Humans exhibit a remarkable ability to learn relational concepts from a small number of examples. On the Synthetic Visual Reasoning Test (SVRT), a collection of 23 problems that require learning relational concepts, people typically discover the relational rules from a handful of examples. An important question is what learning mechanisms underlie human ability to acquire relational concepts so quickly. Here, we examine whether analogical comparisons serve as an efficient strategy to facilitate learning with small samples. We conducted two experiments that varied the display format and problem order for the SVRT. When problems are presented in an easy-to-hard order, people learn more efficiently when prior examples are displayed in spatially-segregated sets, consistent with the use of analogical mapping as a learning strategy. However, when the problems are presented in a random order, the advantage of spatially-segregated displays is eliminated. We propose that when hard problems are encountered early in a problem sequence, analogical mapping becomes too demanding, causing people to fall back on a less efficient learning strategy that does not require the comparison of multiple examples.


Human Relational Concept Learning on the Synthetic Visual Reasoning Test

Competence and morality are two of the most important dimensions in social evaluation. Recent studies have suggested the primacy of morality, showing that information about immorality of an ordinary target person decreases evaluation of their competence. We examined the effect of moral taint on multiple non-moral judgments: ratings of the competence, accomplishment, and contribution of fictitious professionals who were described as highly successful in various fields. Moral taint significantly decreased participants’ non-moral social evaluations of professionals regardless of their field. Mediation analyses showed that the negative impact of immoral character on competence judgments is more strongly mediated by the decrease in participants’ psychological involvement with the target, rather than a decrease in perceived social intelligence of the target. These findings suggest that motivation to distance oneself from immoral others plays a critical role in the revision of social evaluations.

Disgraced Professionals: Revelation of Immorality Decreases Evaluations of Professionals’ Competence and Contribution

The QAnon conspiracy posits that Satan-worshiping Democrats operate a covert child sex-trafficking operation, which Donald Trump is destined to expose and annihilate. Emblematic of the ease with which political misconceptions can spread through social media, QAnon originated in late 2017 and rapidly grew to shape the political beliefs of millions. To illuminate the process by which a conspiracy theory spreads, we report two computational studies examining the social network structure and semantic content of tweets produced by users central to the early QAnon network on Twitter. Using data mined in the summer of 2018, we examined over 800,000 tweets about QAnon made by about 100,000 users. The majority of users disseminated rather than produced information, serving to create an online echochamber. Users appeared to hold a simplistic mental model in which political events are viewed as a struggle between antithetical forces—both observed and unobserved—of Good and Evil.

Rise of QAnon: A Mental Model of Good and Evil Stews in an Echochamber

Is analogical reasoning a task that must be learned to solve from scratch by applying deep learning models to massive numbers of reasoning problems? Or are analogies solved by computing similarities between structured representations of analogs? We address this question by comparing human performance on visual analogies created using images of familiar three-dimensional objects (cars and their subregions) with the performance of alternative computational models. Human reasoners achieved above-chance accuracy for all problem types, but made more errors in several conditions (e.g., when relevant subregions were occluded). We compared human performance to that of two recent deep learning models (Siamese Network and Relation Network) directly trained to solve these analogy problems, as well as to that of a compositional model that assesses relational similarity between part-based representations. The compositional model based on part representations, but not the deep learning models, generated qualitative performance similar to that of human reasoners.

Visual Analogy: Deep Learning Versus Compositional Models

Explicit similarity judgments tend to emphasize relational information more than do difference judgments. We propose and test the hypothesis that this asymmetry arises because human reasoners represent the relation different as the negation of the relation same, so that processing difference is more cognitively demanding than processing similarity. For both verbal comparisons between word pairs, and visual comparisons between sets of geometric shapes, we asked participants to select which of two options was either more similar to or more different from a standard. On unambiguous trials, one option was unambiguously more similar to the standard; on ambiguous trials, one option was more featurally similar to the standard, whereas the other was more relationally similar. Given the higher cognitive complexity of assessing relational similarity, we predicted that detecting relational difference would be particularly demanding. We found that participants (1) had more difficulty accurately detecting relational difference than they did relational similarity on unambiguous trials, and (2) tended to emphasize relational information more when judging similarity than when judging difference on ambiguous trials. The latter finding was captured by a computational model of comparison that weights relational information more heavily for similarity than for difference judgments. Our results provide convergent evidence for a representational asymmetry between the relations same and different.


Keith Holyoak

5

1

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES