I have a moral objection to the government using mechanized algorithms for e.g. sentencing, because the government is (or ought to be) accountable to the public, and must be able to justify all actions. Proprietary algorithms are particularly odious, because it is impossible to "justify" a decision even in the sense of tracing how the algorithm got to that point.
On the other hand, private individuals ought not to be accountable to anyone for their decisions (unless they contractually agreed to be), so I have no issue with e.g. a bank making loan decisions with a computer. After all, if the bank's algorithm is innacurate, this actually hurts their bottom line (because they either deny good customers or accept bad customers), so there is a strong incentive to have good algorithms. On the other hand, no such incentive exists for e.g. sentencing.
banks and other financial institutions have certainly been punished for their bad decisions in the past. Definitely don't see anything wrong with letting banks do whatever they want.
But as he said, there is no reason for them to do such policies. They just hurt the bottom line and benefit their competitors. Statistical algorithms are not prejudiced. They just want to make the most accurate predictions possible.
Yes, this was a viable solution back in the day because there was a 90% chance (made up number, but you get the idea) that denying someone from a redlined district was a good financial choice, and it was too expensive to do a detailed analysis to determine if an applicant was in the other 10%. With the automated credit checking and statistical analysis available these days, this isn't really an issue anymore. Banks can afford to do a more detailed check of each applicant, and they will, because failing to do so would be an economic disadvantage.
The example of redlining is actually a perfect rebuttal to "private individuals ought not to be accountable to anyone for their decisions."
Biases can be encoded in algorithms. For example, what if data they used to build their models was biased? It's conceivable that a bank's algorithm could be racist, and still be okay for their bottom line. Just letting the market decide is not enough.
> Yes, this was a viable solution back in the day because there was a 90% chance (made up number, but you get the idea) that denying someone from a redlined district was a good financial choice,
No That wasn't the reason why redlining existed. It existed because it was explicitly racist policy enabled directly from the National Housing Act of 1934[0] -- which was basically just codifying the existing racist attitudes of the day.[1] It stemmed from the racist idea that blacks were inherently bad neighbors in all ways you can imagine. To say this was merely a good enough heuristic, or even the result of government over reach is grossly misunderstand the historical context of these policies. These policies were not any more controversial than saying, we should not zone a preschool next to a oil refinery.
Your first link indicates that, as I said, the purpose of redlining was to direct wise economic decisions. Some researched (linked from wikipedia) suggests that the data may have been biased against black neighborhoods due to bias of the appraisers. It does not suggest that the purpose was to specifically screw over black people.
I disagree. There's no evidence that human judges are less biased or more accurate. There are studies that have shown unattractive defendants get twice the sentences as a attractive ones. Not to mention race. There are studies that show judges give much harsher sentences right before lunch, when they are hungry.
Judges are not better at predicting recidivism than even very crude statistical algorithms. Sentences are basically arbitrary anyway.
Algorithmic sentencing may not be perfect. But humans are algorithms too, and they are way less perfect. Humans don't have transparency either. Unless you want to propose putting judges in MRI machines while they make sentences.
Anyway no one has seriously proposed that algorithms should decide sentencing. Only that they make recommended sentences to the judge. Who overrules the algorithm about 20% of the time.
Your argument in the second paragraph seems to be that arbitrary discrimination, for instance racial discrimination, is economically irrational; discriminatory companies will be out-competed by non-discriminatory companies.
But this ignores the historic reality of racial discrimination, which is that in some areas of the US, at some points in time, a significant number of white people were bigoted, and strongly (for no good reason) disliked black people. If you opened a restaurant that served people of all colors, your white customers would disappear. If your bank made loans to black people who wanted to buy homes in traditionally white neighborhoods, your white depositors would take their business elsewhere.
The civil rights laws of the 1960s made it economically feasible for businesses in these regions to stop discriminating; there were no longer any "white only" businesses for white bigots to patronize.
If you want to argue that in our more "enlightened" times, such anti-discrimination laws are no longer necessary, that's fine, but you will need to provide a certain amount of proof that things have changed.
You are missing a very important part of historic reality - pro-discrimination laws. Prior to the Civil Rights Act, we had a lot of local/state level economic regulation designed to prevent a race to the bottom (e.g. whites being forced to compete economically with colored folk) and to prevent greedy businessmen from acting against the public interest.
