft. It's just not nearly as pronounced as the Max is.
>The Boeing 737 MAX MCAS system is there ONLY to meet the FAA longitudinal stability requirements as specified in FAR Section 25.173, and in particular part (c) which mandates "stick force vs speed curve:, and also FAR Section 25.203 — "Stall characteristics".
One, it's not going to be clear from the request that the MCAS would want to be used in unethical ways.
Since the Hippocratic oath is the argument here, how many software developers want to work in a system closer to physicians? A national cartel controlling membership and licensure - tough luck if you want to hire more developers because there's an artificially limited supply. Mandatory academic training - goodbye self-taught developers. Follow-on training with pay 1/5th or less of your attending physicians - I know residents in specialties where attendings are paid $500k a year to start, and they are making $60k a year. Brutal shift work - residents work 70-80 hours a week easily. Toxic leadership - I've heard horror stories of residents being forced to lie on ACGME forms regarding their hours under penalty of being outright fired from their residency slot, which would make it nearly impossible to get a job as a physician (mainly because you'd have to apply to a different residency program and explain your termination).
I know they're not suggesting bringing the entire medical education & training structure over to tech workers, and everyone here likes to think that they're brilliant and changing lives every day but most of us are just throwing shitcode JS into a computer for 3-4 hours a day for an ad tech company and not much more. The comparison falls apart pretty quickly.
Edit to add: If you are being asked to perform illegal or unethical acts as part of your employment, then perhaps termination is an ideal course of action? Unless of course your personal enrichment outweighs legalities or ethics in your worldview?
All these "companies take on a life of their own" arguments sound a lot like executives priming the pump of potential jurors with excuses. If decision makers cannot bear responsibility because of a company size or organizational structure then we can make some sizes and structures illegal before they stumble/march into devastating incompetence.
I should phrase it differently. Why is an absolute freedom of association more important then the freedom from being harmed by large associations with amoral machinations. The original argument asks that if large corporations inherently obscure moral outcomes, maybe they are immoral, which is an argument that puts these two moral axioms in conflict. Simply stating that one side wins is thought terminating; its important to argue for why its better.
The only argument that actually matters here is whether or not restrictions on corporate structure actually do violate freedom of association or not.
I'm reasonably skeptical that they do, given that the 1st Amendment hasn't stopped us from enforcing antitrust and monopoly legislation in the past. Yeah yeah, Citizens United and all that, but we regulate companies all the time.
But I'd still want an actual lawyer to weigh in on that, I wouldn't feel confident saying that there aren't limits on how far we can go in that direction.
> I'm reasonably skeptical that they do, given that the 1st Amendment hasn't stopped us from enforcing antitrust and monopoly legislation in the past. Yeah yeah, Citizens United and all that, but we regulate companies all the time.
> But I'd still want an actual lawyer to weigh in on that, I wouldn't feel confident saying that there aren't limits on how far we can go in that direction.
It doesn't necessarily hold that because one thing is legal, everything is legal. For example, we have 1st Amendment restrictions on threats and libel, but in the US hate speech is still protected speech. 1st Amendment exceptions are generally pretty narrow and specific in the US.
In the same way, clearly some corporate regulation is OK. It does not follow that there's literally no limit on what the government can dictate about how a company can operate. I would prefer to get input from a lawyer before asserting that so confidently.
Structures can and should be changed in this case. But shouldn't be outlawed.
Was with you until this part. Just hold them personally liable if someone gets hurt should they create an uncontrollable system and predictably fail to control it.
Keep in mind the US already has laws around corporate structures and conflicts of interest. (Even if they're selectively applied.)
We heard the "too complex to understand" excuse a lot regarding the pricing of subprime debt. Except a lot of people did understand it was a problem. It's basically the "I'm too stupid to know what I was doing" defense. If we accept that defense and try to make regulation to protect them from failing (as was done in finance back then), we basically allow stupid people to continue to be in charge rather than being replaced as they need to.
No, I wouldn't say that. In many cases, management and engineering share the blame jointly and severally since they both have an opportunity to stop it.
> Said another way, engineers now need to be technical and legal experts in the business domain?
