Why care about anything at all? Wouldn't it be simpler just to carve a wake of ruin and suffering through life in pursuit of ever more perfect optimization?
I mean it is what the bilionaires do and look where it got them. Seems our ancestors built a society that rewards it - and maybe we're in a place where it is unfixable due to the inertia of power.
Because it effects people's lives and mental health. I wouldn't want to live in a society or work for an organisation where a person's value is only measured by their output. If that means accepting (temporary) inefficiencies, so be it.
If you're in the US, you very likely already live in a society that only values lives based on their output as given to someone else (unless you're in the capital holder class and happen to be on the other side of the table).
I think the tech industry is especially susceptible to arguments based on efficiency because after all, that's what technology provides at a fundamental basis. We need to realize, as with many aspects within technology itself, there are tradeoffs to efficiency itself. You still have educated people who will tout efficiency above all else. I'm all for progress and improving efficiency but it has costs (and benefits) on the human condition and we need to understand and factor those costs in instead of only looking at benefits from efficiency gains.
As a kid I was naive and believed this same philosophy of efficiency and didn't understand why we didn't always push to technology to improve situations. As I've grown over the years I realize life is a lot more than obsessing over efficiency. Obsessing over efficiency only brought me misery while accepting a reasonable degree of inefficiency in my life in strategic areas brought me a lot of happiness. It's always important to think about what you're optimizing on and what the costs are. Imagine going to a restaurant and making your dinner purchase decision based on a metric like: $/calorie. It's probably the most economically efficient way to look at food (perhaps with a nutritional distribution factor as well), but who wants to live like that out of choice?
I see your decision to accept "a reasonable degree of ineffiency... in strategic areas" as an optimization as well. I'm thinking that it could be seen as improving your long-term efficiency.
I think we can distinguish between them:
Long-term efficiency, where we get the most output over time by not burning anybody out, and by keeping morale high
And short-term efficiency, which is what everybody focuses on, because that's all shareholders care about (gains now!) and it results in the kind of soul-sucking capitalist rat race that you seem to think the entire country lives in.
I'm thinking also that this is a very city-minded view, or maybe a corporate-minded one, and if you were to make the decision to sacrifice some (a lot) of potential earnings and go found or work for a small business in a small town, you might find that not all parts of America value people only based on their contribution to the GDP.
The "max efficiency" only does not work when you are forgetting about externalities (those are defined as "a side effect or consequence [...] that affects other parties without this being reflected in the cost of the goods or services involved")
In your restaurant example, using "$/calorie" metric will minimize money spend, but will have a negative externality of reducing your happiness. Once you realize this, you'll want to update your cost function to something like "max(satisfaction) limit to total(calories)>needed_nutrition and total($)<budget" etc... and you can use your efficiency search again.
(and then you get to third-order effects and start to think about time spent planning... Python's sort() switches to simpler algorithm for small arrays because overhead of starting up a complex algorithm is too high. The same way, you probably don't want to formulate and solve an efficiency optimization problem for the task of choosing your lunch because this is actually not that efficient :) )
How good it is will depend upon the quality of decisions that are made in designing it, starting with: What is the definition of "usefulness"?
I definitely think there is a room here to design a program that is more just in firing than a person would be. I guess it depends a lot on the benevolence or malevolence of our new computer overlords.
That unsentimental optimization created the depression and anxiety inducing pits of social media with all its manipulation of feelings and opinions and fake news leading to a mistaken attempt to overthrow the us government
Same reason I don't care about eating meat, but wouldn't be able to kill a cow (or at least would feel bad about it): emotions are low-intelligence.
I can visualize the unemployment caused by some optimization that I wrote, but emotions don't engage with abstractions. Until that means that Jared will have to go home and tell Laura that he has been fired and now they can't afford their house it doesn't engage emotions.
And my logic says, why should I care about a person I have never met? I am strictly worse of for having them in the first place.
It's not clear if you talk about instant messaging or social networks - there is nothing in smartphones themselves that differentiates them from "dumb" phones as far as ignoring others is concerned.
I think the main discerning function is that most smart phone communication mechanisms are asynchronous first. For traditional dumb phones, the only async feature was really voice-mail and message recorder systems which, from my experience, lead right back to synchronous communication ("call me back please"). With the advent of SMS, we slowly pushed more into the world of async and people learned how to find more and more socially acceptable ways to increase those time delays to the point of ignoring. For myself, almost all of my comms on my phone are async. Synchronous communication is now the exception instead of the norm. If you call me, unless you're my family or very close friends, you will go to voice mail and you will be screened.
If you ignore the message that says you're fired, you lose the cost of going to work and going home once they fire you in person, plus you might waste a day or two that you could have spent job-hunting.
I take it that you are referring to the async messaging capability of a smartphone.
For me, the primary feature of a smartphone is that it is a telephone. I also have a smartphone app to control a 'smart' device. I can't type more than a word or two on a smartphone without making mistakes; maybe my fingers are too fat and blunt.
