The Role of Individual Differences in Collective Behaviour
Final Report Abstract
Social interactions often result in emergent collective dynamics which play a key role in both animal and human decision making. These dynamics have often been investigated assuming over-simplified interaction mechanisms and very similar-behaving agents. The characteristics of interacting agents are, however, expected to greatly depend on individual differences in their performance, preferences and social learning strategies. This project, therefore, investigated the role of these individual differences on social dynamics by extending established theories and frameworks from individual decision making into the social realm across four subprojects. In subproject 1 we investigated the roots of individual differences in performance and how this, in turn, affected the use of social information and the collective outcomes. Building on the Lens model we analyzed how participants categorize images based on cues, the meaning of which had to be learned from feedback. Over the course of the experiment, participants developed different beliefs about the cue meanings. This diversity in cue beliefs was, in turn, an important factor for improving the quality of social information they exchanged among each other. Participants, however, failed to harvest the full potential of this diversity because they relied too heavily on their personal information. Simulating different social learning strategies, we show that this strong reliance on personal information impedes participants from benefiting from this diversity. The role of differences in performance can, thus, be only understood in interaction with individuals’ social learning strategies. Subproject 2 builds on this finding by exploring in more detail the social learning strategies individuals use. In an estimation task, participants were first asked to provide an independent estimate, after which they observed the estimates of others. By investigating their behaviour with a model encompassing a cognitive toolbox of social learning strategies, we showed that participants differed in their social learning strategies. Furthermore, we show that the incorporation of others’ opinions strongly depended on how consistent those opinions were with an individual’s own opinion and the agreement of the opinions among themselves. These results contribute to our understanding of how individuals’ strategies shape opinion and information dynamics in social systems. In subproject 3a we focused on interactions where multiple individuals made decisions sequentially as such dynamics are expected to be driven by differences in performance in interaction with social learning strategies. To shed light on the role of these characteristics, we developed the social drift-diffusion model (social DDM). We showed that correct information spreads when more confident participants are more accurate and decide faster. Under these conditions, later-deciding participants were likely to adopt social information and thereby amplified the correct signal provided by early-deciding participants. The social DDM successfully captured the key dynamics observed, revealing the cognitive underpinnings of information spread in social systems. We used these insights in subproject 3b to investigate the role of individual differences in preference for a particular option (a so-called response bias) and the degree of selfishness. Here, we combined an agent-based version of the social DDM with an evolutionary algorithm, allowing the identification of advantageous strategies under asymmetric error costs. We found that large groups should avoid response biases because such biases rapidly amplify in groups. Selfish individuals undermined the group’s performance for their own benefits by developing high response biases and waiting longer for social information. These results show that individuals facing asymmetric error costs need to carefully trade off the expressed response bias and sensitivity to social information to avoid the more costly error but simultaneously benefit from the collective. Taken together, these projects deepened our understanding of social dynamics by explicitly taking into account individual differences in performance, preferences and social learning strategies. Our work highlights the importance of accounting for such differences to predict the emergence of beneficial or detrimental social dynamics.
Publications
- (2018) Individuals fail to reap the collective benefits of diversity because of overreliance on personal information. Journal of the Royal Society Interface 15: 20180155
Tump AN, Wolf M, Krause J & Kurvers RHJM
(See online at https://doi.org/10.1098/rsif.2018.0155) - (2020). Strategies for integrating disparate social information. Proceedings of the Royal Society B 287: 20202413
Molleman L, Tump AN, Gradassi A, Jayles B, Herzog S, Kurvers RHJM & van den Bos W
(See online at https://doi.org/10.1098/rspb.2020.2413) - (2020). Wise or mad crowds? The cognitive mechanisms underlying information cascades. Science Advances 6: eabb0266
Tump AN, Pleskac T & Kurvers RHJM
(See online at https://doi.org/10.1126/sciadv.abb0266) - How social learning strategies boost or undermine decision making in groups. Free University, Berlin, GER, July 2020
Alan Novaes Tump
(See online at https://doi.org/10.17169/refubium-27728)
