Surprise! Minimally-Counterintuitive Stimuli Trigger Greater Curiosity than Merely Improbable Stimuli

• In Event: The Dynamics of Children’s Information Search

Abstract

Violations of expectations can trigger curiosity in both children and adults (Berlyne, 1966; Stahl & Feigenson, 2015). Such violations can differ in kind as well as degree - for instance, an event could be merely improbable or violate an intuitive theory (Shtulman & Carey, 2007; Stahl & Feigenson, 2019). Theories of curiosity have posited an inverted U-shaped function, such that curiosity is maximally piqued by moderately surprising stimuli (Loewenstein, 1994), but this hypothesis has not been explored for different kinds of violations. Drawing upon work on the cognitive science of religion (Banerjee, Haque, & Spelke, 2013; Boyer & Ramble, 2001), we explore the hypothesis that minimally counterintuitive stimuli - those that involve a single violation of an intuitive theory - will be especially effective at triggering curiosity.
We recruited 77 adults via mTurk and 36 children through a research database (18 4-year-olds, 18 5-year-olds) and presented items that were ordinary, statistically unlikely (e.g., “a rose that can grow to be more than five feet tall”), minimally counterintuitive (MCI, e.g., “an alligator that can talk just like a human can talk”), and very counterintuitive (VCI). Items were presented in pairs (Ordinary vs. Unlikely, Unlikely vs. MCI, MCI vs. VCI), with three trials for each contrast type. Participants were asked which in each pair they would like to learn more about (with half of adults explicitly asked to report their “curiosity”).
We fit multilevel models and compared them to reduced models using likelihood ratio tests, collapsing adults’ question type (learn vs. curious), which did not affect responses. The best-fitting model included contrast type and age group as predictors, but not their interaction. Participants were more likely to choose Unlikely over Ordinary (see Fig. 1; OR=7.95, 95% CI [6.67, 9.48], p<.001), and this preference for the more surprising item was not significantly different from the MCI vs. Ordinary contrast (OR=1.03, 95% CI [0.83, 1.27], p=0.81), but was significantly different from VCI vs. MCI (OR=0.68, 95% CI [0.55, 0.84], p<.001). Consistent with these results, adults chose Unlikely over Ordinary (M=2.10 out of 3, t=6.01, p<.001) and MCI over Unlikely (M=2.12, t=6.09, p<.001), but not VCI over MCI (M=1.64, t=1.37, p=.17), more often than chance. Children only chose MCI over Unlikely more often than chance (M=1.83, t=2.58, p=.01; Unlikely vs. Ordinary: M=1.78, t=1.74, p=.09; VCI vs. MCI: M=1.56, t=.41, p=.68). These effects were driven by older children: age in months had a significant effect on children’s likelihood of choosing the more surprising item across contrast types (b=0.03, 95% CI [0.00, 0.05], p=.02).
Our results suggest that (i) minimally counterintuitive items trigger greater curiosity than merely unlikely items, (ii) surprisingness has diminishing returns, consistent with an inverted U-shaped function, and (iii) sensitivity to surprisingness increases with age, appearing in our task by age 5.

Authors

• Casey Lewry, Princeton University
  Presenting Author

• Sera Gorucu, Princeton University
  Non-Presenting Author

• Emily G Liquin, Princeton University
  Non-Presenting Author

• Tania Lombrozo, Princeton University
  Non-Presenting Author