Empirical and theoretical investigation of associative learning in groups

University Of Rochester, Rochester NY

Investigators

Abstract

Learning is a central psychological construct that has been studied primarily in individuals. However, humans and other social animals often learn with others, yielding outcomes that are often superior to those obtained by learning alone—a phenomenon known as collective learning. This project investigates the extent to which fundamental principles of associative learning, which operate at the level of individual organisms, contribute to collective learning. Just as individuals learn about correlations in the environment (e.g., rain typically follows dark clouds), they may also learn about correlations between stimuli in the environment and the actions of other individuals (e.g., rain typically follows if you see people carrying umbrellas). The investigators will test a mathematical model that postulates how associative learning in individuals contributes to collective learning using animal model systems. The specific animal models are selected because (a) like humans, they are highly social and learn collectively, (b) they afford the necessary level of experimental control under laboratory conditions to test the proposed mathematical model, and (c) the evolutionary distance between them is large enough to identify general mechanisms of collective learning. The basic rules of collective learning found in these animal groups can be leveraged to promote advantageous aspects of collective learning and to anticipate its detrimental consequences. This project aims to identify fundamental processes of collective learning. A series of associative-learning experiments will test which form of training—individual or collective—results in faster and more durable learning that is resilient to interference from uninformative cues. For each experiment, a prediction will be made based on a simple computational model that assumes that the behavior of other group members can be learned as an informative cue. The generality of the proposed model will be tested on nest-seeking behavior in social insects and food-seeking behavior in social mammals. Aside from their sociality and convenience, specific species were selected as animal models because they lack the higher cognitive sophistication of humans. This makes it easier to uncover the basic associative mechanisms underlying collective learning, without having to control for potential interference by cognitively complex non-associative processes, such as verbal communication. Nonetheless, aspects of the model that are validated with different parameters are expected to generalize to human learning and can be the basis of further examination in humans. These findings will, therefore, aid in determining the factors that can promote or impede collective intelligence when humans or machines (e.g., artificial intelligence) perform tasks as a group. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.