If it's just about async messaging: 25 years before the first smartphones, we had email. You could send a message when the recipient was asleep, in the bath, or at the pool. If someone sent you a message, you could read it when that was convenient, in the knowledge that the sender wasn't expecting you to leap to your keyboard to reply instantly.
You could even send the same message to many people at once. And there was best-efforts delivery, which meant that the system would keep trying to deliver, even if your mailserver was pretending to be deaf and dumb, for at least 4 days.
And if the message you received was spam, or simply didn't call for a reply, then you simply didn't reply.
And if you 'did' email on a proper mains-powered computer, then the battery never ran out.
I hate smartphones. I only own one to run this particular app. I never take mine with me when I go out - it stays on my desk.
Maybe I’m naive or just dumb, but: isn’t this better?
One of the big complaints about modern society is all the isms. Sexism, racism, homophobia (ism). Computers don’t know your skin color, gender, religion, sexual preference, etc. Not saying it will solve all problems, but wouldn’t algorithmic decisions be preferable to biased human judgement?
That's not even necessary. Some of those -isms creep in implicitly, because e.g. certain performance metrics assumed by the algorithm may be statistically less likely to be fulfilled by women, or stuff like that.
Doesn't this make it an intractable problem though? Attribute X is more common in group Y first require a population based study, which can be tampered left and right and doesn't always stay the same over time. (commonality of the attribute can shift with generations)
Besides if the goal is "make an algorithm that only base its judgment on population studies" all human decision making algorithms will be the same, and you can endlessly cry wolf.
Probably not, but so what? It happens all the time, because the non-racist logic still affects people unequally due to racial reasons. Here's the top result from "racist algorithms" in Google - https://www.nature.com/articles/d41586-019-03228-6
TL;DR: One of the factors for referring people to treatment was doctor visits over the last year. Makes sense; unwell people go to the doctor, so those going to the doctor more often are more likely to be unwell. But, of course, due to historical segregation, generational wealth, etc, black people on average tended to go to the doctor less. So even if they exhibited the same risks, symptoms, etc, as white people, because they saw a doctor about it less often the algorithm was less likely to refer them for treatment.
No, by they might filter for traits like less likely to commit a crime. Given that African Americans commit roughly half of all crime in America (last time I read the FBI data), such an algorithm would get really racist, really fast if left to itself.
Black people made up 26.6% of 2019 arrests according to FBI data.[1] And arrests and crimes aren't the same. Few wage thieves are arrested for example.
I mean, kind of. If you develop an algorithm and the user you're picturing in your mind looks like you, talks like you, and lives like you do you're embedding all kind of gender-related, racial, and socioeconomic biases. The most famous example being automatic soap dispensers that don't detect black hands[1].
Now combine that fact with the demographics of Silicon Valley companies and you get a recipe for products and algorithms that are "fair" only if you match a set of implicit criteria.
None of this is intentionally malevolent, just a product of hubris and "move fast and break things".
Doesn't having someone who doesn't speak like you, doesn't share your culture, etc make working together worse and therefore not as good of a match? If I went to China and tried to get a job I wouldn't fault them for choosing someone or equal qualification who's also Chinese since by that very fact they'd be a better team fit. Does that make sense?
A lot of -isms are "hidden" -- for example, a manager might be saying to everyone that they do not discriminate, but in practice, they'll never choose people of certain race for promotions.
For human managers, this can be hard to prove, especially if they do not promote people very often. But for algorithms, it is much easier -- multiple people can inspect the source code for suspicious logic. And even if this is some sort of impenetrable AI model, you can still run it with different inputs, and check if the output changes if you change the input race.
They almost certainly will not encode any explicit biases of the algorithm authors (e.g. “she’s pregnant, can we get rid of her?”), which is already a great step forward, as explicit bias is still sadly far too common in the workplace.
When implicit biases find their way into algorithms, it is usually a specific edge case that’s a function of the training data (e.g. facial recognition preforming poorly on darker skinned people due to training set imbalance). The biggest case of implicit bias in algorithms I’m aware of is when the training set is the behavior of society as a whole, as with search engines: for example, women search less for C-level jobs and so are less likely to be shown postings for C-level jobs when doing job searches (thereby perpetuating a vicious cycle), or websites depicting Black teens are more likely to show them in a criminal context than White teens, so searches for “black teenager” show pictures of criminals whereas searches for “white teenager” do not.
These implicit biases are not encoded by the algorithm developer but rather by the dataset the algorithm is applied to.
In the case of automated firing, I don’t see how implicit bias can creep in if the metrics are strictly work-related (e.g. fraction of on-time deliveries). For the record, I do not agree with automating performance metrics, since they cannot account for nuance (e.g. delivery drivers assigned to gated communities have issues opening the gate and thus deliver fewer on-time packages, which is not accounted for in the algorithm.) However, this is not a form of demographic bias, explicit or implicit.
