Amazon is using algorithms with little human intervention to fire Flex workers
arstechnica.com
arstechnica.com
But what happens when there is no one to ask for forgiveness from?
Living in the country that produces and exports the most weapons, and that also extracts the most hydrocarbons, I came to believe that I lived a superior life via effective propaganda. Now I come to see that these actions are not only against my values, but that they also cause great harm.
There must be superior alternatives.
Like you, science fiction has been a cornerstone of my life since early childhood. For the longest time, I believed our digital selves we're capable of being functional alter egos in a new digital medium separated from physical existence, but that is not what happened. Instead, our real selves and the digital world have merged and in many ways they are inseparable. The pitfalls of this intertwined world are all too visible as society finds itself in drastic divides ranging from political views to music tastes. This coming from someone who believed positive technology advancements would outweigh the negatives, but I can only see the downsides lately. I think I lost the plot and it is all too blurry now. Maybe I need a vacation.
Though gross, That one sentence should tell us their entire line of thinking.
Absolutely nothing will happen to them and they know it. Mistreat one guy, there are 10 others to take his place. There is zero incentive to treat people well.
Unionizing would help, but that doesn’t seem to be happening either
Almost everyone has been groomed to defend these corporate behemots' rights to stomp on our lives like if they were defending the right of a small mom and pop shop to kick out a rowdy drunkard.
When a company has a bigger revenue than most contries' gdp they shouldn't play by the same rules, they need to be held to a different set of standards.
If workers didn't think these jobs were better options than whatever existed before, they'd still be working jobs other than these.
People opted in to this.
Those laws and the regulators may be strongly influenced by deep-pocketed employers such as Amazon.
technology is more ruthless in its application
When I see people slaving away on a bicycle in any kind of weather, delivering groceries or boxes of food, I feel pity. I know they earn very little, have very few benefits, and literally zero chance of moving up. It's a dead end job. It's modern poverty. It's the tiny house equivalent of a job.
I've personally met those refugees, I've personally heard their slavers boasting about it. I'd estimate their proportion in Paris around 30% of all Deliveroo drivers. In smaller cities it's not as common.
These slavers will open accounts in the name of their whole family, and they keep 75% of what those poor guys will make.
This is the part where it stops making sense to me. I don't see why it's morally more just for a mom and pop shop to abuse workers than it is for a big corporation. If it's wrong to treat workers poorly, it should be wrong for everyone to treat workers poorly. This logic implies that the amount you're "allowed" to abuse your workers is on a sliding scale with income. Big corporations should be nice to workers, small businesses are allowed to not be nice to workers, so are businesses in poor countries allowed to be outright awful to their workers? It sounds like the start of a justification for slavery in third world countries.
You can make the argument that Amazon can afford it, and mom and pop shops can't, but I don't find that a convincing argument. It bears the same issues as above. If we establish a sliding scale, we can't simply ignore the implications at the extremes.
Personally, I think we just need to raise the bar across the board. It's bad when Amazon does it, and it's bad when small businesses do it. Small businesses shouldn't get a break on labor violations.
You are of course right, and we don't want abuse at either scale. I think GP is also relying on mom and pop shops not having enough economic power to abuse workers at that level.
A local business usually can't afford burning through workers like cigarettes, it would go out of business at the 3rd contractor (and hopefully these 3 people can take legal action without a barrage of lawyers sucking their life savings).
Amazon can afford to burn through thousands without seeing any clear impact on its bottomline. That's where rules should be stricter IMO.
"The UK GDPR gives people the right not to be subject to solely automated decisions, including profiling, which have a legal or similarly significant effect on them." I guess that would probably be buried in an employment contract at Amazon, but it might form the basis for a court case if it wasn't actually spelt out.
Just because a human has designed the algorithm doesn't make it non-automated: "A process might still be considered solely automated if a human inputs the data to be processed, and then the decision-making is carried out by an automated system. A process won’t be considered solely automated if someone weighs up and interprets the result of an automated decision before applying it to the individual."
Lawyers have a tendency to create employment for other lawyer-type professions it seems.
This is called fraud. It isn't legal.
I don't think human vs automation is the question here. This would not look better if a human boss would evaluate you, and just say in one morning that you're fired, so don't come in. Like in the Back to the Future movie.
Automation applies policy evenly. An employee can be sure that the firing wasn't due to a boss having a bad day, not liking the employee personally, not noticing worse performance of colleagues, etc.
