Social Cognition: Wisdom Of The Crowd

Many organizational decisions require knowledge of how people move around in space. Consider, for instance, managing a crowd during events, building safe infrastructure for pedestrians, and using behavioral interventions to limit infection risk during epidemics. It is therefore not surprising that engineers have spent considerable effort on capturing pedestrian flow. However, while these models are aimed at capturing human movement, they typically lack the human component and fail to incorporate individual differences in people’s walking behavior and the goals that underly this behavior. In this talk, I will present the recently developed Minds for Mobile Agents (M4MA) model as an alternative framework that allows for pedestrians to have their own personality and goals, thus capturing more realistic walking behavior than competitor models. I will furthermore present results of ongoing studies that attempt to validate the M4MA model.

This is an in-person presentation on July 22, 2024 (10:00 ~ 10:20 CEST).

How do we rapidly perceive a crowd’s general direction of movement or the average color of a picture? Ensemble perception is the automatic and rapid extraction of the summary statistical properties of collections of stimuli. Ensemble perception is surprisingly accurate given people’s recall for individual stimuli. Despite a large empirical literature on ensemble perception, a comprehensive cognitive model of the process is lacking. Here, building on the efficient coding literature, we develop and explore a model of ensemble perception. In the model, people are assumed to behave as if they first infer a CDF of a set of stimuli using kernel density estimation and then use this inferred CDF to make ensemble judgments. We call this model Distribution Inference and Surface Tracing (DIST). We show that DIST accurately predicts central tendency estimation in the face of changes in the size, variance, and skewedness of a stimuli set, even when constrained to parameters that are resource-rational. Further results are presented that show how DIST performs when fitted to 12 different datasets. DIST is found to fit equivalently to or better than alternative models of ensemble perception, including recent computational models. We also present experiments that aim to test a novel prediction that DIST makes about ensemble perception behaviour. The results of the experiments are taken as preliminary support for DIST.

This is an in-person presentation on July 22, 2024 (10:20 ~ 10:40 CEST).

Contrary to common intuition, groups of people recalling information together remember less than the same number of individuals recalling alone (i.e., the collaborative inhibition effect). To understand this effect in a free recall task, we build a computational model of collaborative recall in groups, extended from the Context Maintenance and Retrieval (CMR) model which captures how individuals recall information alone (Polyn, Norman, & Kahana, 2009). We propose that in collaborative recall, one not only uses their previous recall as an internal retrieval cue, but also listens to someone else's recall and uses it as an external retrieval cue. Attending to this cue updates the listener's context to be more similar to the context of someone else's recall. Over an existing dataset of individual and collaborative recall in small and large groups (Gates, Suchow, & Griffiths, 2022), we show that our model successfully captures the difference in memory performance between individual recall and collaborative recall across different group sizes from 2 to 16, as well as additional recall patterns such as recency effects and semantic clustering effects. Our model further shows that the contexts of collaborating individuals converge more than the contexts of individuals who recall alone. We discuss contributions of our modeling results in relation to previous accounts of the collaborative inhibition effect.

This is an in-person presentation on July 22, 2024 (10:40 ~ 11:00 CEST).

The wisdom of the crowd aims to aggregate people's knowledge. One role psychological models can play is debiasing people's expressions of their knowledge. An example involves debiasing estimates people provide that may have been impacted by anchoring effects. Because of anchoring caused by initial comparison, what people estimate is a systematic distortion of what they know. Cognitive models can serve to infer the underlying knowledge from the observed behavior. Aggregating inferred cognitive knowledge can then lead to better group aggregates than can be achieved by statistical aggregation of the estimates. We consider data from a new experiment in which 194 people answered questions like "What year did the first McDonalds franchise open?" and "What year did Ronald Reagan become president?". Some people were asked initial comparison questions involving years much later than the answer, some were asked comparisons involving years much earlier than the answer, and others were not asked any comparison question. We show these manipulations lead to significant anchoring effects, and test the ability of a simple cognitive model of anchoring to debias people's estimates and improve the crowd aggregate.

This is an in-person presentation on July 22, 2024 (11:00 ~ 11:20 CEST).