Where I think you are wrong here is that you fail to consider the fact that the same programmer must decide which dataset/training data they will use for the AI. So implicit biais is not just external to the programmer, it's definitely caused by the programmer.
And implicit biais goes way beyond their training dataset.
Did they put a clause "if pregnant, then...", likely not. However, what conditions did they use for testing this algorithm and ensure if it works? Testing is really much part of the process as well.
> So implicit biais is not just external to the programmer, it's definitely caused by the programmer.
I agree in the case of things like facial recognition, but in the case of search results, Google can’t exactly choose whether websites it indexes present “Black teens” in a poor light and “White teens” in a good light. Perfectly unbiased algorithms trained on data produced by society will always exactly reflect the unfortunate biases of society. (Ironically, attempting to correct for society’s biases in the algorithm would by definition bias the algorithm itself!)
This is (likely) not the case with Amazon drivers, since the training data are (presumably) strictly job performance-related, e.g. fraction of on-time deliveries, or number of times a worker showed up late. And even if these happen to correlate to demographics (e.g. if Black workers lived further from the warehouse than White workers [or vice-versa] and were thus more likely to show up to work late), tracking employees based on their on-time performance is not bias on Amazon’s part, since a White worker who showed up late would be just as likely to be fired as a Black worker who showed up late. The fact that one demographic (by no fault of its own) happens to show up late more often is not bias on the part of Amazon.
Sure, but I can't imagine you'd plan for being fired for being locked out of your work area. There's a difference between having your house blow down by a hurricane vs having your house blown down by a slight breeze, and this would appear to be the latter.
Also, even in the US, you can sue for wrongful termination depending on circumstances.
Because the algorithm just bases its decisions on performance, without respect to context, you don't hear any meaningful accusations of bias.
When you give supervisors agency to make decisions you get accusations (real or imagined) about various biased actions, or get accusations that the non-binary pregnant Pacific Islanders are under-represented, or do not represent some other constituency. It's bullshit 80% of the time and generates alot of wasted time, cost and bad PR.
It's a policy decision made because it is better for the company and isn't unlike how supervisors in industrial settings operate. If you've ever seen a very physical factory workplace (where the employees are at-will), they tend to eject broken people with chronic injuries as they age out. When I worked at a mall food vendor in high school in the 90s, it was routine for employees to be terminated for being 5 minutes late or too slow. The bigger lesson is go to school and avoid working in some shitty job where your welfare is a lower priority than the machines you serve.
I think the point of the OP is that these factors don't matter. Performance is the only goal. In general I think this is fine, but it's clear that the algorithm is not yet sound. Maybe we'll get there one day.
In America, ADA absolutely matters. Algorithms like this can blatantly violate the law and open the company to significant liabilities. I've seen a company repeatedly lose expense lawsuits due to "following process" where the process did not take federal laws protecting workers into account.
Such algorithms usually emphasize biases instead of dissipating them.
Sometimes I think though, maybe we should admit in the longer run that some of these biases are actually the truth and all talk of "fair" and "unbiased" is some kind of wishful thinking against all the facts that belongs to the land of teletubbies.
> wouldn’t algorithmic decisions be preferable to biased human judgement?
At this moment? Big NO. Algorithms
- are programmed by humans
- trained on past data (which by definition includes all the "isms")
- have a gajillion corner cases in which they perform hilariously bad
- don't understand the data they are processing so they can't reason something must be garbage in
How's that Amazon recommendation system working out for you compared to a human recommendation? That's the current state of the art algorithm on a very simple problem.
I kind of doubt Amazon's algorithm even uses machine learning, it probably assigns a score based on perf metrics (using a human-written algorithm), then fires employees below some cutoff.
>” trained on past data (which by definition includes all the "isms")”
Is that necessarily so?
Can’t they measure more objective data points like deliveries per hour, missdeliveries, mishandled/broken items (controlled for traffic, pop density, data from deliveries to similar areas, etc) rather than other things like “doesn’t get along with boss:peers, has bad breath, does cat-calls, etc?
Now, I admit this has too much Taylorism in it. To much of squeezing the last drop out of people, but I don’t think it has more soft bias.
> Can’t they measure more objective data points like deliveries per hour, missdeliveries, mishandled/broken items (controlled for traffic, pop density, data from deliveries to similar areas, etc) rather than other things like “doesn’t get along with boss:peers, has bad breath, makes cat-calls, etc.
In principle this sounds great. In practice, there are so many other edge case variables that an algorithm cannot control for (but a human manager can). Some amount of discretion is required when enforcing rules, which an algorithm is incapable of.
Aside from what others have already replied, I think about how bad automated moderation on big platforms can be, and it doesn’t give me high hopes for this type of software.