The real problem comes when automated assessment is poorly designed. And it's a shame not only because of the heartless actions that impact real people's lives and the needless public relations damage that is bound to emerge. It's also a shame because of the lost opportunity to help workers' performance improve via feedback at a level of detail that the automated systems could provide.
Damn the secrecy. Develop an open rating system that provides continuous feedback to the worker. Build in dampers that provide ample time to learn, correct mistakes, and improve. Publish every factor, every consideration, every weight that goes into the assessment. Openly. Refine over and over until both management and the workforce see it as fair and helpful.
I would take that system over the subjective, unevenly applied, unclear criteria that's historically been used to assess performance. No surprises to the workers, and management has to own the criteria rather than hide behind unpublished automation. Objectivity should be a virtue. Without transparency, it's going to prove to be quite the vice.
The only way I see capitalism surviving long-term is if we can embrace data collection in all aspects of life and make it work for us to internalize negative externalitites on an ongoing basis.
However, I'll just add that, having tried to design such systems in the past (for large-scale Mechanical Turk work), it is fiendishly complex. Not only is it incredibly difficult to turn our intuitions of "what's fair" into formulas, but then you've got to deal with people trying to game those rules for their own advantage once they figure them out, which ultimately means you can only ever be partially transparent at best.
In fact, I'd go so far as to say it's impossible for all pretty much practical purposes that need to be cost-effective, sadly. It's a really tough problem.
I mean, good for them. They deserve my money for obeying the letter of my requirements. That's exactly what I want them to do. If the letter of the requirements don't accurately and precisely represent the spirit of the requirements, that's on me.
IMHO that sort of adversarial relationship is extremely helpful at developing a good formal spec for the requirements in the first place. It's TDD, but with a monetary incentive on the part of the developer to "code to make the test pass" and nothing more. You don't get to rely on the developer's charitability—the only knob you get to turn, is what test you write!
Of course, in the end, the "tests" might need to be Sufficiently Advanced for this to work, e.g. a spec that specifies something as "what a layman would understand X to be" — where actually evaluating that in an automated way would involve automating a "jury of your peers" type setup to act as the "layman."
The rules might be set up to allow 4 excuses of "building couldn't be accessed" per week before a warning, and 3 warnings in any 60-day period before firing.
If you don't know this, then you simply do your job, and report when a building can't be accessed.
If you do know this, then sometimes you're feeling lazy, want to go home a bit earlier and not do your job, and you lie and report the building can't be accessed, because you know you're under the threshold and so won't be fired.
This isn't about requirements. It's about setting thresholds to catch people who are intentionally not performing their job.
At it happens, this exact thing is 99% of the reason Amazon requires its delivery people to take+submit a photo proof of a delivery when they mark the package as "left at door." Without that requirement, they'd be able to just toss the packages out and mark them as "left at door." Amazon doesn't show those pictures to customers to show them where the package was left — that part should be obvious. They show them to customers so the customer can drive the complaint process when the picture shows that the driver did something stupid! The fact that customers are handed this evidence, in turn disincentivizes drivers from doing something stupid with customers' packages in the first place.
(This personally happened to me just last week: I ordered an Amazon package, and the driver left it at — and took a picture of — the door of a condo within a building that, per the picture, was a completely different building from the one I live in. That picture was very helpful in getting [Amazon's customer service to contact the driver to get] me my package. That was an accident on the driver's part, but it should be clear to the drivers that this is a system that would track — and automatically punish — any attempt at intentional malfeasance at this step, on the basis of this same evidence.)
Amazon requiring its workers to photograph packages is a great idea, and one I applaud them for.
But most instances of employees running into difficulties don't have such elegant solutions. Unless you want employees to be wearing bodycams their whole shift... and then you're going to have to pay other people to review the footage which can take serious time...
Proof can (and should) be positive, rather than adversarial. You could have a bodycam that's recording for the whole shift, but — like in the case of the Amazon driver's photograph — the onus is on the worker to (hopefully with app assistance) trim their bodycam footage down to a short clip that proves the building was inaccessible, and then submit that clip as part of their report. If they choose "building was inaccessible" as an option, then this clip would be a required field on the form; they can't "complete" their job without it.
Yes, there's a bit of video evidence to review. But it's very small amounts of evidence at a time, and evidence only generated in an edge-case. (At least, at game-theoretic equilibrium. There might be more evidence generated at first as workers test the water to see what they can get away with.)
