The effect of difficulty manipulation on confidence in causal strength judgments

Abstract

Petrusic and Baranski (2009) concluded that confidence is a general property of human judgment. However the claim may be premature, as there is a lack of enquiry into how confidence affects causal strength judgements. Within this domain, a complete account of empirical estimates of causal strength has remained a challenge for the two dominant normative models, namely; the co-variational model, and Power PC theory. But the conflation hypothesis, posited by Liljeholm and Cheng (2009), suggests that the co-variational model may actually be an account of participants' conflating confidence with estimates of causal strength; if probed separately, and given a less ambiguous task scenario, the two should be dissociable, and empirical strength estimates should accord with Power PC. Although, if confidence is a general property of human judgement, the conflation hypothesis could be extended to incorporate the most robust finding within the broader literature on decisional confidence - the difficulty (hard-easy) effect: the general tendency for participants’ subjective estimates to exceed objective probability for hard decisions, and to underestimate objective probability for easy decisions. Our aim is to assess the predictions of the two normative accounts of causal strength estimates, as well as formulate hypotheses for how the difficulty effect may effect causal strength judgements, using a modification of a previous causal strength task. We were able to remove a prior bias in previous tasks, and while the two models could partially explain different aspects of strength responses, participants' consistently rated the strength of preventive stimuli higher than generative stimuli, particularly in smaller sample sizes when these judgements were more difficult to make. We suggest this constitutes evidence of the difficulty effect in empirical causal strength judgements, and suggest how the finding may be related to more recent Bayesian models of strength judgement.