TL;DR: Group stereotyping, when based on actual group data, is most valuable where it is most unfair and vice versa.

Group stereotyping seems like it would be most useful, and also most unfair, where one uses a proxy for a information that is difficult to obtain. It is hard to come up with an example that is not a political or identity-based mind-killer. So here's a metaphor, with the wariness that a metaphor can mislead as much as it elucidates.

Let's say that we are in the business of basket-weaving. It turns out that the median left-handed person makes baskets worth 5% more than than comparable baskets made by the median right-handed person. As an industry, we have no idea why, but it is demonstrably true, and significant to our business. People invent all kinds of reasons, but no research proves out any of the reasons.

A basket business can test for the value of any individual's baskets by hiring them, having them produce baskets for a couple months, and track the sales price of their baskets. But that is a substantial investment just to get the information. The problem here is the cost of information. Group-stereotyping is the most useful when the cost of information is high. So an approach might be to prefer to hire lefties. But (unless there are asymmetries in the cost of information), it also where it is most unfair to the group member, because it is most costly to provide the information to rebut the stereotype. We end up ignoring the earnest righties who tell us for sure that they can make better baskets than the lefties we are hiring -- and they might very well be correct.

It also seems to me that using group stereotypes is most justified and least unfair where there is high asymmetry in the cost of obtaining (and verifying, if needed) the information, such that the group member can provide at trivial cost the information that is highly costly for the decision-maker to get, and the situation prompts the group member to do so. For example, if our righty basket-maker had a letter from a prior employer that explained how unusually profitable the basket-maker's baskets were, that would defeat the stereotype, because we would know to update our stereotype with individualized data that is actually probative. (An aside: in this situation, we have to avoid being distracted by things that are not probative, such as emotional appeals, irrelevant but positive information, the good looks of the applicant, and all the other things that can lead to an unreliable decision.) In our scenario, the availability of a letter of recommendation doesn't help all the novice basket-makers who are applying for their first basket-making job, so it is not a 100% solution.

One potential solution to this problem is prices, but they have their own problems. If, on average, lefties are worth 5% more than righties in making baskets, the basket industry could adopt pay practices that are directly related to the value of the baskets produced. The problem with that is complexity. That's a broad category, but I can't come up with anything else that holds all the instances. Prices are not determined just by one party; they are determined largely by the market, which means that they are path-dependent, but also evolved. In our particular situation, changing compensation models can run into issues that economists study under the heading of agency or the theory of the firm -- basically, the idea here is that we could have all kinds of unanticipated effects by changing how we put prices on the work of our basket-makers. Still, one could see adopting a test period where lefties got paid 5% more than righties, until the results were in. That doesn't really change things all that much, except that it puts a limit on the period of unfairness, and puts a deadline on updating our information.

Technology is another solution to this problem. For example, one could invent the basket-value test. It's a cheap test that is based on an academic observation of a strong correlation between your ability to identify certain visual patterns with the value of baskets produced. Presumably, if businesses are really missing the boat by failing to hire talented righties, then there is an incentive for someone to invent this technology, because it will lead businesses to use a deeper poo of labor (which presumably lowers their wage costs). What we'd really be doing is substituting one group stereotype (performance on the test) for another (handedness). That would be worth doing if the test were more specific or more precise than handedness in predicting the value of basket production.

But until that technology comes along, it does seem unfair to the specific righties to judge their productivity based strictly on membership in a group (even if for a limited period until real data arrives). If you don't agree, steel yourself against mind-killing, then take the metaphor and map it to race and conviction rates.

Don't use the word "better", use specific criteria along which you want to measure people. I think that widespread conflation of betterness according to some criteria and some kind of absolute scale betterness (how do you even define it? Is it simply a social status by another name or something different?) is precisely the sort of thing that allows some people to strategically equivocate between the two concepts (or claim that other people are strategically equivocating).

Credit and accountability seem like good things to me, and so I want to live in a world where people/groups receive credit for good qualities, and are held accountable for bad qualities.

If this is your concern, then you should take into account what sorts of groups are appropriate loci for credit and accountability. This will, of course, depend on what you think is the point of credit/accountability.

If you believe, as I do, that the function of credit and accountability is to influence future behavior, then it seems that the appropriate loci of credit/accountability should be "agential". In other words, objects of credit and blame should be capable of something resembling goal-directed alteration of behavior. Individual people are appropriate loci on this account, since they are (at least, mostly) paradigmatic agents.

Some groups might also qualify as agential, and thus as appropriate loci of credit and blame. Corporations come to mind, as do nations. But that is because those groups have a particular organizational structure that makes them somewhat agent-like. Not every group has this quality. The group of all left-handed people, for instance, is not agent-like in any relevant sense, so I don't see the point of assigning credit or blame to it. Similarly for racial groups or genders.