But the point of Mechanical Turk is to gain a scaling advantage by designing your task as a non-iterated prisoners' dilemma (i.e. so that arbitrary new untrained people can come in and do one of your tasks once, and then not necessarily ever interact with you / your tasks again.)
This requires a very different mindset; one that not many people possess. Even most of the people who are using Mechanical Turk "successfully" don't design their tasks to actually be reliably accomplished by the firehose of arbitrary new untrained people that come through the service. Most employers on the service just use it as a temporary-worker recruiting platform — using work-sample test tasks to cull down to a set of reliable workers, and then feeding real tasks to only that pool of pre-tested workers.
But this throws away the key advantage of Mechanical Turk — the scale and elasticity to demand you get by being able to rely on "literally any human" to complete your tasks, rather than needing to be filter them down further and thus expose yourself to winnowing (which is especially harsh in the MTurk world — no worker on the service stays with it for very long, as it's one of the least marginally profitable things you can do with your time, only worth it if you have literally no better options. And most people find another option fairly quickly.)
It's not every task that can take advantage of "literally any human" — but the workloads that can are far more numerous than these tasks' designers think. They just don't understand how to create specs that stand up to an adversarial implementation.
Designing tasks for MTurk is, essentially, writing a wish for a malevolent genie, such that you can automatically verify whether you got what you wished for, and such that you can prove to the malevolent genie that that verification step is what's going to happen, such that they have no reason to defect. It's the baby brother to the AI control problem.
A good example of a well-designed MTurk task (if it were implemented as one) would be Google's older implementation of reCaptcha, the one that shows the "worker" text from books / street-view street-facing signage, and gets them to write what they see. The whole "present one known and one unknown" and "cross check inputs across several workers" parts of the design are there to allow the system to work in a non-iterated prisoners' dilemma context.
Setting up any KPI will reward those that optimize for them. Other workers need to emulate that behavior, which makes their work insufferable. You won't see medium to long term advantage from this.
Define what you need to produce and check if you meet targets. Don't surveil people, it is a waste of time.
This is true with and without policy automation because policy itself has to evolve to address, as you say, a changing world, as well as the organization's changing goals and its outlook on the effectiveness of the existing policy in meeting those goals.
What could maybe work is an automated system for the "easy 90%", paired with an appeals system well-staffed by human beings. But that isn't Amazon's plan (or Google's, etc), because that 10% will cost much more than the first 90%, so it's easier for them to just write it off as a loss. And at their scale, that margin of error translates to thousands of potentially-ruined lives.
I'm saying no policy could ever be defined well enough to cover all possible contingencies of a messy, human world. The conceit that it could is one of the biggest driving factors of our nascent tech-dystopia.
An open policy would of course be better than a closed policy, if only because it would lay bare all of its flaws, but even then it could never be completely fixed.
"All models are wrong, but some are useful."
That approach hasn't worked, and complicating rules and procedures to cover all edge cases will fall short too. That's why judges are still humans. The list of edge cases, the number of factors that can go into a decision, is correlated to the number of things possible in reality. It's basically unbounded.
If I've misunderstood and you are arguing instead for a clear and minimal set of rules to flag and trigger a human review, that sounds sensible to me. In my experience though when companies find an exception they update the policy, and before you know it it becomes a tangled mess.
In that context, consider that a single law, "do the right thing," might require too many judges and might produce too inconsistent of outcomes to work well. "The right thing" may have to be better defined.
Similarly for company policy, it is not the automation used to express bad, ill-defined, or secretive policy that's the problem. The problem is the policy itself.
In the US federal criminal justice system during the 1970s and early 80s judges had a ton of discretion over sentencing. Some were “hanging judges” that gave really harsh sentences. Some never gave any harsh sentences. Some gave harsh sentences to black people but not to white people, and so on.
The legal profession and Congress looked at this situation and decided it wasn’t fair. The difference between probation and ten years in prison shouldn’t be whether or not the judge had indigestion. The federal sentencing guidelines were put in place—-a very detailed system that considered the crime, mitigating and aggregating circumstances, the background of the convict, and several other relevant factors. Judges were supposed to use this system to impose uniform, fair sentences.
Judges hated the sentencing guidelines and kept on trying to depart from them “in the interest of justice”. For a while the Supreme Court said no—-it forced them to follow the system. But in the mid aughts a new majority reversed these precedents. The sentencing guidelines were declared to be advisory.
It took a while for the guidelines era habits to fade but now we are back to a situation where which judge you get is a significant factor in how harsh a sentence you get.