I don't doubt that such laws existed, though I know very little about them. I might surmise that the "public interest" upheld by these laws was based on a rather restrictive definition of "public".
But I don't quite understand your point. My point was that if your goal is to end racial discrimination, free-market forces left to themselves are not sufficient. Sometimes legislation is needed to level the playing field and make non-discriminatory businesses economically viable.
Does your post relate to that, and I am just missing something? Or are you heading in a different direction?
I think the point is that we don't know if free-market forces are sufficient, because there wasn't a point in which they acted; there was always legislation pushing to one side or the other. Many of the Jim Crow[1] laws were only overturned after the Civil Rights Act, so the market went from mandatory discrimination to prohibition of discrimination without ever having a free phase.
My point was that if your goal is to end racial discrimination, free-market forces left to themselves are not sufficient.
Your point - or at least the argument you justify it with - doesn't really make any sense.
Lets consider a result of the same argument applied somewhere different:
"...a significant number of people were environmentalists, and strongly favor clean air. If you opened a factory that polluted or purchased supplies from one, your customers would disappear. If your bank made loans to polluters who wanted to finance new machinery, your depositors would take their business elsewhere."
Suddenly your mechanism for enforcing racism seems implausible, right? After all, if it worked, virtually every popular left wing economic regulation would be superfluous.
"In Wisconsin, for instance, the risk-score formula was developed by a private company and has never been publicly disclosed because it is considered proprietary. This secrecy has made it difficult for lawyers to challenge a result."
Shouldn't that on the contrary make it extremely easy for lawyers to argue that the score evidence should be thrown out?
It's tremendously disheartening to see the mainstream media repeating ProPublica's lies. They ran a statistical analysis. Their R-script said that bias was statistically insignificant. So they repeated a bunch of anecdotes in the story and left that part of the analysis out.
Reporting null results won't get you cited by the NYT, I guess.
I might be misunderstanding the article you linked, but it seems very misleading to me (though I agree with the conclusion that the ProPublica article is misleading). They claim to conclude:
"The predictor is probably not biased against any particular race - the race_factorAfrican-American:score_factorHigh term is not statistically significant. Or, as ProPublica puts it, it's "almost statistically significant"."
I don't think you can conclude that the predictor is probably not biased against any particular race -- you can only use significance to reject the null hypothesis, not prove the null hypothesis (especially since the significance is still somewhat high). Am I misunderstanding the claim in the article?
I wrote the article. It's always tricky to me to figure out how to phrase a statement about a frequentist method (since frequentist methods formally say so little). I may have phrased this incorrectly.
So what I'm attempting to say is that they ran a statistical test, were unable to reject the null hypothesis, and then wrote an article phrased as if they did exactly that. But I think you are right that my phrasing is incorrectly, however.
the scores are dependent. It's talking about a subset of the data compared to the entire set, rather than a similar subset or random set. The term unlabeled Random_Race is the same as Random_Race:score_factorHigh+score_factorMedium, as opposed to each one individually. So while neither high nor medium alone have statistical significance for African Americans, together they do, a statistical puzzle.
Or rather, if you had also copied line 44, less of a puzzle, since `score_factorWhatever` becomes significant in a race specific model. It isn't race alone nor the score alone. It is something else (age? another not modeled thing?)
Why the hell would we use this for sentencing guidelines? It isn't very specific, nor sensitive. It has a fairly high false positive and false negative rate.
When you jail someone, you take away that person's rights and you are probably as much as not going to jail them again. This sentencing guideline basically disguises that, and does it in such a way where it isn't specific nor sensitive, and to boot, 1/3 of the time you'll be wrong in your decision. Furthermore, you'll already know in your gut which 1/3 it is going to be - probably someone young or black.
Propublica's writers are totally justified to write the article. They go further than many journalists and show you what they did, and there is no rule in the 4th estate that says they have to do that. If you don't like the fact that statistics as a subject, and policies dependent on statistics, for most people, have to be framed with a story, that's a different issue altogether. But why not come out and just say that.
I think the problem is vastly more subtle and fundamental than the author of this piece (and ProPublica in general, given their previous coverage of the same story) is giving credit.
