Skip to main content
eScholarship
Open Access Publications from the University of California

UC Irvine

UC Irvine Electronic Theses and Dissertations bannerUC Irvine

Insights into Problem Solving, Algorithm Aversion, and Theory of Mind

Creative Commons 'BY' version 4.0 license
Abstract

This dissertation explores several important topics in the cognitive sciences: Insight, algorithm aversion, and theory of mind. First, I tackle the challenge of understanding insight, or the “aha!” experience. In Chapter 1, I use a well-defined class of problems (compound remote associates: Bowden and Jung-Beeman, 2003; Mednick, 1962) to test if various lexical and morphological properties affect solution retrieval and the likelihood of insight. While performance is only affected by one property (familiarity), other findings contest popular assumptions about insight. Namely, the reported magnitude of insight decreases with trial time (challenging the impasse hypothesis) and increases with the number of cues solved (challenging the all-or-none hypothesis). In Chapter 2, I introduce a new insight problem task: Joke completion. I find that performance and magnitude of insight within it correlate with an established task (rebus puzzles: MacGregor and Cunningham, 2009), though the distribution of reported insight is not bimodal, as was expected. Further, self-estimated and externally-rated joke funniness correlate with reported insight. Lastly, performance and reported insight decrease with trial time, again refuting impasse. In Chapter 3, I shift focus to a more recent concern: Algorithm aversion. As AI becomes an integral part of our daily lives, it is crucial to identify and anticipate biases regarding it. Since aversion mostly occurs in subjective contexts, I test whether people find jokes less humorous if they believe an AI created them. When joke source is ambiguous, people exhibit bias toward jokes they identify as human-created. However, this bias disappears when the (purported) source of jokes is stated. This demonstrates that such biases are weaker than proposed and are dependent on framing. In Chapter 4, I conclude by exploring another perennial topic: Theory of mind. Namely, I examine whether a few words provide an accurate estimate of another person's domain knowledge. This was done by having one group of people ("informants") describe images depicting various domains (e.g., video games, astronomy), then having a second group ("evaluators") make pairwise comparisons between these informants regarding who they believe is more knowledgeable, based on these descriptions. Strikingly, evaluators perform above chance at identifying the more knowledgeable informants when only one description is available (around seven words, on average). Further, the most knowledgeable informants produce the most specific facts and the most knowledgeable evaluators are the most sensitive to false information. However, less knowledgeable evaluators treat specific statements interchangeably, regardless of their factuality. These results show the inferential power a mere few words hold.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View