Think of the stories of an overly aggressive algorithm banning someone for unclear reasons. This has happened to people’s Google accounts. Since Google has limited customer support, people often have a hard time contending the decision. They are locked out of an important slice of their digital lives (email, photos, etc). Can you imagine this happening except it’s your employer who fired you?
You're neither naive or dumb, but this is peak HN.
No, it's not better. Algorithms aren't some magical unbiased utopia. They're created by people, with their own biases. They just add a level of separation between the human bias and the person getting fired.
Humans are, at least, accountable to other humans.
This is the problem, the software is just trained by human decisions. So in many cases they are set up to become even worse.
Generally it is bias cast into software.
The worst part is that they manage to give the false impression of neutrality.
And that doesn't even include all the other problems (cases being ignored/overlooked, false analysis and errors etc.).
Also who guarantees that neutral/fair treatment is ever the actual goal?
Your mistake is in assuming uniform treatment is actually the goal. People have mostly publicly given up that pretense, and now there are plenty of explanations for why fairness isn’t enough and we need to actually have a preference for <group> to <do moral thing>. And even that is just a pretense at some level; most of this stuff is actually politics that has little to do with its stated or even first-order-revealed preference.
The whole concept of at-will employment and employees made to look like contractors is the issue here.
This is a labour rights issue, not an “Amazon is evil” issue.
I don’t see any problem with an algorithm that fires unproductive workers, provided it has real cause to do so.
Instead of bellyaching about how evil Amazon is, perhaps we should be asking why it is so easy to classify someone as a contractor so that you can deny them benefits and terminate them without cause?
Beyond that, why is at-will employment even a thing?
For Canada and most of Europe, the idea that you can be fired for any reason on the spot just because your employer feels like it is ridiculous.
They either have valid cause, or they should have to pay you severance. The courts should not by default side with the mega-corps.
Clearly not a perspective born out of much empathy. Very few people have the requirements to even attempt to start a company and out of that limited pool, most fail within the first year, with luck playing a large role in which companies survive the first 3 years.
It's also bullshit - they do not want anyone to take that advice because employers cannot achieve anything without their employees so if even 20% of their workforce took that advice, their company would implode.
> For Canada and most of Europe, the idea that you can be fired for any reason on the spot just because your employer feels like it is ridiculous.
The counterbalance is supposed be that you can quit whenever you want and aren't owned by your employer. Economic imbalances and capital having more power than labor affects that.
What if the company starts struggling to make payroll 3 months in? What if the employee works hard the first 3 months then stops showing up?
Regulations like that just entrench the biggest players because they can afford the legal headaches of firing. Great way to incentivize people to start companies elsewhere.
Apparently all the companies in the building I'm in didn't get the memo. There is a risk yes, then again there are also way more rewards for the company than the employee so you give some, you get some.
My guess would be zero more than now. None of the startups is limited by getting employees and I'm almost certain no funding would be gained by allowing at will employment as it is a non-issue anyways (as it is already possible to do so by just hiring contractors but nobody does this).
edit: Consider that most people would request considerable more than they earn here now if at-will employment would be the case, offsetting any "gains" you might consider. In Europe wages are considerably lower than the US.
Not being able to afford an employee is obviously a valid reason to dismiss them. So is them refusing to work. But the scenario you describe is contrived anyway: Someone motivated enough to work hard won't randomly stop being so.
The "legal headaches" are only a serious problem if your legal system is broken.
One side of the sword makes hiring more difficult, the other side means you become homeless, lose your health insurance, and cannot provide for your family. It can literally be life threatening to the worker.
The power differential favours the employer so much in this situation that it is ridiculous.
I agree if there was a social safety net like UBI this would be a non issue, but we’re nowhere near that.
There is unemployment insurance and COBRA, although I agree unemployment (without the temporary pandemic increase) isn't really enough especially with COBRA premiums. I do think things are slowly going in that direction though.
However, if a company has eager competitors which can leverage an advantage, then keeping the 10 or 20% of a workforce which is minimally productive or counter productive can jeopardize a whole company and all its employees.
Say intel were to not fire low productivity employees but AMD does, then AMD can capitalize on this, intel loses share and then intel folds subsidiaries and so on. On the other hand AMD would be on a hiring spree to meet the added demand.
I agree with your logic here, but ideally Intel and AMD have to follow the same rules and find cause to terminate their employees OR pay out severance based on duration of service.
You might follow this up by saying that this decreases global competitiveness as well, and you’re absolutely right.
When say for example HiSilicon’s employees work 12hrs a day 6 days a week and live in bunks on company property it would certainly be hard for Intel/AMD to compete. We then start getting into issues of worker rights being baked into international trade agreements.
When do we collectively agree to stop the race to the bottom ? Or do we just accept that labour has no value and humans are expendable?
I think most of the problems with this issue are rooted in the fact that capital is becoming increasingly more profitable than labour. It reminds me of “Capital In the Twenty-first Century” by Thomas Piketty.
