As this example illustrates there are often complex and interesting relations between normative and descriptive theories in this area (and I believe in the study of cognition more generally) . On the one hand, we can ask, regarding a normative theory, whether as an empirical matter people conform to it at least to some significant extent. In fact a great deal of recent empirical work on causal cognition suggests that this is so for normatively defensible theories — causal cognizers are “rational” to a substantial extent in their judgments and reasoning. At the same time, finding, as an empirical matter, that people’s causal reasoning exhibits certain features can prompt us to ask whether there may be some normative rationale (perhaps not previously appreciated) for the presence of those features and, if this is so, lead to improvements in normative theorizing. CHF describes a number of examples of this dialectic– one is Patricia Cheng’s causal power theory in which certain patterns in ordinary subject’s causal judgments, previously dismissed as “irrational”, are shown to be normatively appropriate once one has a correct normative theory about such judgments. Assumptions about rationality can thus serve as a kind of bridge between normative and the descriptive claims.
One consequence of this view is that it gives invocations of intuition roughly the same status as (some ) results about causal cognition that emerge from ordinary psychology experiments or from X-phi research — in all three cases, we are looking at empirical claims about patterns of judgment and reasoning involving causal cognition. So appeals to intuition have, in this respect, have no special status. In fact, CHF argues that philosophers who appeal to judgments about cases in causal scenarios often turn out, as an empirical matter to be fairly reliable in the sense that they correctly anticipate how others will judge, as shown in experimental studies, but CHF also discusses a number of examples in which this is not the case. As explained in CHF, although sometimes useful, judgments about cases also have a number of limitations when understanding causal reasoning. In particular such intuitive judgments often do not provide reliable information about the causal factors affecting causal reasoning– typically real experiments are required to accomplish that. Moreover, judgments abut cases tell us, at best about verbal behavior, and (at least usually) not about non-verbal behavior which is often a crucial source of evidence regarding causal cognition.
The previous sentence provides a segue into another topic which is discussed at some length in CHF— the role of appeals to intuition or judgments about cases in constructing theories of causation. Appeals of this sort are standard fare in much of the philosophical literature– both on causation and of course more generally: Suzy and Billy throw rocks at a bottle, Suzy’s rock gets there first, and the bottle shatters. We have the intuition/judgment that impact of Suzy’s rock caused the shattering and Billy’s did not. This in turn is treated by many philosophers as a datum to be captured in any philosophical account of causation. In CHF I consider the status of such “intuitions”. I reject the suggestion that they should be treated as sources of rationalistic insight into the “nature” of causation and, more controversially, also reject the suggestion that they inform us about “our concept” of causation. Instead, I claim that to the extent that such judgments about cases provide reliable information about anything, they are best viewed as claims about how the speaker herself judges (that part is fairly trivial ) and how she expects others will judge. That is, the actionable information in the above judgment should be understood as something like “I judge Suzy’s rock was the cause and I believe that others will judge similarly” This is of course an ordinary empirical claim which may well be mistaken since the person involved may be mistaken about how others will judge. Additional experiments can assess such claims.
My talk of normative correctness naturally raises the question of where the grounds or standards of such assessments come from. CHF treats this as a matter of ends/means justification: we have certain ends or goals in causal reasoning and assessment of judgments, reasoning patterns and so on is in terms of conduciveness to these goals. This in turn is connected to my advocacy of a “functional” account of causal reasoning, and to the idea that human forms of causal reasoning is often successful in enabling us to cope with the world and satisfy our goals. I see one of the most important goals or functions of causal reasoning as having to do with the discovery of relationships that support manipulation or intervention, as described in my 2003 book, Making Things Happen which defends an “interventionist” account of causation. Very roughly, interventionism is the claim that X causes Y when appropriate manipulations or interventions on X are associated with changes in Y. Given some body of evidence about X and Y (which e.g. might consist in part of information about patterns of correlation), claims that it is normatively correct to judge that X causes Y or to follow a reasoning pattern that leads to such a judgment are assessed (in part– more is involved as noted below) in terms of whether it is true that such judgments/reasoning patterns will lead to the identification of relationships that support manipulation and control– that is, that these achieve goals associated with causal reasoning. This is rather abstract, as is unavoidable in a word-limited post of this nature, but justifications of this means/ends sort are standard in discussions of causal reasoning in statistics and machine learning. They contrast with many of the standard attempts at normative justification found in the philosophical literature, as when the standard for the true or correct theory of causation is taken to be the one which best systematizes our ordinary causal judgments or “intuitions” or that satisfies certain metaphysical constraints.
How should philosophers (and others) approach the topic of causation and causal reasoning? Causation with a Human Face (CHF) proposes an approach that brings together results from a number of different disciplines, both descriptive ( e.g., the empirical psychology of causal cognition) and normative (statistics, econometrics, machine learning, philosophy). The guiding idea is that causal reasoning should be understood in “functional” terms– that is in terms of the role that it plays in human life and the human goals and purposes that it serves. I see this as contrasting with philosophical inquiries into what causation “is”. Along the way I discuss a number of other issues, including the role of appeals to “intuitions” or “judgments about cases” in philosophizing about causation, the adequacy of “associationist” models of human causal cognition, “rational” models of causal cognition, relations between normative theorizing (how one ought to reason causally) and descriptive models (how as a matter of empirical fact people do reason), what I call distinctions within causation and much else. I also explore issues related to the descriptive adequacy of several philosophical theories of causation understood as proposals about causal reasoning.
Although normative theorizing about causal reasoning is carried out in a number of other disciplines, CHF argues that many prominent philosophical theories are naturally viewed (and ought to be viewed) as normative proposals as well, although they typically also have a descriptive component. For example, David Lewis’ well-known counterfactual account of causation can be regarded as a normative proposal to the effect that causal claims should be understood in terms of certain counterfactuals and that these counterfactuals in turn ought to be evaluated in terms of his possible world semantics and the criteria he proposes for evaluating similarity of worlds. At the same time, Lewis clearly expects that it is true, as an empirical matter, that people often associate counterfactuals and causal claims in something like the way he describes. These observations suggest a number of questions which are explored in CHF. Do people in fact connect causal claims and counterfactuals and use counterfactuals to reason about causal claims ? (Yes). Is it normatively appropriate to connect causal claims and counterfactuals (Yes, definitely– CHF argues that it is useful and clarifying to do so and often facilitates reliable inference. )