Engineers should know enough about their business domains to understand the ethical impacts of their work. Ethics and law are orthogonal, so thankfully this is generally much easier than being a legal expert.
> (Remember employees in the US depend on the company for health insurance. Saying 'no' could cost a lot more than just ones position.)
Thankfully, the healthcare safety net in much of the US is far better than it gets credit for, and the pay and availability of opportunities for software engineers in the US has generally been quite good. I'm sympathetic to this argument in general, which is one reason I don't think there should be an oath for, like, Amazon warehouse workers, but I'm far less sympathetic for anyone making 5+ times the median US income.
> Most software engineers are not like doctors. We have little autonomy over what is created. Our responsibility is primarily the how. And with devops sometimes the actual deployment and maintenance itself.
Your responsibility as framed to you by the business is the how, but upstream of the how is the question of whether or not to do it at all. If you contribute to a piece of software, you've tacitly answered yes to that question.
This would be more reasonable in the era of 10-20 years in the same company or industry. Needing to job hop every 2-3 years for a decent raise, and software skills applying to a vast array of industries, makes it less reasonable IMO.
> ...but I'm far less sympathetic for anyone making 5+ times the median US income.
Not everyone here or in software makes that kind of money. Some of us in the Midwest--or who aren't as skilled at negotiating--don't pull down nearly that much.
Yeah, to be clear, I don't think software engineers have an infinite level of responsibility for understanding the ethical implications of their work. If you were a software engineer at a credit rating agency in 2006, and you didn't see the ethical dilemma because you didn't anticipate that contagion would be exacerbated by the shadow banking system to bring down the global economy, you get a pass. But if your prospective employer is, like, locking children in cages, or spreading disinformation on political candidates, you should probably find that out during the interview process.
> Not everyone here or in software makes that kind of money. Some of us in the Midwest--or who aren't as skilled at negotiating--don't pull down nearly that much.
Good point - I'm also in the Midwest and make less than that, for what it's worth. I've naturally had FAANG in mind as I type these comments, and more generally I think salaries for the more unethical roles tend to skew higher.
Yep. That's more the responsibility of product managers, upper managements and chief architects/engineers.
I suspect a lot of philosophers would disagree with you that ethics is much simpler than interpreting laws.
Ethics forms basis on which law is built.
And, given that, it is not simpler to make ethical decision, it is harder. The decision should be worth being basis for a law, how's that simpler than following law?
Even if you take definition of ethical decision from Wikipedia: "An ethical decision is one that engenders trust, and thus indicates responsibility, fairness and caring to an individual." These words can bear negative connotations - if I beat people, I should be trusted that I will beat people, I should beat people fairly and I should care to beat an individual thoroughly.
Yes, I fully see you talk about principles. It can be seen that it is easier to make decisions from principles. But, you can misguide yourself about application of these principles.
This is a red herring. The oath is not needed merely for illegal work. In fact, the more common use cases will likely be legal. It's a common sentiment, but: Don't conflate ethics with legal.
> Most software engineers are not like doctors. We have little autonomy over what is created. Our responsibility is primarily the how. And with devops sometimes the actual deployment and maintenance itself.
This is not a dichotomy - there can be a spectrum. You can restrict it to those who do know what the product is used for, or at least have good guesses for them.
And while not everyone is this way, I wouldn't really want to work in a job for long if I'm not told what the code I'm writing is for. It's not even an ethical concern for me - it just makes for a boring job. Ideally I want people to tell me the problem they are solving and give me some leeway in crafting a solution. Don't come to me with a solution and ask me to implement it.
But money dude.
If you're an actual engineer of any kind, you always have some choice on this. You make architectural decisions every day, and you generally work for places that do, in fact, take your input into consideration. If you don't work for any of those kinds of places, then you are still responsible because you wrote the code to enable it. You can always say "no". There are consequences for that, for sure. You can always quit as well. And it may still get made. But it won't be by you.
And sometimes, that's still better than the alternative.
On the contrary, a doctor's desision affects the life of an individual patient in a very clear and understandable manner.
Illegal stuff is illegal because not knowing the law is no excuse. For your own safety and good, you better have some grasp on legal stuff in your domain. Don't have to be an expert.