Companies providing welfare is inefficient, worse for workers, and doesn't help the jobless (of which there will be more if companies are responsible for welfare). Welfare is better left to the state.
The Amazon drones sprinting around doing package deliveries are certainly governed by a similar algorithmic approach.
When I and I think a lot of the tech community see that, it looks ugly and dehumanizing, and the algo under the hood plays no small role in why.
This news has similar markings. If you feel there is no issue with Amazon’s Algos and their leverage over a vulnerable workforce: that’s your ethics and I won’t fault you.
But, your blind trust that an algorithmic solution to PeopleOps is problem-free is indicative of a common blindspot for techies. It causes a fair amount of social issues downstream when the algos meet the people being governed by them but the industry that creates the algos can’t see the nuance.
I’m not suggesting that there are no issues with algorithmic HR operations.
I’m suggesting that the abuses of such a system will only be solved in the courts, and that those courts decide the rules based on legislation that outlines worker protections.
Corporations exist to make a profit and you can’t expect them to act against the profit motive out of the kindness of their own hearts, you must compel them.
This is ultimately a human decision. The algorithm gives a score but a human decides what a good score is. Doing this at scale is no different than hiring a hundred middle managers to eye ball it.
At the end of the day- if your company wants to fire the least productive 10% of your workforce on a regular basis to motivate employees, you'll be able to do so.
This is a business practice issue and not a technology one
This is a labour rights issue, not an “Amazon is evil” issue.
There's a lot of room for both to be true. There's no need to set up a dichotomy here, especially when you consider Amazon's attempts to curtail unionization [0] and their lobbying to exempt employees from labor protection [1].
At-will employment is a thing because big corporations (like Amazon) have lobbied for it [2].
> According to a report by Bloomberg, Flex drivers, who are Amazon contract workers and not granted the protections reserved for full-time employees, are being hired and fired via an app. A software program monitors each worker to determine whether they are working quickly enough, whether they are driving safely enough, and whether they are efficiently meeting their delivery quotas. That this program is rife with errors and punishes workers for things that are not their fault, from traffic problems to incorrect delivery directions, does not seem to concern Amazon. Workers have often complained about the unfair monitoring and lack of human oversight, but Amazon has maintained its system.
Three reactions:
1. The Bloomberg piece[1] interviews one of the drivers (63-year old Army vet), who had this to say:
> “I’m an old-school kind of guy, and I give every job 110%,” he said. “This really upset me because we're talking about my reputation. They say I didn’t do the job when I know damn well I did.”
I'm not sure we're ready for the kind of society that results when the Amazon way is universally adopted.
2. There's a fascinating and chilling exploration of how far something like Amazon's policy can go in the audio series The Program:
It would interesting to surrender the whole process to AI.
Let's have AI decide what the productivity requirement should be, given biological constraints of us meat sacks. Feed it all the data - churn, sick days, the Geneva convention, employment law, recruit & training costs, development goals....the whole fucking shebang.
I'll have a sportsman's bet with any of you, that the targets for performance set by AI, would be easier to meet than those devised by mere human cruelty.
99 comments
[ 0.22 ms ] story [ 71.7 ms ] threadSo yeah.
I think the tech industry is especially susceptible to arguments based on efficiency because after all, that's what technology provides at a fundamental basis. We need to realize, as with many aspects within technology itself, there are tradeoffs to efficiency itself. You still have educated people who will tout efficiency above all else. I'm all for progress and improving efficiency but it has costs (and benefits) on the human condition and we need to understand and factor those costs in instead of only looking at benefits from efficiency gains.
As a kid I was naive and believed this same philosophy of efficiency and didn't understand why we didn't always push to technology to improve situations. As I've grown over the years I realize life is a lot more than obsessing over efficiency. Obsessing over efficiency only brought me misery while accepting a reasonable degree of inefficiency in my life in strategic areas brought me a lot of happiness. It's always important to think about what you're optimizing on and what the costs are. Imagine going to a restaurant and making your dinner purchase decision based on a metric like: $/calorie. It's probably the most economically efficient way to look at food (perhaps with a nutritional distribution factor as well), but who wants to live like that out of choice?
I think we can distinguish between them:
Long-term efficiency, where we get the most output over time by not burning anybody out, and by keeping morale high
And short-term efficiency, which is what everybody focuses on, because that's all shareholders care about (gains now!) and it results in the kind of soul-sucking capitalist rat race that you seem to think the entire country lives in.
I'm thinking also that this is a very city-minded view, or maybe a corporate-minded one, and if you were to make the decision to sacrifice some (a lot) of potential earnings and go found or work for a small business in a small town, you might find that not all parts of America value people only based on their contribution to the GDP.
In your restaurant example, using "$/calorie" metric will minimize money spend, but will have a negative externality of reducing your happiness. Once you realize this, you'll want to update your cost function to something like "max(satisfaction) limit to total(calories)>needed_nutrition and total($)<budget" etc... and you can use your efficiency search again.
