Study of mental load measured by pupillary diameter during a causal diagnostic probability estimation task

Jean-Louis STILGENBAUER, Researcher, University Paris 8, CHArt Laboratory, France

Causal reasoning reveals the causal structure of the systems that surround us, and is therefore fundamental to an intimate understanding of the world. Here, we focus on an elementary form of diagnostic reasoning that consists in estimating the chances that a particular cause triggers a certain type of effect. From a formal point of view, this involves a reasoner estimating the diagnostic conditional probability Pr(cause|effect). Baratgin & Stilgenbauer, (2020) and Stilgenbauer & Baratgin, (2018), showed that individuals can follow two main strategies to estimate this probability. The first is to perform an uncertain inference that takes the form of a defeasible Modus Ponens (DMP): effect; if effect then probably cause; therefore probably cause. The second possible strategy consists in forming an abductive reasoning that relies on a defeasible version of the Consequent Assertion Pattern (CAD): effect; if cause, then probably effect; therefore probably cause. Recent results (Stilgenbauer, Baratgin, & Douven, 2017; Stilgenbauer & Baratgin, 2019) suggest that the estimate of Pr(cause|effect) is overall more accurate when individuals reason with DMP than when they use DAC. In this research, we attempt to understand this result and hypothesize that estimating diagnostic probability through DMP facilitates the mobilization of cognitive resources needed to complete the task. Conversely, we believe that the use of DAC prematurely blocks the recruitment of these resources. To test this hypothesis, we measured the pupillary diameter of two groups of participants tasked with estimating a diagnostic probability value. Variations in pupillary diameter are indeed known to correlate with cognitive effort (Gavas, Tripathy, Chatterjee, & Sinha, 2018; Hess & Polt, 1964). In our experiment, one group of participants were prompted to produce an estimate through DMP-like reasoning and a second group were prompted to reason according to DAC. The results confirm our hypothesis, on average the pupillary diameters of the participants are larger in the DMP group than in the DAC group. These new data suggest that the diagnostic probability estimation strategy (DMP vs. DAC) modulates more or less strongly the amount of cognitive resources engaged during a diagnostic reasoning activity.