Software is in a dangerous place right now[1]. Half[2]-way between stupid and smart. Recent advances from fields like Machine Learning have moved the software out of world of the purely deterministic. Stochastic methods have given us software that can live in the gray that is reality; this has made it mighty. Particular subdomains that were once entirely the purview of human workers are rapidly moving towards automation. These algorithms have gotten good enough at what they do, accurate enough at what they do, that they are approaching the semantic-work-outsourcing limit that is 'trustworthiness'; You can treat them more like agents than tools, and trust that they do their job sufficiently well that you, the consumer of their work, need not worry about the details; you get to take the executive role of dealing only with the abstractions they provide. "Just tell me yes or no if we should do this". You implicitly trust that the system will "Do The Right Thing"; of course it will, it's got a fantastic resume with some great recommendations[3].
The problem is, the algs are also still dumb. Very dumb. They cannot model themselves. They cannot introspect. And, perhaps most crucially, they cannot interface with their newly-promoted executives in the lingua franca to explain why they're dumb. When you want to figure out why the intern decided that it would be a good idea to name all their variables some permutation of the words 'herp' and 'derp', you march over to their cube and ask them. A conversation occurs. Different perspectives are exchanged via a common protocol[4]. New knowledge is acquired. A mutual understanding is reached. When you want to figure out why your Facial Recognition Software isn't acknowledging Black People[5]... you go get a Masters in Statistics, Distributed System, Probability Theory, with a minor in Anthropology and Demographics. Then you spend a month reading code and running experiments. The software, briefly perceived as a trustable agent that knows how to do its job, suddenly becomes a tool again, because you can't just ask it why it screwed up so badly. And not just any tool; an incredibly complicated, fragile, and opaque tool, with a million different knobs and dials and a Gordian nest of pipework and conduits that would make even the bravest chaotician sweat a little bit. Even when you've open sourced the data and implementation and the deployment architecture and the napkins you've been scribbling hyperparameters on, the box is still pretty damn black.
And so, while auditability or accountability is important, it's only a small (and very, very, very hard) piece of the societal changes that might be needed. Changes in ethics (who goes to jail when a UAV confuses a hospital for a barracks? What does a smart gun do if its wielder pulls the trigger while pointing at a civilian?). Changes in focus (Is it correct to make such deterministic choices about the world? Maybe Hume was right and the predictability of human behavior is slightly harder than we currently state it to be[6]. Maybe accuracy is bounded lower than we like, and so justification should be the primary target). Changes in education (and a reduction in the magical thinking about computers. I hope some distant ancestors of mine finally see the day that computers are as boring and obvious as hammers).
I worry about focusing so much on accountability because as it stands now, a highly autidable system as defined by the article still needs deep domain knowledge to even begin theorizing about. All of these things above and more will probably need to shift as well, whether deliberately or not, in order to accommodate these smart-dumb tool-agent hybrids we have now, systems that we're building right now, that are just powerful enough to be dangerous.
So first of all this isn't a new thing. The statistical methods vs human judgement debate goes back decades. Long before computers, people were training simple linear models with pencil and paper. The earliest paper I found is from 1928, on a similar issue as in the article. Where a very crude statistical algorithm was better at predicting recidivism of inmates, than three prison psychologists. This isn't a new thing at all.
Second the models in question aren't as complicated as you imagine. These are just decision trees with a few hundred inputs at most. Even with more complicated models, you can train simpler models to mimic them, and then inspect at those. There are other ways to make algorithms more transparent, but my point is these aren't complicated image recognition deep neural nets. They don't need to be either.
Third I don't think these algorithms are "dumb". By all accounts they significantly outperform humans. Humans, even experts, are terrible at doing even basic statistics in their head. Often our decisions aren't much better than random chance. Even when humans are allowed to see the results of an algorithm, and tweak it's output when they think it's making a mistake, they do worse than just the algorithm alone.
There's now evidence to suggest humans are irrationally biased against algorithms. Search for "algorithm aversion". A study shows that even after watching a statistical algorithm do better, people still prefer worse human judgement.
Yep, like I said, it's absolutely an old problem, even when it comes to statistical techniques.
I understand that the particular models in the article are on the simpler side. I was speaking more generally. As to whether the models in question here need to be more complicated, well, that is an interesting question.
Humans are absolutely terrible at doing basic statistics in their head. Algorithm aversion is probably a thing, no doubt motivated by similar anthrocentric biases to luddism. My point was there are other ways of measuring how 'good' a system is. In the space of a well-defined problem with well-defined parameters, sure, maybe accuracy is king. This is strikes me as akin to trying to build a bridge while assuming that the world is a frictionless vacuum. If everyone has agreed that the problem is correctly framed, then yes, a statistical system can be trusted, the mathematics are inevitable. But a statistical model (the current iterations of them, at least) is never going to ask whether or not it -should- only be considering the inputs its given, or whether more is needed. The world that these systems understand consists of the subset of the world provided to them by a set of sample data The patterns they learn can only ever be as good as that.