(and then you get to third-order effects and start to think about time spent planning... Python's sort() switches to simpler algorithm for small arrays because overhead of starting up a complex algorithm is too high. The same way, you probably don't want to formulate and solve an efficiency optimization problem for the task of choosing your lunch because this is actually not that efficient :) )
a matter impacting human lives -> sentimental, i.e. cognizant of human cost
a matter impacting my program's performance -> not very sentimental, i.e no human cost to be cognizant of
How good it is will depend upon the quality of decisions that are made in designing it, starting with: What is the definition of "usefulness"?
I definitely think there is a room here to design a program that is more just in firing than a person would be. I guess it depends a lot on the benevolence or malevolence of our new computer overlords.
I can visualize the unemployment caused by some optimization that I wrote, but emotions don't engage with abstractions. Until that means that Jared will have to go home and tell Laura that he has been fired and now they can't afford their house it doesn't engage emotions.
And my logic says, why should I care about a person I have never met? I am strictly worse of for having them in the first place.
IMO: you practically lose nothing by avoiding them.
For me, the primary feature of a smartphone is that it is a telephone. I also have a smartphone app to control a 'smart' device. I can't type more than a word or two on a smartphone without making mistakes; maybe my fingers are too fat and blunt.
If it's just about async messaging: 25 years before the first smartphones, we had email. You could send a message when the recipient was asleep, in the bath, or at the pool. If someone sent you a message, you could read it when that was convenient, in the knowledge that the sender wasn't expecting you to leap to your keyboard to reply instantly.
You could even send the same message to many people at once. And there was best-efforts delivery, which meant that the system would keep trying to deliver, even if your mailserver was pretending to be deaf and dumb, for at least 4 days.
And if the message you received was spam, or simply didn't call for a reply, then you simply didn't reply.
And if you 'did' email on a proper mains-powered computer, then the battery never ran out.
I hate smartphones. I only own one to run this particular app. I never take mine with me when I go out - it stays on my desk.
The dystopia for HR is the burden these apps alleviate.
Stop accepting contrived gate keeping and gifting unwarranted political privilege on corporations if you all are sick of this.
It's public agency that’s the problem. Everyone hates the way the world works while they keep shuffling into that office or opening that laptop.
I have little sympathy for smart people who fail to realize Bezos and the rest literally have no leash attached to them.
https://marshallbrain.com/manna
One of the big complaints about modern society is all the isms. Sexism, racism, homophobia (ism). Computers don’t know your skin color, gender, religion, sexual preference, etc. Not saying it will solve all problems, but wouldn’t algorithmic decisions be preferable to biased human judgement?
Besides if the goal is "make an algorithm that only base its judgment on population studies" all human decision making algorithms will be the same, and you can endlessly cry wolf.
TL;DR: One of the factors for referring people to treatment was doctor visits over the last year. Makes sense; unwell people go to the doctor, so those going to the doctor more often are more likely to be unwell. But, of course, due to historical segregation, generational wealth, etc, black people on average tended to go to the doctor less. So even if they exhibited the same risks, symptoms, etc, as white people, because they saw a doctor about it less often the algorithm was less likely to refer them for treatment.
[1] https://ucr.fbi.gov/crime-in-the-u.s/2019/crime-in-the-u.s.-...
Now combine that fact with the demographics of Silicon Valley companies and you get a recipe for products and algorithms that are "fair" only if you match a set of implicit criteria.
None of this is intentionally malevolent, just a product of hubris and "move fast and break things".
[1] https://reporter.rit.edu/tech/bigotry-encoded-racial-bias-te...
For human managers, this can be hard to prove, especially if they do not promote people very often. But for algorithms, it is much easier -- multiple people can inspect the source code for suspicious logic. And even if this is some sort of impenetrable AI model, you can still run it with different inputs, and check if the output changes if you change the input race.
When implicit biases find their way into algorithms, it is usually a specific edge case that’s a function of the training data (e.g. facial recognition preforming poorly on darker skinned people due to training set imbalance). The biggest case of implicit bias in algorithms I’m aware of is when the training set is the behavior of society as a whole, as with search engines: for example, women search less for C-level jobs and so are less likely to be shown postings for C-level jobs when doing job searches (thereby perpetuating a vicious cycle), or websites depicting Black teens are more likely to show them in a criminal context than White teens, so searches for “black teenager” show pictures of criminals whereas searches for “white teenager” do not.
These implicit biases are not encoded by the algorithm developer but rather by the dataset the algorithm is applied to.
In the case of automated firing, I don’t see how implicit bias can creep in if the metrics are strictly work-related (e.g. fraction of on-time deliveries). For the record, I do not agree with automating performance metrics, since they cannot account for nuance (e.g. delivery drivers assigned to gated communities have issues opening the gate and thus deliver fewer on-time packages, which is not accounted for in the algorithm.) However, this is not a form of demographic bias, explicit or implicit.