In theory the magic answer is guided discretion. That allows for mostly uniform treatment and a safety valve for unusual situations. But how do you enforce that on a nuts and bolts level? Everyone thinks his particular case is special. Thoughtful empathetic judges are likely to be persuaded that they are.
I don’t know what the answer is but I do know there isn’t an easy one.
Laying people off is very expensive. However, constantly fire people at random and you can quickly grow and shrink your workforce for free!
There's even a name for it: "work to rule".
Poor objective functions can be gamed; good ones align goals such that gaming and aligned goal attainment become the same.
I'd rather face the (admittedly formidable) challenge of designing such objective functions in the open than rely on undefined or secretive policy that's hidden.
Assessment requires definition, by definition.
Suppose I reverse engineer how the system works and find a way to steal, but have that loss attributed to different workers. How will that get discovered? how many honest workers will get fired with no recourse?
You can't replace trust with automation. You still need subjective critical thinking. If the algorithm can find the bad apples, it should be capable of producing metrics showing why they're underperforming to be reviewed by supervisor.
A single photo is a picture. 150 thousand photos is a movie.
Using automation for worker surveillance is the least innovative employment of tech. That workplace will never be good. The thing you are correct about is that it isn't due to the technology itself.
> It's also a shame because of the lost opportunity to help workers' performance improve via feedback at a level of detail that the automated systems could provide.
Unrealistic with the one exception that you provide the info only to the specific worker. Will never work that way if that info isn't kept private.
Source: Experience of anyone that ever worked at a company where middle management must defend its existence to higher ups.
It is just a level of controlling that isn't needed. The time sink alone is a larger waste than the performance increases you net from such measures. And it really gamifies you work, so don't use it on anyone smarter than a jellyfish.
Automation applies policy based on it's inputs/systems.
If someone went out of their way to insure/test that it applies policy evenly, there's a decent chance it will. If not, it'll apply policy based on whatever it's inputs/systems are.
While I don’t morally agree with Amazon policies, we must understand that companies will always optimize to win the capitalism’s game whatever the human cost.
However, as a society, our role is not to ask Amazon (or whatever misbehaving company) to stop being bad guys, but to change the rules of the game so that immoral behavior ceases to be a competitive advantage.
Plus, you really have to have 50% + 1 to make a difference, and then there are still mechanisms against you. You just need to come up with half of $1.74 trillion as of today, and of course enough people who want to sell.
No, I think there are more effective strategies than that.
A lot of AGMs are simply rubber-stamping exercises (look at how many directors are voted in again and again with 80-95% of the vote?!) It does not have to be this way.
You don't have to literally own Amazon to change their ways - in theory you can buy one share and ask some pointed questions at the AGM. If the other shareholders see that change is in their best interest, they will vote accordingly. Sadly, most are asleep at the wheel.
(My money is it's other reasons, and thus the idea of "we must pressure our political leaders! is just a feel good thing we tell ourselves because the truth is very unpleasant to acknowledge)
Between Andrew Jackson to U.S. Grant, there were no 2 term presidents. Maybe they knew something we didn't.
While it had been mentioned in the past, it was the delivery on a Sunday that really helped drive the point home, somehow.
If these algorithms work better with lighter-skinned people (as has happened in the past), could they have a discrimination issue on their hands?
Image recognition works great, as long as you have enough reserves that its failures don't hurt you.
Regarding Amazon, Uber and other "instant employment" type companies - it's a weird world, where you can basically get automatically hired and fired by an algorithm. It creates, in theory, full employment (although not very good employment, at least right now) and gives workers more leverage if they want to quit another job, but also more instability if the rules for the new job aren't made clear or are applied arbitrarily. It also feels like we're in a very nascent stage for this. Maybe this could be done better - it seems like an area that is ripe for disruption.
For real though. We need some labor protections in this country. Remember when we robo-signed ourselves into fraudulent foreclosures?
I, as a developer, would not choose to work at Amazon because of how those who are affected are treated. That is because I have options.
I do not blame anyone who has to take a position there due to their economic position. That is not the general position of people talking here, though.
I'm not comfortable working for any company who treats any of their employees like garbage.
I think I may watch the movie Brazil again.....
You'll find them on Twitter falling over themselves to tell everyone how morally upstanding they are
It's no more natural or inevitable than rows of 12 year old sewing clothes or dead mine workers.
A business relationship like employment is by mutual consent. Either party can opt out at any time.