It seems to me that "People in Group A are better than people in Group B" is often a piece of rhetoric used to make it harder for people from Group A and Group B to cooperate with each other. This is frequently to the benefit of a small subset of one or the other group.

In short: Who benefits from elevating this sort of hypothesis to consideration? Usually, not you.

If one is perfectly rational (omniscience would even be better), yes, otherwise I do not think it is a good idea for a lot of reasons. Just on the top of my head :

It is very hard to be accurate, let alone objective, when analysing "impact on society" or "quality of character", and the result is dependent on the criteria used.

When there is a big variability within a group (race, genre or whatever), statistics are not very useful and you should end up with a better model by getting to know the person.

Anchoring effect : People are bad at updating evidence when given a first information, there are already enough problems with stereotypes without making it official.

Given a set of parameters, there would be strong incentives to neglect others parameters or to game the system.

Personal responsibility : One qualities depends on a lot things, what are we taking into account? Nature? Nurture? Nothing?

We should not forget that from an evolutionary perspective (if we regard groups as the players) it is advantageous to have at least some bias in favor of the group you belong to. Groups which don't do this, are out-competed by groups who do.

Of course, too much bias leads to extremism. However, no bias at all might lead to the extinction of the group in question.

Humans are already biased towards thinking that various positive characteristics are correlated with each other. Keeping track of an explicit "goodness" variable would make that even worse. So while I don't see anything wrong with comparing specific characteristics between people or groups of people, I endorse the norm that it is not acceptable to make statements of the form "Person A is better than Person B" or "Group A is better than Group B". "Quality of character" is nowhere near specific enough.

Define qualityOfCharacter and impactOnSociety.

I mean someone else might score you very low on qualityOfCharacter because your family has totally failed to provide a sacrificial infant for the yearly solstice ceremony. (Failure to support important social traditions)

Another might rate your qualityOfCharacter as very low because you reported a neighbor to the police for selling weed on the street corner. (betrayal of member of ingroup to a member of the outgroup)

Another might rate your impactOnSociety as terrible because you eat meat and thus help increase the suffering of creatures around you.

Though if you can come up with a definition of impactOnSociety() that almost everyone from a wide range of backgrounds can agree on then you're probably half way to a utility function for a friendly AI.

Basically your question is whether it makes sense to reduce the judegment of the value of a person to a single dimension.

In practice that's what China is reported to be planning with social credit whereby your credit score will also be shown on your dating profile to make sure that all the high quality people get dates while the low quality people have it harder.

Such a system provides a lot of accountability for individual actions. Is that what you want to have?

It's really funny to me that your "terrible" quote and your "doesn't seem wrong" quote are mostly the same things, just with the second emphasizing over and over that you'd be willing to change your mind.

But what's going on here is just our old familiar dilemma of justice vs. truth. It SHOULDN'T be profitable to use someone's skin color as a quick proxy for what's inside their heads. That would be monstrously unfair. People can't help their skin color. It would totally be bullshit that certain shades of human would turn out to be more likely to be late every morning. Imagine if that was also applied to gender? Holy information asymmetry Batman!

Thing is, though, these beliefs will save you money. Extend more trust to folks with the 'good' characteristics, less to those with the 'bad' ones. Hiring a clerk for your urban youth patronized Footlocker? Got 50 applicants? Odds are, the white girl isn't in a gang. She could be, of course. If Warriors has taught us nothing else it is that anyone might be in a gang. But she probably isn't. What about the 30 black dudes? They probably aren't either. The majority of hu-nams aren't. But each of them is more likely to be than she is.

So, you've got a simple set of rules (which I'll snidely sum up as 'Always prefer white dudes for everything'), which seems like, by the numbers, it would save you money. You like money. As Danny DeVito said in that movie, "Everyone does, that's why its called money!". But there's a cost. You don't want to be known as a racist/sexist/age-ist/mental health-ist (even though, you actually would be if you did this).

Solution: Use these rules and rigorously deny that you do. Call it the Silicon Valley strategy, or the Patriarchy, if you are being mean about it.

What's the outcome of this measurement and acknowledgement? How will you use the belief that a median member of a group has some better qualities than a median member of another group?

Why focus on groups rather than individuals, and why on qualities rather than accomplishments?

One day if we'll have the explicit formula for human Coherent Extrapolated Volition, it may become possible to measure one's impact on the society on a single scale. Until then, there is no such thing as a scalar "a.impactOnSociety(time)".

Also: there are many sources of evidence about a person. Some people stop immediately after taking one piece of evidence (such as race) into account, even if there are other pieces available. Of course such motivated thinking is a mistake separate from mere "using evidence about people / groups".