> WW2 prisoner camps
For starters, those camps would shoot anyone who tried to leave. Not the same thing at all.
Cute line, but obviously people have always wanted criteria to be objective and fair.
If the rule was "take the ordinal value of each letter in employee's name, sum mod 365, and if the result matches the day of the year, they're fired!" -- that's completely objective, and also bonkers. Nobody ever thought all objective formulas were good, and it's sophistry to suggest otherwise.
You do understand that the is a logical impossibility, as "fairness" is a subjective standard...
What is "fair" to me, likely will not be what you believe is "fair" thus something can not be objective and "fair"
> "take the ordinal value of each letter in employee's name, sum mod 365, and if the result matches the day of the year, they're fired!"
My Objection to this standard would not be "it is unfair" but rather it makes no logical sense to have employment choices made by that system and would likely cause many objectively valuable employees to be separated from the company.
Objective rules need to make sense in the context for which they are apply, "fairness" would not come into play here
There are good reasons why judicial systems have higher instances in them to which you can appeal to if you don't agree with lower instance.
Programmed, or even worse trained, which means there’s no general intelligence around that really understands why it does what it does.
There is no objective answer to "How productive is this employee", it's not something you can measure empirically.
While I think you can measure 'how productive is this employee' according to some metrics, the problem is in the metrics. Someone who is mediocre at his job in the quantified sense can be a huge boon to the company due to soft skills or knowledge.
Someone not living for their job could be important to an entire community.
It is a philosophical problem, in addition to finding any kind of metric that actually kinda works.
I don't know about most people but objectivity can fly out the window when it comes to my essential needs. I want another group of humans involved who have similar needs and may consider things an objective metric likely won't.
Some mixture of a utilitarian performance metric balanced with some empathy, morale, and ethical reasoning seems like a better blend to me. Either extreme leads to undesired outcomes in my opinion.
With pure human involvement you can have extremely biased no justifiable firings based on emotion, hatred, racism, sexism, ageism, whatever and in those cases, a metric is your friend. It can potentially show that undesired bias for what it is to others.
On the other hand, that bias may be a good bias. Perhaps your manager is a reasonable human being, they know data the algorithm doesn't like how many additional hours you put in unclocked or that you helped your coworker through a rough problem a few months back, perhaps that you saved the company at a critical junction. They also know your performance recently dropped because there's a pandemic and subequent emotional distress might be effecting you. Perhaps they also notice your performance dropped just like the metric identified, but they know the cause of this is because you just became a parent and have had to divert focus and time towards your family. Your manager has the ability here to step in and override the metric because they know better, have empathy, and understand the world isn't about constantly working at maximal efficiency. This is, to me, a desirable bias.
If we keep chasing and supporting utilitarianism pushed by corporations full-steam, we're going to create quite the hellscape that no one wants to live in.
It is an exaggeration, but with the health & social care systems that the US has it isn't a million miles away either.
Even a judge has limits on comprehensiveness and objectivity. No human judge has perfect knowledge of every law in the United States. When making law, judges choose to uphold stare decisis in most cases but overturn decisions in others because of a certain gravitation towards certain interpretations that largely remain unarticulated. The latter is not done out of some "special" knowledge unobtainable by AI, but by the very subjectivity an AI is meant to rule out in its own considerations.
In contrast, an AI judge could provide an exact numerical weighting and ranking to the sources it cited or present the conditions under which there would be contradictions between one law and another if one were to rule in a certain way. AI doesn't have to be a black box. If done right, it can be most transparent and revealing legal tool to have ever been created.
If these same people got what they wanted, then realised it didn’t turn out the way they envisioned, and had worse outcomes, then used the new information to change their minds, then I don’t see what the problem is.
It's the same as when project managers track how many commits per week each employee makes and use that as a metric for how much work they're doing - an exceptionally dumb example but it literally happens at a lot of companies. And for an algorithm such as this, I doubt they have much better/more useful data to feed it.
Humans have their own set of issues as well though, as you've mentioned. I feel that the problem stems from the way companies are organized with a single entity having full control over resources and decision making, as opposed to responsibility, resources, and power being divided among the people doing the work, in a self-organized and likely democratic way.
As programmers we really should know better than blindly trusting software.
Another programmer has 4 kids, lives pay check to paycheck, and has a wife with a medical condition...
The dept has to be downsized by 1 programmer, both programmers have the same income, the single programmer is slightly more productive..
Should the single programmer should be fired over the other because of there personal life situation?