I'm not sure this is something that needs fixing. You seem to feel differently. What am I missing?
There are tons of places hiring right now; Amazon isn't the only game in town for labor.
I realize it’s easy to forget you’re discussing human beings when computers make much more sense. I don’t know, call me crazy, European, idealistic, whatever, but putting a human in the loop to say “eh, that’s a bad call” isn’t a lot to ask nor the demise of capitalism. If a human being double checking and delivering bad news is the only thing holding the economy back from grand designs of glorious, automated scale, I’d prefer another economy that pays more attention to its purpose - which is the sustenance of humans, not the concentration of wealth based on flagrant disregard for the welfare of people it serves.
I like the way Sorkin put it: “we used to wage war on poverty, not poor people,” and I think your initial question about empathy illuminates much of that sentiment.
This strikes me as ad hominem; I'm talking about standard unskilled workers. Anyone who doesn't have enough savings on hand to buy food and shelter for the short period of time while switching jobs in one of the best labor markets for workers that our generation has ever seen is simply irresponsible.
There are so many unfilled jobs right now.
> I realize it’s easy to forget you’re discussing human beings when computers make much more sense
Oh, come on. We're talking about consent to a business transaction between human beings here, obviously. Please do try to comment in good faith.
Luckily for us, I’m not debating you and need not concern myself with whatever you’d like to call the argument. I’m simply appealing to whatever sliver of empathy you have to consider that just maybe firing someone with a batch job is a good place to think about an ethical line. I didn’t even advocate to draw one, just introduced nuance to think about as a counterweight to your extraordinarily simplistic view of labor and its role in society.
If you can’t meet me even there and instead take an opportunity to call people who just might starve to death on the other end of such an experience “irresponsible,” based on ostensibly a distant understanding of finance and employment for lower social classes, we are not going to agree on anything. The ad hominem would be noting that from your views it reads like you have next to no experience with the financial circumstances you’re discussing, which given the topic seems an interesting but unnecessarily personal observation.
If these comments are in bad faith in this forum then I have absolutely no interest in baptizing my views.
Honestly, I’d be curious how your interview at Dominos would go. I’m not being snide. You should take a break from this line of work and go after the so many unfilled jobs you’re talking about for a while, because I think it would refine your viewpoint on this (speaking from personal experience). You know how one can interview for a cybersec gig and disclose that they were fired once for some bullshit and everyone shares a good laugh about the story while cutting an offer letter? That doesn’t happen at Dominos.
Some views here are very ‘boolean’ and lack nuance... presented as simplifying complex matters. But it’s rather a simplistic mode of thinking, which reduces our lives to lives devoid of empathy. This mode of thinking also tends to see ‘working class’ folks who can’t manage their finances as folks who simply lack character.
In the words of Rutger Bregman... poverty isn’t a lack of character, it’s a lack of cash (recommended read: his book ‘Utopia for Realists’)
And don’t forget, what this pandemic has showed us is that lots of working class folks really are essential workers.
Yes, there are some folks who are lazy, but you find that in all sectors.
Letting simple algorithms designed by folks with a very ‘boolean’ worldview determine who is worthy of employment (survival) and who is not leads us to world as illustrated in the movie Elysium.
I for one appreciate your view.
Please continue baptizing ;)
I’m not sure at this point, both are horrible corporations I’ve resisted as long as I can can, yet both regularly siphon money from my checking account.
A tremendous amount of retail just died in the 2000s, with only a few survivors. Either folding, or being bought out and merging with other companies that folded.
Amazon accelerated what you describe in every quantifiable way and it was undoubtedly baked into the overall strategy. Yes, the situation was already terminal thanks to the Internet and poor online shopping execution by the incumbents, including Walmart, but overlooking Amazon’s involvement (going so far as to call it not their fault) and pointing at Walmart as the disease itself is a tiny bit of revisionism.
Am I the only one old enough to remember discussions in older forums like these about how brilliant Bezos was to identify and exploit such a now-obvious opportunity? It was obvious then to the point of being damned near Amazon’s entire thesis.
Amazon certainly didn't help their bottom line, but it wasn't the nail in the coffin either.
Feels like this sums up so much of our world, it's scary. Anyway.
Main issue is with "mutual consent", or more specifically, negotiating power.
A contract negotiated between two parties with similar amounts of power will have terms that protect both sides. These will often define things like advance notice for cancellation, or penalties for early cancellation. The paying party usually wants the most flexibility for cancellation, the counterparty wants certainty of revenue, so a balance is found. Assuming everyone did their work, both sides should be reasonably happy with the terms of the contract.
