The talk was based on his paper What Causal Illusions Might Tell us about the Identification of Causes, written jointly with Robert Thorstad. They argue that causal judgements are based on two kinds of process:
- ‘A fast and intuitive process that identifies potential causes on the basis of perceptual cues’ – particularly, observed correlation, and
- ‘A slow and reflective process that identifies potential causes on the basis of causal mechanisms.’
They call this the Dual-Process Hypothesis of Causal Identification.
Several experiments support this hypothesis. In one, the Jedi powers experiment, participants were in a lift (elevator) and were subjected to a ‘causal illusion’: a man in the lift appeared to repeatedly open the lift doors by gesturing with his hands. Unbeknownst to the participants, there was someone hidden from view outside the lift who was opening the doors by repeatedly pushing the call button. The participants initially judged that the man caused the doors to open, apparently by observing the correlation between the hand gestures and the doors opening. These judgements weakened, plausibly as doubts about the mechanism crept in.
Perhaps causal judgements in daily life can tell us something about causal judgements in medicine. In comparison to mechanism discovery, conducting a clinical study is quick and easy. So, observed correlations are easy to come by. The temptation is to attribute causation just on the basis of clinical studies. This needs to be tempered by the slow, more reflective process of mechanism evaluation. Otherwise we may be too easily convinced of jedi powers and snake oil, and too slow to identify real effectiveness.