How Does Explanatory Virtue Determine Probability Estimation?—Empirical Discussion on Effect of Instruction

Abstract

It is important to reveal how humans evaluate an explanation of the recent development of explainable artificial intelligence. So, what makes people feel that one explanation is more likely than another? In the present study, we examine how explanatory virtues affect the process of estimating subjective posterior probability. Through systematically manipulating two virtues, Simplicity—the number of causes used to explain effects—and Scope—the number of effects predicted by causes—in three different conditions, we clarified two points in Experiment 1: (i) that Scope's effect is greater than Simplicity's; and (ii) that these virtues affect the outcome independently. In Experiment 2, we found that instruction about the explanatory structure increased the impact of both virtues' effects but especially that of Simplicity. These results suggest that Scope predominantly affects the estimation of subjective posterior probability, but that, if perspective on the explanatory structure is provided, Simplicity can also affect probability estimation.

How Cognitive Biases Affect XAI-assisted Decision-making

“…Existing work on XAI desiderata and psychology of explanations has evidenced that people look for specific qualities in explanations. In our corpus, we also observe such preferences for "broad" [13], [35], [38], "simple" [13],…”

“…Similarly, Kulesza et al [41] argue that more comprehensive explanations help to significantly improve participants' mental models, but other work [42], [43] found complete explanations can lead to overreliance [35]. In [35], [38], the authors contend that coherent and broad explanations are preferred, with scope being even more important than simplicity, consistently with Lombrozo's point of view in cognitive science that broader and simpler explanations are better [24]. Based on these findings, it can be challenging to gauge the right level of complexity in explanations.…”

“…Our corpus also contains articles discussing under reliance issues. These were manifested through various effects, such as "the escalation of commitment" [60], the "illusion of validity" [61] or "perceived goal impediment" [59], which concerned mainly domain experts, and "negativity bias" [11], [17], [38], [40], [62]. Several works have highlighted the role of user expertise in under reliance problems.…”

Simplicity as a Cue to Probability: Multiple Roles for Simplicity in Evaluating Explanations

“…Furthermore, these simplicity preferences might be stronger in cases where posteriors are more uncertain or difficult to compute, such that people have greater need to rely on alternative cues. Indeed, in previous work, effects of simplicity are not reliably found when explanations’ posterior probabilities are explicitly provided or trivial to compute (Bonawitz & Lombrozo, 2012; Lagnado, 1994; Lombrozo, 2007; Shimojo, Miwa, & Terai, 2020).…”

Revisiting the Narrow Latent Scope Bias in Explanatory Reasoning

“…Psychological research suggests that these explanatory virtues are shared by lay people (Lagnado, 1994;Lombrozo, 2007;Pacer & Lombrozo, 2017;Read & Marcus-Newhall, 1993;Shimojo, Miwa, & Terai, 2020;Vrantsidis & Lombrozo, 2022), even by children as young as five years old (see, e.g., Bonawitz & Lombrozo, 2012;Corriveau & Kurkul, 2014). A typical test situation revealing a preference for explanatory simplicity (Lombrozo, 2007) is abstractly summarized in Fig.…”