And implicit biais goes way beyond their training dataset. Did they put a clause "if pregnant, then...", likely not. However, what conditions did they use for testing this algorithm and ensure if it works? Testing is really much part of the process as well.
I agree in the case of things like facial recognition, but in the case of search results, Google can’t exactly choose whether websites it indexes present “Black teens” in a poor light and “White teens” in a good light. Perfectly unbiased algorithms trained on data produced by society will always exactly reflect the unfortunate biases of society. (Ironically, attempting to correct for society’s biases in the algorithm would by definition bias the algorithm itself!)
This is (likely) not the case with Amazon drivers, since the training data are (presumably) strictly job performance-related, e.g. fraction of on-time deliveries, or number of times a worker showed up late. And even if these happen to correlate to demographics (e.g. if Black workers lived further from the warehouse than White workers [or vice-versa] and were thus more likely to show up to work late), tracking employees based on their on-time performance is not bias on Amazon’s part, since a White worker who showed up late would be just as likely to be fired as a Black worker who showed up late. The fact that one demographic (by no fault of its own) happens to show up late more often is not bias on the part of Amazon.
e.g. getting fired because you couldn't fulfill your "quota" because some neighborhoods have their gates locked over the weekend.
Imagine getting fired for something you have no control over, or didn't even realize was getting penalized for.
This system is cruelty, pure and simple. Is it any wonder that Amazon execs are worried about running out of people to hire?
That is life, that could always happen, the right thing is to accept the fact and plan accordingly.
If i took any kind of job I acknowledge that there always possibility that I'm going to be fired for something I have no control.
Also, even in the US, you can sue for wrongful termination depending on circumstances.
The reason of being fired is beside the point.
Employer should be able to fire anyone for any reason.
Likewise employee should be able to quit for any reason.
Because the algorithm just bases its decisions on performance, without respect to context, you don't hear any meaningful accusations of bias.
When you give supervisors agency to make decisions you get accusations (real or imagined) about various biased actions, or get accusations that the non-binary pregnant Pacific Islanders are under-represented, or do not represent some other constituency. It's bullshit 80% of the time and generates alot of wasted time, cost and bad PR.
It's a policy decision made because it is better for the company and isn't unlike how supervisors in industrial settings operate. If you've ever seen a very physical factory workplace (where the employees are at-will), they tend to eject broken people with chronic injuries as they age out. When I worked at a mall food vendor in high school in the 90s, it was routine for employees to be terminated for being 5 minutes late or too slow. The bigger lesson is go to school and avoid working in some shitty job where your welfare is a lower priority than the machines you serve.
An algorithm also can't apply correctional factors (e.g. age, physical strength, disabilities).
Sometimes I think though, maybe we should admit in the longer run that some of these biases are actually the truth and all talk of "fair" and "unbiased" is some kind of wishful thinking against all the facts that belongs to the land of teletubbies.
At this moment? Big NO. Algorithms
- are programmed by humans
- trained on past data (which by definition includes all the "isms")
- have a gajillion corner cases in which they perform hilariously bad
- don't understand the data they are processing so they can't reason something must be garbage in
How's that Amazon recommendation system working out for you compared to a human recommendation? That's the current state of the art algorithm on a very simple problem.
Is that necessarily so?
Can’t they measure more objective data points like deliveries per hour, missdeliveries, mishandled/broken items (controlled for traffic, pop density, data from deliveries to similar areas, etc) rather than other things like “doesn’t get along with boss:peers, has bad breath, does cat-calls, etc?
Now, I admit this has too much Taylorism in it. To much of squeezing the last drop out of people, but I don’t think it has more soft bias.
In principle this sounds great. In practice, there are so many other edge case variables that an algorithm cannot control for (but a human manager can). Some amount of discretion is required when enforcing rules, which an algorithm is incapable of.
Think of the stories of an overly aggressive algorithm banning someone for unclear reasons. This has happened to people’s Google accounts. Since Google has limited customer support, people often have a hard time contending the decision. They are locked out of an important slice of their digital lives (email, photos, etc). Can you imagine this happening except it’s your employer who fired you?
No, it's not better. Algorithms aren't some magical unbiased utopia. They're created by people, with their own biases. They just add a level of separation between the human bias and the person getting fired.
Humans are, at least, accountable to other humans.
This is the problem, the software is just trained by human decisions. So in many cases they are set up to become even worse. Generally it is bias cast into software.
The worst part is that they manage to give the false impression of neutrality.
And that doesn't even include all the other problems (cases being ignored/overlooked, false analysis and errors etc.).
Also who guarantees that neutral/fair treatment is ever the actual goal?
This is a labour rights issue, not an “Amazon is evil” issue.
I don’t see any problem with an algorithm that fires unproductive workers, provided it has real cause to do so.
Instead of bellyaching about how evil Amazon is, perhaps we should be asking why it is so easy to classify someone as a contractor so that you can deny them benefits and terminate them without cause?
