Imagine you are visiting a new city for the first time, and dinner time rolls around. How do you choose a restaurant? You could just check reviews and go to the highest-rated option. But how can you be sure that the reviewers share your food preferences, your spice tolerance, or your budget? And how do humans in general manage to learn from others when preferences can vary quite substantially between individuals?

How we use social information
Until now, much research on how people learn from others has focused on settings where everyone has identical goals and preferences. However, in the real world, that is rarely the case. A new study, now published in the scientific journal Proceedings of the National Academy of Sciences (PNAS), closes this gap by investigating how humans use social information to make decisions when the preferences of those around them are not a perfect match. The research was led by scientists from two German Clusters of Excellence – the Cluster Machine Learning at the University of Tübingen and the Cluster Collective Behaviour at the University of Konstanz – together with colleagues from RIKEN (Japan), and the University of St Andrews (United Kingdom).

To study this phenomenon, the researchers created an online experiment that resembled a video game. The game was designed to mimic everyday decision-making situations. Participants completed the game in groups of four. Each participant had a goal, which was individually unique, but similar to that of the other participants. Throughout the experiment, participants were able to see the progress of their fellow players.  

Social information serves as a decision-making aid
The results show that, even in this setting, humans use social information to guide their decisions, but also take these cues “with a grain of salt”. In the experiment, they treated social information as less reliable than information they collected themselves, but still flexibly adapted it for their own circumstances. To explain this phenomenon, the researchers introduced a new social generalization model, which outperformed a number of other models from previous theories in predicting behaviour. “Unlike models from the previous literature, our model assumes that social information should be integrated similarly to individual information, rather than blindly copied”, explains lead author and PhD student Alexandra Witt.

With the help of this model, the researchers showed that humans use social information as an exploration tool. Individual exploration can be costly, both cognitively and in terms of risk. When social information was available, participants relied on it to guide their choices, saving themselves the costly individual exploration process. "The idea that social learning can function as an exploration guiding tool is not new", says Wataru Toyokawa, a co-author of the study, formerly of the University of Konstanz, who is now a group leader at RIKEN. "But what we found not only supported the idea, but helped us extend and generalize the theory to the case of diverse, heterogeneous human societies".

Why does this matter? 
“Although recent advances have demonstrated the power of Artificial Intelligence, it still struggles to learn socially in a similar capacity as humans”, says senior author Charley Wu, who leads the Human and Machine Cognition Lab at the University of Tübingen. “Indeed, it is our extraordinary capacity for social and cultural learning that has played a key role in the success of the human species. A better understanding of this ability could let us incorporate similar principles into AI, such as in virtual assistants or recommendation algorithms”. Ultimately, social learning is one of humanity’s most powerful tools, and this research brings us closer to understanding this impressive ability.

Key facts:

Publication: Witt, A., Toyokawa, W., Lala, K. N., Gaissmaier, W., & Wu, C. M. (2024). Humans flexibly integrate social information despite interindividual differences in reward. Proceedings of the National Academy of Sciences, 121(39), e2404928121.
https://doi.org/10.1073/pnas.2404928121

• The research was led by doctoral researcher Alexandra Witt and group leader Charley Wu from the University of Tübingen. Wu is a member of the Cluster of Excellence Machine Learning and the Tübingen AI Center. Additional researchers include Wolfgang Gaissmaier, professor of social psychology and decision sciences at the University of Konstanz and PI at Konstanz’s Cluster of Excellence Collective Behaviour, Wataru Toyokawa, formerly at the Cluster of Excellence Collective Behaviour, now at RIKEN, Japan, and Kevin Lala, professor of behavioural and evolutionary biology at the University of St Andrews (United Kingdom).
• The Cluster of Excellence Centre for the Advanced Study of Collective Behaviour at the University of Konstanz is a world-leading research centre for the study of swarm behaviour. It tackles pressing questions across species and organizational levels in an interdisciplinary manner, from neuronal mechanisms to individual perception and preferences to collective behaviour in groups or entire societies.
• The aim of the Cluster of Excellence “Machine Learning: New Perspectives for Science” at the University of Tübingen is to enable machine learning to take a central role in all aspects of scientific discovery and to understand how such a transformation will impact the scientific approach as a whole. 

Note to editors:

You can download photos here:

1) https://www.uni-konstanz.de/fileadmin/pi/fileserver/2024_EXTRA/nicht_blind_folgen_02.jpg

Caption: Professor Wolfgang Gaissmaier is a professor of social psychology and decision sciences at the University of Konstanz and PI at Konstanz’s Cluster of Excellence Collective Behaviour.

Copyright: University of Konstanz / Inka Reiter

2) https://www.uni-konstanz.de/fileadmin/pi/fileserver/2024_EXTRA/nicht_blind_folgen_01.jpg

Caption: Dr. Wataru Toyokawa, formerly at the Cluster of Excellence Collective Behaviour, now at RIKEN, Japan.

Copyright: private

3) https://www.uni-konstanz.de/fileadmin/pi/fileserver/2024_EXTRA/nicht_blind_folgen_03.jpg

Caption: Lead author Alexandra Witt is PhD student at the University of Tübingen.

Copyright: private

4) https://www.uni-konstanz.de/fileadmin/pi/fileserver/2024_EXTRA/nicht_blind_folgen_04.jpg

Caption: Dr. Charley Wu is a member of the Cluster of Excellence Machine Learning and the Tübingen AI Center.

Copyright: Tübingen AI Center / Elia Schmid.