I believe that this is a red herring.
Yes, the drone operator that pushes the button is a murderer, even if it was the POTUS that gave the order for it to happen.
Just not the minority of violent ones.
Also, the Hippocratic Oath is not terribly complicated. It basically says, I will not furtively and maliciously hurt people in an abuse of my authority, and I will try to heal them when they are sick. I don't think it's a lot to ask that software engineers to agree not to create knowingly malicious software. It actually addresses exactly the problem you describe. If everyone has taken this oath, and adheres to it, there is no "someone else" to do that evil work.
Last point, there is a wide variety of software engineering work out there. Some of it may be mindless of the bigger picture of what is actually happening, but for any sufficiently advanced behavior to emerge out of a complex software product, some engineer at some level has to have some idea of the path they are going down to create or allow that behavior. And every engineer has the ultimate autonomy over how and what is created because it is our hands on the code. If you don't understand that, you don't understand the power of the profession.
As someone who works in safety critical code nothing irks me more than when people absolve themselves by saying “I’m just a programmer, that’s not my job/problem”. We need to hold ourselves to a higher professional standard
- lock picking was for fun and didnt gain info - some warehouse workers didnt know about critical mass of uranium and stored the stuff to closely together
Until then, management falls on the sword, thanks.
That's not true in any way. Lots of software is written by people who don't even have a degree, others by some who have a computer science degree, but not an engineering one, etc.
The other issue is that software is rarely unethical. The unethical bit often comes from the way it is used.
And I'd have to agree with OP. In an idealist world you could assume software engineers would all be ready to quit their job at any sight of unethical affair, even say, launching something to production with a known vulnerability, or without anything but the most rigorous security review process having passed. But in practice, you're not going to achieve this result, unless you put a framework to incentivize software engineers towards being ethical. If you allowed them to sue their employer, and made it that they more often win the lawsuit, for asking them to build something unethical, or insisting that they do so even after the SE said it was unethical, or to retaliate in any way to an SE refusing to build something on ground of ethics, then you'd maybe start to see results. Otherwise, won't happen, and you've only created a scape goat to make it even easier for companies to push for unethical software, since they can now just blame SE they coerce into building it anyways.
Aside from that, the quantification of attributes/properties of people can have negative implications for many people. Oversharing is a problem on the net, but at least here people just endanger themselves.
I'm having a hard time trying to find examples of this, outside the field of armament development.
And in those fields where a software failure may result in death, e.g. aircraft development, proof of a software engineer willingly causing it, would likely result in jail time already.
1) New regulation forces some branches of software engineering to have some type of oath.
2) Now some software jobs will require only oath takers to do.
3) A well payed and powerful new cast of software engineers is born.
4) They are highly paid and have a powerful lobby working for them.
5) The oath takers become very frisky and only work on jobs with minimal risk. The ones that do screw up have an armada of lawyers, because of course they have a new association with deep pockets.
6) Innovation stalls for a while.
7) Big corps start outsourcing some of the oath-taking jobs. These engineers are not bound by the same regulation. Screw ups happen, people die at some point.
8) Maybe we should have the outsourced engineers also take an oath? Back to square 1
This is exactly what I found happened for medical Doctors in Canada (don't know for US). Not saying doctors are not doing a good job, and I can't imagine the stress and pressure they operate from. But suing for malpractice in Canada can be challenging to say the least. I have personal account of a Family member who was grossly mistreated, and all the Doctor did was changed hospitals, nothing more than a slap on the wrist.
https://diamondlaw.ca/blog/how-canadian-law-discourages-pati...
Nah, it's not the same at all. The fundamental difference between creating a program and medicine is creating a program only has to be done once, or at least a only by a few.
Medicine on the other hand: it has to be redone with each new patient. If the Hippocratic Oath works to prevent 99.9% doctors from doing a harmful procedure then you've hit a home run. Sure, you will never completely stop some bad egg removing a perfectly good limb because a patient suffering from Xenomelia offered enough money. But who wouldn't call a thousand fold reduction a huge win.