A negotiation between a large corporation and an individual (low-level) employee is not really a negotiation, the individual employee can only accept or reject the corporation's offer but they cannot typically negotiate specific parts of the contract. Individually, they have no leverage. If they decline the offer and move on to the next company, they don't have any more negotiating power over there either.
Let's also talk about BATNA a bit here. BATNA for a company: they just hire the next person that comes along, still with the terms dictated by the company. BATNA for an employee: accept an offer somewhere else (again with the terms dictated by the company), or starve.
So what can an individual employee do to get more leverage and get an outcome closer to a "mutual consent" contract? Well, the obvious option is to coordinate their negotiation with others who also have low individual negotiating power but share similar interests. Combined, they should have a lot more leverage.
We could call this negotiating process "collective bargaining", and the group of individuals a "union".
Of course if that was correct we likely would not have seen the rise of unions in the first place because things would simply have adjusted over the many decades if industrial revolution leading up to wide scale establishment of labor unions. When it comes down to being able to survive, people will accept the terms they're given. They don't have the luxury of waiting for a "market correction" on labor rates and benefits.
Open markets can find efficient solutions to complex problems. Unfortunately, much like waiting for evolution to find a solution for a problem, it doesn't always happen at a time scale suitable for human lifetimes.
Because people arrange their lives around their work situation and it can be significantly damaging to them to lose their jobs due to a bad manager or, as in this post's example, a flawed algorithm.
It takes awhile for a new job to cut that first check. In America, usually around a month.
That month can be insanely scary with no money and mouths to feed.
And the CEOs or their lobbyists go to state and federal capitols to lobby for destruction of these programs, or "donate" to politicians' election campaigns with "wink wink, you know what we want."...
For the poor, losing a job arbitrarily, with no protections, even cultural, can be massively life disrupting.
In civilization, we make affordances for people that go against the razors edge of pure efficiency.
It's really only the US that allows employers to unilaterally terminate their employees, and there's ample evidence that companies use it as plausible cover for all sorts of actions that would be illegal (discrimination against protected groups). That, in my book, is indeed something that needs fixing.
But even beyond plausible discrimination: mutual consent is least ambiguous when there's mutual and equal power. No such balance exists in the employer-employed relationship: employees are de facto dependent on their employers for their ability to feed, clothe, and shelter themselves.
What if I want to quit?
If I can't be fired, then the complement to that would be I cannot quit.
There needs to be a balance where I can quit the company and the company can quit me but where companies don't abuse the system (ex Amazon in this example). I don't think there is a great solution, other than a command economy (but there you are not quite free to change "companies" either).
Is there supposed to be a catch here? If you want to quit, then you should be allowed to quit.
To make it explicit: in the vast majority of the developed world, the individual employee's autonomy is given priority. Companies still have broad discretion when it comes to firing; they're simply held to a higher standard of justification than "we don't like you anymore."
This compromise does make sense in a slightly unchanging economy.
Also, if you have highly sought skills this is a detriment, if you have skills that are not in demand, then this is nice to have.
You can also always negotiate an earlier release. These periods offer stability for both parties.
How does this work elsewhere in the world? From a legal point of view at-will employment seems like the default when it comes to buying a service (ie. labor). If you rent a server from amazon you can stop the rental anytime you want, and they can kick you off anytime you want.
I think this gets to the heart of the difference: Americans tend to think of their employment as a continuous exchange of dollars for their labor, whereas most of the world takes the concept of an employment contract to heart.
That contract, in turn, contains both legal and contractual stipulations that prevent firing without cause: employers must give you so-and-so many weeks of notice before beginning the process of firing you, must pay you for so-and-so many months if you present an argument to a labor board, &c. There's nothing informal about it: it's the difference between continuous billing for a service and a fixed-length contract.
Here's a random PDF I found that explains some of the difference between US and French labor contracts[1].
[1]: https://www.bakermckenzie.com/-/media/files/insight/publicat...
Your boss does.
A virtualized server from Amazon does not need food, clothing, and shelter to survive. It does not have a family to support. Simply, it is not alive. and laws around humans and commodities should probably be very different.
More cold-hearted practical arguments for businesses: the virtual machine cannot vote in the next election, it cannot contact the local news media to complain about working conditions, it cannot unionize. If things get very bad it cannot buy torches and pitchforks.
