Return to search

Why Are Some Statistical Generalizations Epistemically Risky?

Moral encroachment theses (MET) operate like pragmatic encroachment theses. When the stakes of belief are high, so are the standards for evidence. This means that evidence which is sufficient in a low stakes-of-belief scenario may be insufficient when the stakes are raised. Simply, METs aim to appeal to the varying moral intuitions that one may have in cases with different moral stakes and build an epistemological difference out of that moral distinction. For example, one might think that in cases of racial profiling, because the moral stakes of belief are high, what would otherwise constitute good evidence for belief is insufficient.
However, most METs assume that the probabilistic evidence on which one relies to form their belief is good evidence. Instead of examining the reliability of statistical generalizations, like those used in cases of racial profiling, the moral encroacher focuses on the moral facts of the circumstance of belief formation to explain why the subsequent belief is wrong epistemically. I will focus on Sarah Moss's account because she focuses on cases in which one forms an opinion on the basis of probabilistic evidence. I use Moss's version of the MET as a target to illustrate the challenges METs face in general. Broadly, Moss holds that a judgment's moral risk bears on its epistemic status.
In Section 1, I briefly outline Sarah Moss's MET and explain why it fails to identify which cases produce epistemically problematic judgments and fails to explain why those judgments are epistemically problematic. In Section 2, I offer an alternative account, which explains why statistical generalizations about marginalized social groups are likely unreliable as evidence. Thus, use of this kind of evidence leads to epistemically problematic beliefs. I conclude by introducing epistemic risk as an explanation for why the inference made in Shopper is epistemically problematic while the inference made in Fraternity Member is not. / Master of Arts / Imagine a shopkeeper who has just realized something was stolen from his shop. There are two possible suspects: a young white man and a young Black man. He did not see the shoplifting occur, and the only evidence he has is the statistical evidence that young Black men are 70% more likely to shoplift than young white men. By all accounts, he is not racially biased, this is simply a statistical fact that he is aware of. Based on this evidence, he forms the judgment that the young Black man is the likely culprit. Let's call this case Shopper.
Now imagine a student on a college campus whose friend has been assaulted. There are two possible suspects: a young man who is not a fraternity member and a young man who is in a fraternity. The only evidence that the student has is the statistical evidence that men involved in fraternities are 70% more likely to have committed sexual violence than average. By all accounts she is not anti-fraternity, she is simply aware of this statistical evidence. Based on this evidence, she forms the judgment that the fraternity member is the likely assailant. Let's call this case Fraternity Member.
I think there's a difference between these two cases. Specifically, I think it's okay to make the inference in the latter case, but not in the former. Even if you don't quite share my intuition, you might still think that however 'icky' it feels to draw the above sort of inference in Fraternity Member, it feels ickier still to draw it in Shopper. Either way, I don't think these intuitions are merely responsive to the moral facts of the cases: I think there's something different about the evidence relied upon in these cases. Specifically, we have reason to thinks that the processes with which we produce the evidence relied upon in Shopper are biased.

Identiferoai:union.ndltd.org:VTETD/oai:vtechworks.lib.vt.edu:10919/114735
Date20 April 2023
CreatorsMarley, Maeve
ContributorsPhilosophy, Parker, Wendy, MacKenzie, Jordan, Corredor, Mercedes Maria
PublisherVirginia Tech
Source SetsVirginia Tech Theses and Dissertation
LanguageEnglish
Detected LanguageEnglish
TypeThesis
FormatETD, application/pdf, application/pdf
RightsIn Copyright, http://rightsstatements.org/vocab/InC/1.0/

Page generated in 0.0021 seconds