That's what I mean by 'dumb'. Not that they aren't good at what they do. That they aren't good at knowing about what they can't do.
Here is a paper [1] discussing the effects of the new European Union legislation mentioned in the article ("right to explanation"). Interesting related read, regardless of the validity of the numbers mentioned in this opinion piece.
I would love—assuming Congress had their heads in the right place—to make a small council (council to prevent Luddite or crazy Presidents/Congresses from completely wreaking havoc) that would understand tech stuff like we do. People who just comment "Yea that's not how this works," or "hey techies think this is a good idea." Because I give a lot of governments good credit on tech stuff (and especially local governments) but on the federal level I think we're a little lost and lobbyists get to influence what congress members believe.
Technology isn't special. Politicians are also uneducated on every other industry and area of life. It's just that when they misunderstand technology you notice.
Well I notice it about a couple areas but in a ton of those areas we have specific branches of the executive to deal with the nitty gritty but not really one for technology. Similar in UK where they have white wall ministries to deal with things that parliamentarians should not be expected to have a deep knowledge of.
I think there are elements to complicated tags in complicated algorithms that people need to see and understand. All the details - nah. I don't know all the details that go into the math of my credit report. I do know I have the right and responsibility to see what debts and savings that are reported.
I do think the author is right - issues that can affect someone's rights and responsibilities that are being embedded in an algorithms are part of the public trust and should be examined a bit more closely by the public. Do I or anyone else need to see all the details - no. Do I need to understand some basics and understand some of the tagging. Yes.
32 comments
[ 0.28 ms ] story [ 93.4 ms ] threadOn the other hand, private individuals ought not to be accountable to anyone for their decisions (unless they contractually agreed to be), so I have no issue with e.g. a bank making loan decisions with a computer. After all, if the bank's algorithm is innacurate, this actually hurts their bottom line (because they either deny good customers or accept bad customers), so there is a strong incentive to have good algorithms. On the other hand, no such incentive exists for e.g. sentencing.
Biases can be encoded in algorithms. For example, what if data they used to build their models was biased? It's conceivable that a bank's algorithm could be racist, and still be okay for their bottom line. Just letting the market decide is not enough.
No That wasn't the reason why redlining existed. It existed because it was explicitly racist policy enabled directly from the National Housing Act of 1934[0] -- which was basically just codifying the existing racist attitudes of the day.[1] It stemmed from the racist idea that blacks were inherently bad neighbors in all ways you can imagine. To say this was merely a good enough heuristic, or even the result of government over reach is grossly misunderstand the historical context of these policies. These policies were not any more controversial than saying, we should not zone a preschool next to a oil refinery.
[0] https://en.wikipedia.org/wiki/Redlining#History
[1] http://www.theatlantic.com/magazine/archive/2014/06/the-case...
Judges are not better at predicting recidivism than even very crude statistical algorithms. Sentences are basically arbitrary anyway.
Algorithmic sentencing may not be perfect. But humans are algorithms too, and they are way less perfect. Humans don't have transparency either. Unless you want to propose putting judges in MRI machines while they make sentences.
Anyway no one has seriously proposed that algorithms should decide sentencing. Only that they make recommended sentences to the judge. Who overrules the algorithm about 20% of the time.
But this ignores the historic reality of racial discrimination, which is that in some areas of the US, at some points in time, a significant number of white people were bigoted, and strongly (for no good reason) disliked black people. If you opened a restaurant that served people of all colors, your white customers would disappear. If your bank made loans to black people who wanted to buy homes in traditionally white neighborhoods, your white depositors would take their business elsewhere.
The civil rights laws of the 1960s made it economically feasible for businesses in these regions to stop discriminating; there were no longer any "white only" businesses for white bigots to patronize.
If you want to argue that in our more "enlightened" times, such anti-discrimination laws are no longer necessary, that's fine, but you will need to provide a certain amount of proof that things have changed.
But I don't quite understand your point. My point was that if your goal is to end racial discrimination, free-market forces left to themselves are not sufficient. Sometimes legislation is needed to level the playing field and make non-discriminatory businesses economically viable.