This is false. This presumes that all employees are incapable of saving money or seeking alternate employment quickly.
Neither is true.
The fact that some (tech/highly skilled) workers have actual leverage doesn't negate the fact that many, MANY others don't have any.
We can't operate on how things ought to be, we have to look at how things are actually...and they're bad.
The privates were all paid the same, every two weeks.
Half were able to pay their bills and did well. The other half would spend their paychecks as fast as possible, and would run out of money after 1 week. The 2nd week was spent begging, borrowing, and in general being a deadbeat.
My dad would sit down with them, go over their spending, and prepare budgets for them. Literally none of them were able to adhere to those budgets.
Again, 50%.
It doesn't presume anything of the sort: you don't need to make a statement about "all" employees to observe that the average American under 35 has under $10,000 in liquid savings[1]. That can disappear pretty quickly with a car or home accident, unexpected injury, or legal liability. Even outside of those, it's not very much to live on outside of LCoL areas for more than a couple of months.
The average American is also not a tech worker: they're somewhere in the lower middle class, and may or may not have the ability to transition rapidly between jobs (our current funhouse-mirror economy aside). You're making extraordinary presumptions about millions of people, presumptions with very real and material consequences.
But to go even further: we should not be punishing people with the threat of hunger or exposure for failing to keep liquid savings! The current economy doesn't encourage liquid (i.e., conservative) saving techniques, and American education does not consistently include financial planning. The vast majority of Americans are blamelessly "winging it" and should not be punished for that.
To tie a bow around it: I don't need to make a universal statement about every employee in America to correctly observe that many employees are materially threatened by at-will employment. Attempting to promote this into universal claims is bewildering at best; it's immediately obvious to everyone that plenty of people can survive an unexpected firing.
[1]: https://www.businessinsider.com/personal-finance/average-ame...
The low hanging fruit analogies are slavery, limited suffrage, and laws against homosexual civil unions, all of which were global norms at one point.
The positive argument is that it's already a fixture and an agreed-upon solution in most of the developed world. The normative argument, i.e. that it's good, is in the next two paragraphs.
For starters, the unequal status of corporations and their employees?
Contract law requires good faith execution of the contract. It is not good faith to terminate contracts using a known bad system.
As are my customers.
It's a person's responsibility to plan for job loss.
This is a childish fantasy.
If you have a policy you require of your employees("You must take a picture of yourself"), and you terminate the employee for not doing it, that is not a without-cause termination. That is a termination for-cause - the cause being the violation of the policy.
Termination without cause would basically be "we are downsizing, you are no longer needed".
The rules about "without-cause" are pretty consistent state to state. However, the rules on "with cause" vary state to state. In a lot of states, you can't terminate someone for violating a policy if they didn't violate it. These policies are a form of "implied contract".
In other states, the employer can act completely and totally arbitrarily, as long as it is not acting illegally.
This is actually the minority of states.
On what basis? Aren't they contractors? Do contractors get protections against arbitrary "firings"?
Plus cancellation terms are likely codified in terms of the license agreement.
Note that getting a computer to do something doesn't make it unbiased. People can, and often do, write bias into programs.
Question is, according to whom. For example, Uber got hit with a pretty big fine in Italy because in the view of the court, the people working for Uber should be viewed as employees.
But I know nothing about pre-Uber Italian taxi regulations.
How is this considered acceptable?
Many years ago, I had a friend who was questioning a lawyer about a general indemnity clause in a contract. The lawyer said, "Don't worry about it too much. The courts don't like it when someone tries to take away your ability to seek legal satisfaction." I guess that has changed.
This is at least super immoral and fucking gross.
I guess I need to read up on legal terms. How are these obvious problems not breaches of good faith? I can admit I don't know much here, but, I'm shocked to find out that none of the examples in the article are considered "even close to" a breach of good faith?!
What would it take for Amazon to fire someone and breach the assumed good faith? Do you have an example? I'm curious and honestly trying to learn. I googled this obviously but since I'm not a legal expert the results are iffy and can be difficult to interpret so I appreciate the conversation on this, cheers.
It would be a breach of good faith if you say to the employee that shaving your beard is okay when they ask but then fire them for it because akshually paragraph 113.5 of your employee handbook prohibits beards. If you did not say anything like that, then the default conditions apply, and in an at-will employment state firing someone because you don't like their face (for example, after they have shaved) is not prohibited.