We demonstrably have the 0.1% of programmers who are willing to break any oath. They make malware, and willingly take out Sony as mercenaries because Kim Jong-Un got pissed off at a movie. All that 0.1% has to do is write the program once. Thereafter you are not trying to discourage hoards of high skilled professional from doing it again, you are trying to stop a legion of dark net operators copying the thing and selling it to anyone. An oath is a waste of time under those circumstances.
A system of ethics within the health system are necessary for customers to retain trust in the health industry. It's also strongly aligned with the selfish interests of workers who must enact that system of ethics. These properties do not neatly translate to software engineering—mostly because the most difficult ethical dilemmas in technology are rarely obvious when looking at source code. The problems with Facebook (for example) are not always inherent in code; many are only revealed after being deployed at scale and external groups begin exploiting the system.
I've quit jobs in the past because of ethical concerns about the way in which those above me have been acting. In one case this involved bribery of senior government officials to push through a project that put at risk the privacy of hundreds of thousands of people.
If you go along with shit like that, you're an accomplice and share partial responsibility. As professionals we have a responsibility to stand up for what is right. It's not good enough to fall back to the lazy excuse of "just doing my job".
However the same argument could be applied to labor abuses in the textile/garment and shoe manufacturing industry. Most people who follow news are probably aware that about 20 years ago Nike, Adidas and other brands went through a period of terrible public relations disasters, after the working conditions in some of their shoe factories in developing nations were exposed by journalists.
The argument that could have been made at that time would also have been "well but if we don't employ these people, somebody else will just do the same thing with equally terrible labor/human rights violations somewhere else in the world, with even cheaper factories".
The situation today is not great, but it is significantly improved from how it was twenty years ago. There are third party neutral inspection/oversight agencies. Companies in the garment industry are forced to make public commitments to labor rights and reasonable working conditions, and to allow external auditing. They can't just hand wave away the problem and say "but if we don't do it , someone else will.."
But that's a thoroughly discredited argument.
https://news.ycombinator.com/item?id=21416115
https://news.ycombinator.com/item?id=21923548
A shoe is a physical artifact. It must be made somewhere out of something. Software is much more flexible. If I tell you to build an 'ethical' piece of software with an extensible API, it will only take a tiny amount of work to make it do something unethical.
And much like the shoe companies abusing labour's, that had nothing to do with the engineers that made a a product line, but the management that is driven to ever lower costs.
I don't want other people making decisions on morality for me. There are people who don't believe that software developers should develop software for the U.S. government because they don't like the way immigration law is being enforced. I'll make my own decisions about what is and isn't "moral", thanks.
There is no point wasting energy and time around such people if they don't share the same values.
It's not complicated (requires some networking) to find in any org, the characters who will "do whatever it takes".
Then getting them kicked out, opposing them, sidelining them, subverting them, avoiding them are all choices every Engineer has.
- developing treatments for chronic symptoms instead of curing diseases
- being unprepared for pandemics
- making health care unaffordable except through employer plans
- promoting wrong nutrition guidelines for decades after the evidence was in
and more.
To have good outcomes you need ethics at both individual and system-wide levels.
https://en.m.wikipedia.org/wiki/W._Edwards_Deming
Software engineers are accountable to themselves before they are accountable to their bosses.
This is where a strict licensing requirement, like Canada's P. Eng, can empower the engineer. If you think what you're being asked to do would violate your professional ethics, not only can you decline to do it, but you have a system to ensure that you won't just get replaced by someone who will do it.
And in the end, if software engineers are to conduct themselves in moral and ethical ways, they must be empowered to do so without having to sacrifice their personal wellbeing or livelihood. Regulation, it seems, is the only way to achieve that end.
Your comment about auditors is such a common misunderstanding that it has a name, the 'expectation gap'. Auditors are not there to detect fraud.
Organizations that went through true iterative process to reduce failure rate like NASA figured out that they needed to allow true authority to specific domain experts to blow the whistle and not face reprisal or suffer for it. Oaths fix nothing, you need organizational change, and if someone is going to do that, its the management in charge.
If we were to introduce an oath we would have to take further inspiration from doctors, e.g having a certification required to do the job, or, failing that at least having an industry-wide union/guild protecting the position.