Does your post relate to that, and I am just missing something? Or are you heading in a different direction?
https://en.wikipedia.org/wiki/Jim_Crow_laws
Your point - or at least the argument you justify it with - doesn't really make any sense.
Lets consider a result of the same argument applied somewhere different:
"...a significant number of people were environmentalists, and strongly favor clean air. If you opened a factory that polluted or purchased supplies from one, your customers would disappear. If your bank made loans to polluters who wanted to finance new machinery, your depositors would take their business elsewhere."
Suddenly your mechanism for enforcing racism seems implausible, right? After all, if it worked, virtually every popular left wing economic regulation would be superfluous.
Shouldn't that on the contrary make it extremely easy for lawyers to argue that the score evidence should be thrown out?
Reporting null results won't get you cited by the NYT, I guess.
https://www.chrisstucchio.com/blog/2016/propublica_is_lying....
The power of the media to push a socially convenient lie is pretty amazing.
It looks like this is an Opinion piece from ProPublica. Not much better, I know, but the trail of bias is at least a little less obscured
"The predictor is probably not biased against any particular race - the race_factorAfrican-American:score_factorHigh term is not statistically significant. Or, as ProPublica puts it, it's "almost statistically significant"."
I don't think you can conclude that the predictor is probably not biased against any particular race -- you can only use significance to reject the null hypothesis, not prove the null hypothesis (especially since the significance is still somewhat high). Am I misunderstanding the claim in the article?
So what I'm attempting to say is that they ran a statistical test, were unable to reject the null hypothesis, and then wrote an article phrased as if they did exactly that. But I think you are right that my phrasing is incorrectly, however.
That seems to me to be a fault whose location is not in your stars.
Or rather, if you had also copied line 44, less of a puzzle, since `score_factorWhatever` becomes significant in a race specific model. It isn't race alone nor the score alone. It is something else (age? another not modeled thing?)
They say so in this link, https://www.propublica.org/article/how-we-analyzed-the-compa... Where they explain a bit more in detail what the chart labels mean than what they are labeled in the R script.
Score == dependent variable. So why you ran your blog comments as Score == potential independent variable is a little strange.
As for why you want to critique the very last chart (which does appear in 36, and does as a big chart not to work at all)
Who cares?
Overall, every other time, if we look at a Pearsons fit (line 51)
All defendants Low High Survived 2681 1282 0.55 Recidivated 1216 2035 0.45 Total: 7214.00 False positive rate: 32.35 False negative rate: 37.40 Specificity: 0.68 Sensitivity: 0.63 Prevalence: 0.45 PPV: 0.61 NPV: 0.69 LR+: 1.94 LR-: 0.55
Why the hell would we use this for sentencing guidelines? It isn't very specific, nor sensitive. It has a fairly high false positive and false negative rate.
When you jail someone, you take away that person's rights and you are probably as much as not going to jail them again. This sentencing guideline basically disguises that, and does it in such a way where it isn't specific nor sensitive, and to boot, 1/3 of the time you'll be wrong in your decision. Furthermore, you'll already know in your gut which 1/3 it is going to be - probably someone young or black.
Propublica's writers are totally justified to write the article. They go further than many journalists and show you what they did, and there is no rule in the 4th estate that says they have to do that. If you don't like the fact that statistics as a subject, and policies dependent on statistics, for most people, have to be framed with a story, that's a different issue altogether. But why not come out and just say that.
Software is in a dangerous place right now[1]. Half[2]-way between stupid and smart. Recent advances from fields like Machine Learning have moved the software out of world of the purely deterministic. Stochastic methods have given us software that can live in the gray that is reality; this has made it mighty. Particular subdomains that were once entirely the purview of human workers are rapidly moving towards automation. These algorithms have gotten good enough at what they do, accurate enough at what they do, that they are approaching the semantic-work-outsourcing limit that is 'trustworthiness'; You can treat them more like agents than tools, and trust that they do their job sufficiently well that you, the consumer of their work, need not worry about the details; you get to take the executive role of dealing only with the abstractions they provide. "Just tell me yes or no if we should do this". You implicitly trust that the system will "Do The Right Thing"; of course it will, it's got a fantastic resume with some great recommendations[3].