If you explicitly say every now and then "we may arbitrarily terminate you" and then actually do terminate people for arbitrary reasons (for example, fire everyone whose name starts with 'F' or employee ID ends with '13'), that's not a breach of good faith, as you're not breaking any promises, neither explicit or implied, you're doing exactly as you warned them you may do.
"Examples of bad faith terminations include an employer firing an older employee to avoid paying retirement benefits or terminating a salesman just before a large commission on a completed sale is payable. There have been relatively few cases in which employers were found liable under an implied covenant of good faith and fair dealing theory."
Being fired by an opaque, capricious algorithm is not bad faith. It's likely more legally sound for Amazon than firing someone the "old fashioned way"—There's no potential for personal biases to get in the way, and I'm sure the software squirrels away all the necessary documentation should a wrongful termination suit arise.
Given the well-known problems with computer vision and race (and the lighting-related problems this have which have similar source), I’d be very surprised if this didn't constitute disparate impact racial discrimination if someone actually did the legwork to gather the relevant data.
Every person in the article seems like they were screwed over when they were fired. All of them.
> Being fired by an opaque, capricious algorithm is not bad faith.
Yes, it is. When the 'algorithm' is already known by executives to make horrible mistakes, it definitely is.
From the article: “Executives knew this was gonna shit the bed,”
> There's no potential for personal biases to get in the way,
I do not agree with this at all. Someone - people - wrote the code that they use to do the firings. Just because a 'computer' took the steps does not mean that there is "no potential" for personal biases to get in the way. In fact there is unlimited potential for that, and it's sneaky, because you can just blame the software!
> and I'm sure the software squirrels away all the necessary documentation should a wrongful termination suit arise.
Are you really sure about this? Why are you so sure? I personally would be confident that it does not do that. If it did, wouldn't it be possible for it to not make these decisions in the first place?
In the sense that 50/50 is “most”, yes.
30 days? Did you order from aliexpress rather than amazon?
I’ve also been buying less on Amazon, but their easy returns if something arrives truly broken or mismatched from the description are a lot more peace of mind than one typically gets with random small businesses. I’ve had to file a chargeback before after a small biz refused to let me cancel an order after informing me after charging me that one thing I ordered was no longer available in the configuration I ordered it (auto parts so it literally wouldn’t have worked). It was a huge hassle and a few hours of wasted time, plus I still had to find the hard-to-source part somewhere else. On Amazon I’ve been refunded immediately, no questions asked, when I’ve been sent the wrong thing. I’m sure that people who return stuff at a higher rate are super annoying for merchants with how Amazon puts seemingly all of the burden on them, but for consumers who mostly buy stuff in good faith (e.g. not buying a tool to use it then return it when done), it’s a great experience. I’m sure they have metrics on customers around how much they return and what that “free shipping” really costs.
Yes, they highlight transparency issues which is the problem. When using a tech like this objectives must be clear, workers should know the system and know the expectations and productivity criteria
The algorithm could be biased at first place.
if(employee.isnt_somehow_white()){ employement_service_provider.fire_quicker(employee)}
if(employee.is_pregnant()){employement_service_provider.fire_the_quickest(employee))
One can speculate about possible (and/or highly personal) correlations while not realizing that the outcome is completely fair.
That's irrelevant. My point is that no bias is eliminated at first place, since the algorithm is still man made. You could argue that since there is no human intervention the same input would lead to the same output, but again, neither you or I have read the code to claim that it is somehow "bias free" or even completely deterministic, especially if the input relies on computer vision.
This is a straw man. if specific criteria are being automatically evaluated bias is most certainly reduced (if not eliminated).
2. "especially if the input relies on computer vision" - evaluation criteria can be infinitely improved upon
3. "neither you or I have read the code" - I highlighted transparency concerns in my original comment
4. Unavoidable, unintentional implementation flaws (bugs) are expected but so are fixes and continuous improvements.
I would certainly trust an algorithm to make a decision about my employment over an unpredictable, flawed, and (often times) incompetent human management
The question is, how do you detect such a bias, and how do you prevent it in the first place? We have some ideas when it comes to humans, but for code it's much more challenging.
https://www.reuters.com/article/us-amazon-com-jobs-automatio...
The subject of AI and ethics is, in my opinion, something that is not considered enough by the actors of the field, including developers.
https://www.propublica.org/article/machine-bias-risk-assessm...
https://www.boredpanda.com/google-translate-sexist/
https://openai.com/blog/multimodal-neurons/#biasandovergener...
https://www.theverge.com/2018/1/12/16882408/google-racist-go...