Downward Causation and the Epiphenomenalist Revenge
Woodward (2021, 2022) argues that ideal interventions---suitably qualified---may be used to justify the existence of mental causation. I argue that his argument fails because it runs into a paradox. In fact, in conjunction with standard assumptions in this debate, which Woodward endorses, Woodward's argument entails both that there is downward causation and there is no downward causation. This renders his position vulnerable to what I call the "epiphenomenalist revenge": the most plausible way out of the paradox is to give up mental causation altogether.
High-level Causation and Causal Inference (with A. Moneta)
Experimental methods for causal inference (e.g. randomized controlled trials) are believed to conclusively identify causal relations in virtue of realizing ideal conditions (Woodward 2003) that avoid confounding. We observe that many high-level aggregate variables have potentially ambiguous effects on other variables due to their heterogeneous causal role in the population of interest (Spirtes and Scheines 2004). We argue that, when heterogeneity is present and when data on individual units are unavailable, experiments provide a much weaker inferential leverage. The reason is that the ideal conditions on which a conclusive inference would depend are in principle unrealizable. Contrary to non-heterogeneous cases, no possible evidence may invalidate an experiment because confounding remains always compatible with the evidence. Thus, the problem arises of how the widespread practice of causal inference may be justified. We propose a rationalization based on a form of abductive reasoning.
Ill-defined Variables and Instrument Validity (with A. Moneta)
Angrist and Pischke (2010) argue that instrumental variable (IV) methods lend credibility to estimates of causal effects and, consequently, to policies based on such estimates. Here we focus on one reason for scepticism, namely failures of exogeneity. A valid instrument Z on X wrt Y must be exogenous wrt Y, meaning that it must not induce confounding on Y. Deaton (2010) notes that exogeneity is an a priori identifying assumption, which is hard to verify by empirical tests (e.g., overidentification). He concludes that this undermines the credibility of IV studies. This paper has a twofold goal. Firstly, we provide yet another reason for scepticism. Many studies investigate relations between aggregate, "ill-defined" variables (Spirtes and Scheines 2004). As it turns out, there exist no exogenous instruments on these variables. In such cases, IV estimates are necessarily inconsistent. However, secondly, under plausible assumptions one may at least identify the contexts where instruments may not be exogenous and, as a result, adjust one's expectations on the usefulness of IV methods.