I think there’s a fundamental goal conflict between “fairness” and precision. If the socially-unpopular feature is in fact predictive, then you either explicitly want a less-predictive algorithm, or you end up using other features that correlate with S strongly enough that you might as well just use S.
If you want to ensure a given distribution of S independent of classification, then include that in your prediction goals: have your cost function include a homogeneity penalty. Not that you’re now pretty seriously tipping the scales against what you previously thought your classifier was predicting. Better and simpler to design and test the classifier in a straightforward way, but don’t use it as the sole decision criteria.
Redlining (or more generally, deciding who gets credit) is a great example for this. If you want accurate risk assessment, you must take into account data (income, savings, industry/job stability, other kinds of debt, etc.) that correlates with ethnic averages. The problem is not that the risk classifiers are wrong, the problem is that correct risk assessments lead to unpleasant loan distributions. And the sane solution is to explicitly subsidize the risks you want to encourage for social reasons, not to lie about the risk by throwing away data.
“If you want accurate risk assessment, you must take into account data (income, savings, industry/job stability, other kinds of debt, etc.) that correlates with ethnic averages.”
While not strictly true this is true in essence. The failure point is telling though. What you need is to make generalization that are more general than single individuals. Why that categorization dimensions needs to be ethnicity is not forced at all. Why it would not be gender? Why is it not that you have a certain gene?
When you take such a grouping of indivudals and say that “this average is meaningfull to the decision that I am going to make” that is no longer strictly need.
In dissocaited theorethical talk you could argue and backup as some groupings being more meanignful than others. But the whole discriminatory problems come from people applying a set of groupings that are just common or known without regard to the fit or justifiabliy for he task at hand. That is we first fix the categories and then argue about their ranks rather than letting rankings define categories.
Redlining seems to go beyond what’s economically efficient, as far as I can tell (see wikipedia).
Redlining (or more generally, deciding who gets credit) is a great example for this. If you want accurate risk assessment, you must take into account data (income, savings, industry/job stability, other kinds of debt, etc.) that correlates with ethnic averages.
Er, that’s precisely my point here. My idea is to have certain types of data explicitly permitted; in this case I set T to be income. The definition of “fairness” I was aiming for is that once that permitted data is taken into account, there should remain no further discrimination on the part of the algorithm.
This seems a much better idea that the paper’s suggestion of just balancing total fairness (eg willingness to throw away all data that correlates) with accuracy in some undefined way.
I may have been unclear—if you disallow some data, but allow a bunch of things that correlate with that disallowed data, your results are the same as if you’d had the data in the first place. You can (and, in a good algorithm, do) back into the disallowed data.
In other words, if the disallowed data has no predictive power when added to the allowed data, it’s either truly irrelevant (unlikely in real-world scenarios) or already included in the allowed data, indirectly.
The main point of these ideas is to be able to demonstrate that a classifying algorithm—which is often nothing more than a messy black box—is not biased. This is often something companies want to demonstrate, and may become a legal requirement in some places. The above seems a reasonable definition of non-bias that could be used quite easily.
I think there’s a fundamental goal conflict between “fairness” and precision. If the socially-unpopular feature is in fact predictive, then you either explicitly want a less-predictive algorithm, or you end up using other features that correlate with S strongly enough that you might as well just use S.
If you want to ensure a given distribution of S independent of classification, then include that in your prediction goals: have your cost function include a homogeneity penalty. Not that you’re now pretty seriously tipping the scales against what you previously thought your classifier was predicting. Better and simpler to design and test the classifier in a straightforward way, but don’t use it as the sole decision criteria.
Redlining (or more generally, deciding who gets credit) is a great example for this. If you want accurate risk assessment, you must take into account data (income, savings, industry/job stability, other kinds of debt, etc.) that correlates with ethnic averages. The problem is not that the risk classifiers are wrong, the problem is that correct risk assessments lead to unpleasant loan distributions. And the sane solution is to explicitly subsidize the risks you want to encourage for social reasons, not to lie about the risk by throwing away data.
“If you want accurate risk assessment, you must take into account data (income, savings, industry/job stability, other kinds of debt, etc.) that correlates with ethnic averages.”
While not strictly true this is true in essence. The failure point is telling though. What you need is to make generalization that are more general than single individuals. Why that categorization dimensions needs to be ethnicity is not forced at all. Why it would not be gender? Why is it not that you have a certain gene?
When you take such a grouping of indivudals and say that “this average is meaningfull to the decision that I am going to make” that is no longer strictly need.
In dissocaited theorethical talk you could argue and backup as some groupings being more meanignful than others. But the whole discriminatory problems come from people applying a set of groupings that are just common or known without regard to the fit or justifiabliy for he task at hand. That is we first fix the categories and then argue about their ranks rather than letting rankings define categories.
Redlining seems to go beyond what’s economically efficient, as far as I can tell (see wikipedia).
Er, that’s precisely my point here. My idea is to have certain types of data explicitly permitted; in this case I set T to be income. The definition of “fairness” I was aiming for is that once that permitted data is taken into account, there should remain no further discrimination on the part of the algorithm.
This seems a much better idea that the paper’s suggestion of just balancing total fairness (eg willingness to throw away all data that correlates) with accuracy in some undefined way.
I may have been unclear—if you disallow some data, but allow a bunch of things that correlate with that disallowed data, your results are the same as if you’d had the data in the first place. You can (and, in a good algorithm, do) back into the disallowed data.
In other words, if the disallowed data has no predictive power when added to the allowed data, it’s either truly irrelevant (unlikely in real-world scenarios) or already included in the allowed data, indirectly.
The main point of these ideas is to be able to demonstrate that a classifying algorithm—which is often nothing more than a messy black box—is not biased. This is often something companies want to demonstrate, and may become a legal requirement in some places. The above seems a reasonable definition of non-bias that could be used quite easily.