Barrett Emerick (St. Mary’s College of Maryland) and Tyler Hildebrand (Dalhousie University), “Inductive Reasoning Involving Social Kinds”
Journal of the American Philosophical Association, 2024
By Barrett Emerick & Tyler Hildebrand
Suppose you want to develop or defend a real-world social policy. To do so, you’ll probably need to take three steps:
A descriptive step: You start by identifying some pattern in the social world. For example, you might track down some crime rate statistics.
An inductive step: You extrapolate the pattern from observed cases to unobserved cases. For example, you might infer that patterns of crime will continue if left unchecked.
A normative step: You state some goals or normative principles and argue that your preferred policy will bring about the desired outcome. For example, you might argue that policy P will reduce crime.
In comparison to the descriptive and normative steps, the inductive step might seem trivial. Although philosophers are well-acquainted with the difficulty of justifying inductive inferences, we often ignore this in real-world contexts. For practical purposes, it’s usually obvious enough whether our inductive inferences are good ones—or so it seems.
In our paper, “Inductive Reasoning Involving Social Kinds,” we argue that this is a mistake. Reasoning inductively about the social world is tricky. Social kinds change over time, they can be vague, and they are interactive and susceptible to looping effects. As a result, patterns in the social world are relatively unstable. In addition, the three steps of a policy defense aren’t independent. Figuring out how to satisfy the inductive step may turn out to be helpful in evaluating the normative step, at least for some proposed social policies.
Here’s the key idea behind the importance of the inductive step. To defend a social policy, we have to make inductive projections on the basis of the regularities we observe in the social world. Until we can explain why the regularities in question hold—which requires an understanding of the relevant social background conditions—we are not justified in taking the inductive step. (The social world is messy!) Such explanations help us to see that some policies reinforce unjust background conditions, which thereby create or sustain the very data used to justify those policies. In contrast, others correct unjust background conditions, and thereby disrupt flawed inductive loops. Thus, an investigation into inductive inferences themselves can shed light both on which normative principles we should accept and on which policies best promote our goals.
Here's a sample application. Many political liberals are in favor of race-based affirmative action, but are opposed to racial profiling in criminal justice. Some might see a tension here, or feel uncomfortable holding those two commitments, because both race-based policies require reasoning inductively about race. However, on our account, this tension evaporates. In support of affirmative action, one can argue that racial patterns of socioeconomic difference are explained by the history of racial oppression. Under the status quo, we should expect these patterns to continue. However, policies of affirmative action can help to undermine the status quo and correct such patterns. Against racial profiling, one can argue (similarly) that racial crime statistics are explained by the history of racial oppression. Yet we should expect policies of profiling to reinforce the unjust social conditions that lead to racial crime rate disparities in the first place. If we’re right, there’s an interesting asymmetry between these two instances of reasoning about race, and it makes a big difference to how we assess the normative steps in arguments for these two different sorts of policies.
We think this should be of interest to anyone who cares about social justice, but it’s interesting for a more straightforwardly theoretical reason as well. Our way of clarifying the relevant sense of an explanation makes use of some tools from both metaphysics and epistemology—and relatively traditional approaches to these disciplines at that. For example, in our paper we discuss the metaphysics of laws, properties, kinds, and race, and their relevance to the distinction between good and bad inductive inferences. Insofar as our account of the inductive step is correct, it helps to illustrate why metaphysics and epistemology matter. Indeed, it suggests that they can make a difference to the real world.