The problem is, the algs are also still dumb. Very dumb. They cannot model themselves. They cannot introspect. And, perhaps most crucially, they cannot interface with their newly-promoted executives in the lingua franca to explain why they're dumb. When you want to figure out why the intern decided that it would be a good idea to name all their variables some permutation of the words 'herp' and 'derp', you march over to their cube and ask them. A conversation occurs. Different perspectives are exchanged via a common protocol[4]. New knowledge is acquired. A mutual understanding is reached. When you want to figure out why your Facial Recognition Software isn't acknowledging Black People[5]... you go get a Masters in Statistics, Distributed System, Probability Theory, with a minor in Anthropology and Demographics. Then you spend a month reading code and running experiments. The software, briefly perceived as a trustable agent that knows how to do its job, suddenly becomes a tool again, because you can't just ask it why it screwed up so badly. And not just any tool; an incredibly complicated, fragile, and opaque tool, with a million different knobs and dials and a Gordian nest of pipework and conduits that would make even the bravest chaotician sweat a little bit. Even when you've open sourced the data and implementation and the deployment architecture and the napkins you've been scribbling hyperparameters on, the box is still pretty damn black.
And so, while auditability or accountability is important, it's only a small (and very, very, very hard) piece of the societal changes that might be needed. Changes in ethics (who goes to jail when a UAV confuses a hospital for a barracks? What does a smart gun do if its wielder pulls the trigger while pointing at a civilian?). Changes in focus (Is it correct to make such deterministic choices about the world? Maybe Hume was right and the predictability of human behavior is slightly harder than we currently state it to be[6]. Maybe accuracy is bounded lower than we like, and so justification should be the primary target). Changes in education (and a reduction in the magical thinking about computers. I hope some distant ancestors of mine finally see the day that computers are as boring and obvious as hammers).
I worry about focusing so much on accountability because as it stands now, a highly autidable system as defined by the article still needs deep domain knowledge to even begin theorizing about. All of these things above and more will probably need to shift as well, whether deliberately or not, in order to accommodate these smart-dumb tool-agent hybrids we have now, systems that we're building right now, that are just powerful enough to be dangerous.
---
[1] Ev...
Second the models in question aren't as complicated as you imagine. These are just decision trees with a few hundred inputs at most. Even with more complicated models, you can train simpler models to mimic them, and then inspect at those. There are other ways to make algorithms more transparent, but my point is these aren't complicated image recognition deep neural nets. They don't need to be either.
Third I don't think these algorithms are "dumb". By all accounts they significantly outperform humans. Humans, even experts, are terrible at doing even basic statistics in their head. Often our decisions aren't much better than random chance. Even when humans are allowed to see the results of an algorithm, and tweak it's output when they think it's making a mistake, they do worse than just the algorithm alone.
There's now evidence to suggest humans are irrationally biased against algorithms. Search for "algorithm aversion". A study shows that even after watching a statistical algorithm do better, people still prefer worse human judgement.
I understand that the particular models in the article are on the simpler side. I was speaking more generally. As to whether the models in question here need to be more complicated, well, that is an interesting question.
Humans are absolutely terrible at doing basic statistics in their head. Algorithm aversion is probably a thing, no doubt motivated by similar anthrocentric biases to luddism. My point was there are other ways of measuring how 'good' a system is. In the space of a well-defined problem with well-defined parameters, sure, maybe accuracy is king. This is strikes me as akin to trying to build a bridge while assuming that the world is a frictionless vacuum. If everyone has agreed that the problem is correctly framed, then yes, a statistical system can be trusted, the mathematics are inevitable. But a statistical model (the current iterations of them, at least) is never going to ask whether or not it -should- only be considering the inputs its given, or whether more is needed. The world that these systems understand consists of the subset of the world provided to them by a set of sample data The patterns they learn can only ever be as good as that.
That's what I mean by 'dumb'. Not that they aren't good at what they do. That they aren't good at knowing about what they can't do.
[1] http://arxiv.org/pdf/1606.08813v2.pdf
Except lobbyists answer to the people who pay them. Closest thing we have to the tech community paying lobbyists is the EFF.
If we had a professional organization, that would be a reasonable place to organize such a lobbying arm.
I do think the author is right - issues that can affect someone's rights and responsibilities that are being embedded in an algorithms are part of the public trust and should be examined a bit more closely by the public. Do I or anyone else need to see all the details - no. Do I need to understand some basics and understand some of the tagging. Yes.
This is what helps uphold a free society.