Trying to detect a crime like shoplifting beforehand seems to open the door for all sorts of biases, racial and otherwise...
Well you can hope, but it's made by humans, so I wouldn't count on it.
Edit: to people wondering why I do this, I can only say that it’s something that I never even thought of as weird. That’s what growing up rich can do you.
I have a friend that does the same thing and it just seems really weird to me.
My groceries aren't mine until I've paid for them. And also, as a person with not always a lot of money I sometimes (very rarely, but still) find that I don't have enough money for the things I wanted to buy. What if I'd started consuming the goods, got to the till and couldn't pay?
If I needed to drink first I'd buy the drink first, drink it, and then proceed with the remainder of the shopping.
Anyway, people can do what they want. I'm not going to berate anyone for this. But I still think it's really weird.
I think a restaurant is different though because that is explicitly how it works (eat first then pay).
To take your comparison further, what would happen if you sat down in a grocery store and started consuming hundreds of dollars worth of food and drink? I think it would raise a lot of eyebrows, and the employees would probably confront you. So there is a difference.
Furthermore, in a restaurant you place an order and they serve you, and from a legal perspective (IANAL) there might be something about some sort of implicit contract that you will pay them. Whereas with a grocery store you are not allowed to remove the items from the store until you've paid for them. And I feel like consuming the items inside of the store is sort of analogous to removing them from the store.
I would also add that if the store owner came up to you and challenged you to please pay for that right now, you're obligated to do so. Considered as a whole, it's not a good idea for someone to do that over a bottle of soda for all kinds of reasons, but I'd say the right still is essentially there.
I think a potential risk here is from small initial sample sizes. Here's a thought experiment. A system like this is set up and runs successfully for some time. A person with some distinctive characteristic enters the store and steals something and is caught. The system learns that so far 100% of people with that characteristic stole. Next time it detects that characteristic, it flags that person and they're ejected from the store. The system will never learn that this characteristic can be benign, because from now on it will be excluded from it's data set.
That's a bit of an artificial example, making assumptions about how the system might work, but something along those lines could be a potential failure mode, and we might never know what that characteristic is because the system can't express the reasons for it's decision in a meaningful way.
You can’t say if [some race] then that. But you can use other things (are they loitering, barefoot, poor hygiene, has sunglasses, truckers cap, pulls over hoodie on entering, have a decoy, etc)
If the proportion of white shoplifters and black shoplifters being caught is different, AI will just scale up that disparity. If reality is biased, we should work to fix that: not make it more efficient.
To specifically address your point:
"so long as their accuracy is within reasonable range of effectiveness"
This is a far too narrow look at what accuracy means - that's why we usually look at sensitivity, specificity, false negative rate, and false positive rates, and why we break those rates up over important subgroups. To the example numbers you give, let's posit (without evidence) that your numbers are correct - they don't tell us what proportion of people visiting the store belong to each group (is the model just learning to identify ethnicity?), and they don't tell us what proportion of each group is really shoplifting.
In many areas (drug crimes come to mind), there's good evidence that different ethnic groups commit crimes at similar rates, but arrest rates are wildly, grotesquely disproportionate. Imagine that AI is applied to such a system - it learns, correctly, that black people are more likely to be arrested for particular crimes. It then targets black people, and the accuracy numbers look good on paper. In fact, because of the trust we have in technology, such a system might help normalize or justify this crazy preexisting disparity.
If you were building a statistical model to describe predictors of shoplifting, this kind of bias wouldn't make it past a cursory level of peer review. When you're slapping "AI" on a product and implementing it at a huge scale, nobody bats an eye.
My point is about the data generating mechanism - making inferences conditional on a bad selection procedure is a bad idea. It would be bad statistics, and it's bad ML. In this case, my argument is that it's also bad for society.
For example I can legally profile short people, but it isn't moral to do so. And I would rightfully be judged if I did so.
Asking if it is right to judge people based on their mental health is an entirely different topic than asking if we can judge people based on a physical characteristic they were born with and have no control over. Apples and oranges.
Can you elaborate on this? As stated, it sounds not just inaccurate, but impossible: it reads like youre saying that any policy that doesn't result in a uniform (or proportional) racial ratio is illegal, which obviously isn't the case.
I think you're talking about disparate impact (as a legal concept), but this is 1) IIRC likotes to certain domains, like housing or employment and 2) avoidable if the racially-skewed policy is demonstrably related to the job (in the employment case).
As a trivial example, requiring a college degree for a job dramatically shifts the racial distribution of the candidate pool, and obviously isn't illegal.
If the shoplifting system is found to be illegal, AIUI, it would probably be because the case is made that it's _directly_ racially biased (disparate treatment).
Same group that complains when a TSA agent frisks a five year old white kid. All in the name of fairness.
Police has a limited capacity. I want that capacity, paid for with taxes, to be optimized. If that means racial profiling, because some races are more prone to crime, then so be it. That is another form of fairness. Humans employing these techniques make use of common sense, a long sought after feat that AI barely mimicks.
You do not fix societal problems, such as racism or racial crime statistics with algorithms anyway. These communities should put the blame elsewhere, starting with themselves, before they point the finger at "racist" algorithms. It is not statistics fault when certain groups are more or less likely to commit crimes.
So the communities of people who are direct descendants of people were forcibly brought here, enslaved, and then systematically denied access to the mechanisms the rest of our ancestors used to build wealth, should blame themselves because they have higher crime rates?
White families have more than 10 times the wealth of black families on average. This is directly caused by the years of systemic oppression faced by their ancestors.
In addition, to this day, we arrest them at a higher rate rate than White people for the same crimes. We are more likely to convict them for the same crimes. And we give them longer sentences for the same crimes. Elementary school teachers even give harsher punishments to Black students for the same infraction.
Job seekers with black sounding names are significantly less likely to be called back for interviews. Black renters have a harder time renting than White renters with the same credit score. The list goes on and on. When you combine all of this, you end up with a community with a higher crime rate. The problem is caused by systemic injustice, it's not a problem that black communities can fix themselves. If you apply those same conditions to any demographic, you would see a higher crime rate--it's inevitable.
>If that means racial profiling, because some races are more prone to crime, then so be it.
And it's absurd to use that higher crime rate to justify racial profiling, which inevitably has a disparate impact on innocent Black people, which erodes trust in police and institutions, which directly leads to even more criminality. In turn that leads to even more racial profiling. You can't just ignore positive feedback loops by mentioning them and pretending they don't exist.
>That is another form of fairness
Trial by combat, enslaving defeated enemies, and separate but equal are all "other forms of fairness" that our society has rejected.