Isn't that what a hipppocratic oath would solve. They'd be accountable to the oath before their bosses, and that would give them reasonable grounds to refuse unethical work.
Some health practitioners are literally bought by Big Pharma, by their hospital accountant, etc. How would an oath fix that? Same with engineers or any other discipline.
You need to make sure that everyone in the process has skin in the game. For me it's less about control (legislation) than about responsibility and accountability (assessments, eating your own dog food).
The case of Snowden shows how bad it can turn when not everybody in the loop has skin in the game (asymmetry). Ironically, his behavior tells us he has been faithful to some kind of oath, but apparently none of his coworkers or supervisors.
obviously for a hippocratic type oath to work you need the same kind of system in place for qualifying engineers that you do for doctors and not allowing anyone to work as an engineer who failed the ethics board.
They are responsible to their bosses as well as the public who are the end users of their designs/products
This was not always the case.
How the hell did we let that happened?
As a dev you can, but don't have to, think as much about the politics and operations of your org for all kinds of reasons. You are relatively harder to replace so internal politics tends to matter less, and if the org makes decisions you don't like, you can be confident that you can leave and find something else versus the long and often unfruitful process of trying to change an org from within.
Politics can and often is messy, how often have you heard something like "I just want to build things" (it's how I feel for sure), if you can get paid well to do that, why get involved with a messy decision making process?
My personal experience has seen individual contributing software developers have little voice in the matter regardless. Outside of "tech"-forward companies they are of little consequence in the larger political structure of a company.
I've seen it and been it—speak up about a desired direction, voice concerns about a decided direction, concerns about faulty legacy software, etc. Those voices, unless amplified by political clout mean little to nothing to anyone else.
In a lot of organizations, title is everything when it comes to moving discussion.
So my thinking, if we're continuing the comparison to the medical profession, is hierarchies must follow similarly.
A head of Surgery in a hospital ward is going to be a doctor. Hospital directors are going to be doctors. Sure, the CEO may not be, but they rely on the expertise of their directors. If the directors don't follow the same code as the rest of the professionals under them, then they can theoretically impose any ethics they choose and the onus falls to the IC/surgeon/etc which for the intended purposes of the oath/license/regulation at all is tenuous.
After the 2008 mortgage crisis, Netherland required everybody working at banks to take the banker's oath, which is mostly about balancing the interests of the 4 main stakeholders of the bank: shareholders, customers, employee, and society. It's pretty broad, it doesn't magically fix everything, but it does make everybody more aware of their responsibilities. Maybe software companies should require something similar, where everybody needs to be aware of their responsibilities towards, well, primarily user data, I guess. And that goes for not just software engineers themselves, but for everybody involved in the process.
It seems to me that if everyone has to promise to do no evil, the meaning of such oaths would become diluted.
Doctors, bankers and apparently software engineers have a pretty big impact on society, and often in ways that aren't very transparent to most other people. It's quite possible there are other professions that have a similar impact, but I'm pretty sure it's not all.
I think registered accountants also have some sort of oath, again because various stakeholders including society as a whole has to be able to trust them.
the idea of Hippocratic Oath reminds me of Asimov's Three Laws of Robotics in "The Naked sun" (SPOILERS ahead): the detective realises that the normally quoted First Law of Robotics ("A robot may not injure a human being or, through inaction, allow a human being to come to harm.") is actually just an approximation, he argues that the real Law is "A robot may do nothing that, TO ITS KNOWLEDGE, will harm a human being; nor, through inaction, KNOWINGLY allow a human being to come to harm."
This is important because even though robots really try their best, different robots could perform sub-tasks that look very harmless by themselves, but combined kill a human being:
- A robot is instructed to pour this bottle of poison into a caraffe of water and then leave the room
- Another robot is instructed to enter the room, take the caraffe of water and give it to a human to drink
The human is poisoned, but none of the robots are directly responsible (in the first law sense). Is the act of connecting the two dots the evil deed.
Precisely the issue with the original proposal. Would it have mattered whatsoever if PhD's took a Hippocratic oath when developing the Manhattan project?
I feel like this MSFT executive may already know that swearing engineer to "do no harm" is fruitless after reading the article, but it's still unfortunate that statements like his diverts attention from more meaningful proposals.