Beyond that, why is at-will employment even a thing?
For Canada and most of Europe, the idea that you can be fired for any reason on the spot just because your employer feels like it is ridiculous.
They either have valid cause, or they should have to pay you severance. The courts should not by default side with the mega-corps.
As employer you may see it differently. e.g if you don't want to be fired make your own company and rule it as you like.
It's also bullshit - they do not want anyone to take that advice because employers cannot achieve anything without their employees so if even 20% of their workforce took that advice, their company would implode.
The counterbalance is supposed be that you can quit whenever you want and aren't owned by your employer. Economic imbalances and capital having more power than labor affects that.
At will employment wouldn't be such an issue if we had a more substantial government safety net.
Regulations like that just entrench the biggest players because they can afford the legal headaches of firing. Great way to incentivize people to start companies elsewhere.
edit: Consider that most people would request considerable more than they earn here now if at-will employment would be the case, offsetting any "gains" you might consider. In Europe wages are considerably lower than the US.
The "legal headaches" are only a serious problem if your legal system is broken.
The power differential favours the employer so much in this situation that it is ridiculous.
I agree if there was a social safety net like UBI this would be a non issue, but we’re nowhere near that.
However, if a company has eager competitors which can leverage an advantage, then keeping the 10 or 20% of a workforce which is minimally productive or counter productive can jeopardize a whole company and all its employees.
Say intel were to not fire low productivity employees but AMD does, then AMD can capitalize on this, intel loses share and then intel folds subsidiaries and so on. On the other hand AMD would be on a hiring spree to meet the added demand.
You might follow this up by saying that this decreases global competitiveness as well, and you’re absolutely right.
When say for example HiSilicon’s employees work 12hrs a day 6 days a week and live in bunks on company property it would certainly be hard for Intel/AMD to compete. We then start getting into issues of worker rights being baked into international trade agreements.
When do we collectively agree to stop the race to the bottom ? Or do we just accept that labour has no value and humans are expendable?
I think most of the problems with this issue are rooted in the fact that capital is becoming increasingly more profitable than labour. It reminds me of “Capital In the Twenty-first Century” by Thomas Piketty.
Addressing the international aspect via trade agreements is a good step.
Better unemployment benefits between jobs would help the domestic aspect.
When I and I think a lot of the tech community see that, it looks ugly and dehumanizing, and the algo under the hood plays no small role in why.
This news has similar markings. If you feel there is no issue with Amazon’s Algos and their leverage over a vulnerable workforce: that’s your ethics and I won’t fault you.
But, your blind trust that an algorithmic solution to PeopleOps is problem-free is indicative of a common blindspot for techies. It causes a fair amount of social issues downstream when the algos meet the people being governed by them but the industry that creates the algos can’t see the nuance.
I’m suggesting that the abuses of such a system will only be solved in the courts, and that those courts decide the rules based on legislation that outlines worker protections.
Corporations exist to make a profit and you can’t expect them to act against the profit motive out of the kindness of their own hearts, you must compel them.
Legislation is a good idea as well
Do you want to believe on faith that business owners will act in a way that matches your morals?
I certainly don’t.
At the end of the day- if your company wants to fire the least productive 10% of your workforce on a regular basis to motivate employees, you'll be able to do so.
This is a business practice issue and not a technology one
There's a lot of room for both to be true. There's no need to set up a dichotomy here, especially when you consider Amazon's attempts to curtail unionization [0] and their lobbying to exempt employees from labor protection [1].
At-will employment is a thing because big corporations (like Amazon) have lobbied for it [2].
0 - https://www.nytimes.com/2021/03/16/technology/amazon-unions-...
1 - https://apnews.com/article/technology-business-washington-se...
2 - https://mainebeacon.com/corporate-lobby-groups-set-sights-on...
Three reactions:
1. The Bloomberg piece[1] interviews one of the drivers (63-year old Army vet), who had this to say:
> “I’m an old-school kind of guy, and I give every job 110%,” he said. “This really upset me because we're talking about my reputation. They say I didn’t do the job when I know damn well I did.”
I'm not sure we're ready for the kind of society that results when the Amazon way is universally adopted.
2. There's a fascinating and chilling exploration of how far something like Amazon's policy can go in the audio series The Program:
https://www.programaudioseries.com
3. Marshall Brain was right[2]. Automation doesn't spell the end of the low-level worker. It spells the end of the middle manager.
[1] https://www.bloomberg.com/news/features/2021-06-28/fired-by-...
[2] https://marshallbrain.com/manna1
Let's have AI decide what the productivity requirement should be, given biological constraints of us meat sacks. Feed it all the data - churn, sick days, the Geneva convention, employment law, recruit & training costs, development goals....the whole fucking shebang.
I'll have a sportsman's bet with any of you, that the targets for performance set by AI, would be easier to meet than those devised by mere human cruelty.