Demographic profiling consists in drawing inferences about particular persons from statistical generalizations about their demographic group such as their race, gender or socio-economical class. Even if the generalization is statistically well-supported, somebody who relies on it to draw about a particular person strikes many as doing something problematic.
Practical Example:
John Hope Franklin, a distinguished African American historian was awarded the Presidential Medal of Freedom in 1995. The night before he received the award, he hosted a small party with friends at the Cosmos Club in Washington D.C., a historically white club of which he was one of the very first black members, and where most of staff were black. While walking through the club, a woman calls him out inferences, presents him with her coat check, and asks him to bring her coat. (Schroeder 2021: 192).
The woman’s judgment of ‘Franklin is staff’ was wrong. Now, was her judgment wrong, because she ought not to have believed it (epistemically wrong) or because she ought not to have judged based on it (morally wrong)? This distinction can help answer the question what is problematic about demographic profiling.
There are different views one can take when discussing this problem. The Action view holds that it is only a moral wrong : the woman epistemically ought to believe that ‘Franklin is staff’, but she practically ought not to act on it. Pragmatism similarly holds that the woman’s belief is epistemically rational, but still, she morally ought not to believe that ‘Franklin is staff’. Statistical Deficiency, instead, says that it was an epistemical error: The woman’s belief is epistemically irrational as statistical evidence never justifies a belief ‘a is F’. Moral Encroachment also places the error on the epistemic belief: the woman’s belief is epistemically irrational. Given the moral stakes, she needs other evidence for her belief to be rational.
1. An epistemic worry
The pragmatic view holds that the epistemic belief of the woman was justified based on the statistical evidence as the evidence makes it very likely that ‘Franklin is staff’. However, she ought to suspend her judgment (moral) as it is unfair to judge someone differently and disadvantageously just because he or her belongs to a certain demographic group. An argument against this view would be, that the epistemic belief was wrongfully formed. Somebody targeted by demographic profiling could legitimately complain: “you don’t know me!”. Franklin is not only a black person, but is also a man, from a specific age, belonging to a specific socio-economical class (educated, historian, etc). Not only is the woman ignoring that Franklin belongs to other demographic groups, but she also lacking evidence directly related to Franklin himself (and not only as a member of a specific group).
One could argue that the problem arises given the way the woman forms her belief ‘Franklin is staff’, as she bases it on one demographic group Franklin is part of. Therefore, the woman would need stronger evidence (and not only statistical evidence) before judging that Franklin is staff, hence «don’t rush your judgment».
Let’s assume the woman has another kind of evidence for her belief: She sees Franklin wearing a staff uniform while giving out coats and a reliable witness told her that Franklin was part of the staff. In this scenario, it is epistemically rational fort he woman to believe that Franklin is part of the staff. She gained these reasons to form her belief not from statistical evidence.
2. Statistical evidence: highly probable but ‘weak’
In the literature, several philosophers have argued that statistical evidence is somehow deficient even if it is very likely that P. Some argue that statistical evidence does not justify belief, as the possibility of error cannot be sufficiently ruled out. Instead, epistemic rationality can be defined like this: S’s belief that p is epistemically rational only if S’s evidence ruled out all relevant alternatives to p.
According to Jackson (2020), a possibility is salient (noticeable) for a subject S just in case S pays attention to it (2020: 5084). As such, an epistemically rational agent pays attention to error possibilities made salient by her evidence (2020). Even if we grant that statistical evidence always makes salient error possibilities (0.01%), such a slim possibility of error can simply be ignored. Therefore, statistical evidence cannot serve as a basis to form justified epistemic belief, but only as a credence. Sometimes, statistical evidence is good enough when the moral stakes are low.
3. The Morally Relevant Alternative Model
- A particular individual can be an exception to the statistical generalization as it is not universal.
- We have a moral reason to consider seriously just this alternative possibility: it qualifies as a relevant alternative. Persons have a right to be judged according to their own merits.
- If you are unable to rule out this alternative possibility, then the belief is not epistemically rational.
- Statistical evidence alone does not rule out this alternative possibility. Even if the statistic makes it very likely that p, it does not rule out the possibility that not p.
a. No strong connexion: even if the belief that ‘a is F’ turns out to be false, the belief ‘most as are F’ will not be impacted.
b. Statistical evidence does not distinguish between the actual world (in which Franklin is a member of the club) and a close possible world (in which Franklin is in fact a staff member).
Conclusion : If it is not epistemically rational to believe that p, S ought to suspend judgment. The belief ‘Franklin is staff’ is not epistemically rational, so you ought to suspend your judgment.