The nazi officers who committed most of the atrocities used similar arguments. "I was just following orders!"
I expect better from a software engineer on hacker news. You've single handedly convinced most here - through your weak logic - that such an oath is necessary.
A software engineer is more like a chemist working for the pharmaceutical industry than a doctor treating patients. And chemists typically don't have an Hippocratic Oath. Pharmacists sometimes have their own version, but it is mostly about giving good advise to patients and respecting them as human beings.
But it doesn't stop the pharmaceutical industry from being heavily regulated, and while their business practices are often criticized, the drugs that come out of it are generally safe and effective. Many countries also have regulations making important drugs (ex: vaccines) accessible to everyone.
It's interesting to me that this just presumes developing these autonomous weapons systems in the first place is ethical. I understand there is a difference of opinion on this ethical point, but it immediately frames the discussion pretty far away from the Hippocratic oath's requirement to abstain from causing harm.
So does abortion and euthanasia, and probably plenty of other practices as well. Both of those are without doubt harm-causing practices, with their related points of controversy primarily revolving around whether the harm that is caused is worthwhile in the context of the alternative being a potentially greater harm.
Putting aside the fact that the Hippocratic oath is not actually a relevant part of modern medicine (modern doctors are accountable to comprehensive, codified sets of ethics), the fact that there is no such thing as a set of common ethics by which people choose to live their lives kinda points out the futility of this idea.
One person could say developing weapons is bad because they cause harm, another could say it’s good because they can be used to reduce harm that would have otherwise been caused. Who’s right? Neither of them. That’s just two people with different opinions. I would personally suggest that establishing moral authorities like can often be harmful, because lacking any objective truths, it’s a topic people should generally be left to make up their own minds about.
Am I right or wrong? Who’s to say? I’m just a person with an opinion, and so is anybody who would want to agree or disagree with me.
I think the main motivator in almost all of those things is money.
The reason for wars is money, they just get justified by "the greater good".
Same for all the involved technologies.
A less controversial example would be something like chemotherapy. In fact, a lot of treatments for terminal and chronic ailments are pretty harmful.
> Neither will I administer a poison to anybody when asked to do so, nor will I suggest such a course. Similarly I will not give to a woman a pessary to cause abortion.
Chemo is obviously a bit different though, because the potential harm caused by denying abortion or euthanasia is (generally speaking) the potential to deny somebody the right to exercise a form of personal agency over their body/life. The controversy isn’t really a medical one.
The Oath also doesn’t really address treatments that have potentially harmful side effects, and it’s debated whether the oath allows doctors to perform surgery. It’s basically not fit for purpose in 2020. If you wanted to suggest that software engineers adopt a code of ethics similar to that of doctors, what you’d be really suggesting is something like “We need a AMA Code of Medical Ethics for software engineers”. Which obviously doesn’t have the same broad appeal and simplicity of an oath.
You'd be hard pressed to find people who want to abolish the entire field of oncology on the grounds that the treatments are horrible.
It's absurd to take the above-stated "requirement to abstain from causing harm" as a hard restriction out of context without taking into account the main point of the profession which is to help the sick.
The logical conclusion of considering "do no harm" as inviolable above all else is that doctors would have to restrict their treatments to homeopathy and compassionate smiles.
- Borenstein
The way its asked looks like an attempt to shift the Overton window until autonomous weapons of all kinds are treated as a mundane inevitability not worth worrying about, with just the niggling details subject to ethical questioning.
But big shifts like that are exactly the sort of thing serious ethical codes should be used to watch out for. Not the niggling details afterwards.
There is no human in the loop (no pun intended for snares). It decides when to strike using physics and the answer is always "yes" if it is triggered.
What makes the new "autonomous" weapons different is that they attempt target differentiation. Mobility becomes useful then when weapons systems can say "no" when presented a target. Since even the Military Industrial Complex, purveyor of unneeded bullshit which wantonly takes lives would find it impossible to sell a drone that goes around shooting missles at all targets after launch.
https://www.acm.org/code-of-ethics
This sits a layer down in the defense-in-depth stack. And the idea is that if there's a recognized code, and consensus on what constitutes a violation, that employers will conform because if they don't they'll risk not just one "activist" employee leaving but most of them, out of a shared sense of communal ethics.
