16:30 - 18:00
Mon-B16-Talk III-
Mon-Talk III-
Room: B16
Chair/s:
Julia Groß
Is a Simpler Explanation Preferred Even When It Doesn’t Fit The Data? An Attempt To Measure Simplicity Bias Experimentally
Mon-B16-Talk III-04
Presented by: Mateja Manojlovic
Mateja Manojlovic 1, 2, Kaja Damnjanovic 2, 3
1 University of Belgrade, Faculty of Organizational Sciences, Department of Human Resources Management, 2 University of Belgrade, Faculty of Philosophy, Department of Psychology, Laboratory for Experimental Psychology, 3 University of Belgrade, Faculty of Philosophy, Institute of Philosophy
Simplicity bias (SB) is the tendency to prefer simple over complex explanations when reasoning, ignoring the explanatory power of that explanation. Preferences for simpler explanations often match the higher prior probability of a model. One research line interprets simplicity as a cue for a higher probability of a model when the probability information is omitted. On the other hand, within the heuristics and biases approach to human cognition, it is well established empirical fact that reasoning about probabilities violates formal norms of rationality. To explore if the SB can be defined as another cognitive bias, we proposed to set comprehensiveness of explanation as a norm, and aimed to provide experimental evidence that people deviate from this norm, when manifest the SB. Participants were presented with 30 SB tasks that included relations between causes and consequences and then asked to choose between the offered simple or complex explanations. SB tasks were divided into two subsets that differed in comprehensivness. In the standard situation (SS) subset, both simple and complex explanations were comprehensive, and in the conflict situation (CS) subset only complex explanations were comprehensive. The participants’ scores were counted as a proportion of simple explanations chosen in SS and CS separately. The one-way repeated-measures ANOVA reveals that participanta chose simple explanations significantly more frequently in SS than in CS (F(1,194)=148.344, p< .01, η2= .43). This result doesn’t directly support our hypothesis, though it can be discussed regarding the possibility that SB tasks elicited type 2 of thinking in subjects.
Keywords: simplicity bias, reasoning, judging, cognitive biases