Would it work? No idea. My experience is that software people tend to be pretty squishy on matters of personal ethics.
https://www.youtube.com/watch?v=nfRUQh_EHoQ
Architects, Doctors and Lawyers are closer to one extreme, you can make as strong a case for licensing them as you can for not licensing them. Software engineers are vocationally with the florists, where licensing is less a matter of ethics or workmanship or safety as it is an artificial government sanctioned means of preventing new competition from entering the market, or only allowing it in a controlled and revocable manner.
All I need to make a website is to know how to do it. To build a house I need to comply with an endless list of regulations that are essentially intended to make me unable to do so. To buy an asthma inhaler (insert medicine here) I must pay a doctor to write me a prescription. It doesn't matter whether I know what I need and how much. I must pay for it. Lawyering I can do on my own for myself though.
Edit: I'm not saying that all of this is necessarily bad, but I just couldn't resist when you mentioned those 3. They're not great examples, because the ethics there are about protecting the person receiving the service. In software development you rarely build something for a specific human individual.
On the other hand you bet that if you work on missile guidance/plane software you'll need some sort of certification. Your comparison just doesn't work. And there are already laws in place that any website dealing with user data or any sort of ecommerce must follow.
Depends where you live and how much local architects have been regulatory captured the housing market. Other than that, building a house is easy.
Sadly, just about everyone in ML/AI is contributing to an easily weaponized technology. Every time you make it easier to train a network, every time you make it faster to train a network, every time you improve an image recognition algorithm, every time you improve the latency/jitter of inference... you're contributing to the pile of knowledge which will be leveraged to control populations or enable military action. Most people stay unaware of it and just focus on the benefits. Some of us just grow to accept it.
I would never hire someone that treats software engineering as something easier that medicine or other branches of engineering.
The parent is right. Software engineering is advantaged by a much more simplistic feedback loop. Software's "Hello, World" is validated in milliseconds. Medicine's "Hello, World" could take 30 years of clinical trials to ensure that you haven't killed anyone. It is easy because it is quick, allowing a greater understanding within an equal amount of time.
In fact, the adage of years of experience comes from learning that actually take years to encounter different circumstances and see the results play out in order to fully understand what you're dealing with. This is almost never a problem for software engineers, especially in the learning phase. Software engineers can gain "years of experience" each day by seeing the results of what they are doing in practically real-time.
Funny how you can make that statement.
Most Software Engineering Degrees require an ethics course. Yet - we still have the issues we have today.
Asking for good intentions isn't going to change outcomes, because Software Engineers already have good intentions. Nothing will change.
How about Software Engineering Licensure? You have been found to contribute glaring security issues in widespread code? Lose your license and employment. Use dark patterns to defraud people? Lose your license and employment. Leave "debugging" endpoints open or collect unminimized telemetry? Lose your license and employment.
This seems stronger, and much more likely that outcomes will change.
But good luck getting the industry to move.
Also it's not that the underbelly asks would ask their hacker candidates for a license, or that the military/cops would care about a licensing body.
I am honestly much more worried about terrible directives coming from c-suite persons, that result in ethically questionable apps and software.
In addition to ethics training and commitments from the persons doing the actual engineering work.
This is especially rich coming from a faboulsy wealthy executive of a company engaged in many unethical activities
https://en.m.wikipedia.org/wiki/Professional_Engineers_Ontar... https://www.peo.on.ca/licence-applications/become-profession...
That said, this might actually work! If a software engineer can suffer personal harm by working for a business with iffy ethics, then they are incentivized to play it safe by avoiding working for those types of businesses -- thus correcting the market by internalizing the externalities. I doubt anyone would work for Facebook in a world with a Hippocratic Oath for Software Engineers that has real teeth.
Put another way: pointing to decision makers instead of individual engineers is a simple rephrasing of the Nuremberg defense, "I was just following orders!" It is obvious that we should hold leaders accountable. The question here is whether we hold individual software engineers accountable too (they're not mutually exclusive) and the answer is probably yes.