When we make choices, we often rely on the thoughts and experiences of those around us. However, our individual preferences, tastes, and objectives can vary significantly. A research team explored how we learn from others despite these personal differences. They discovered that people typically perceive social information as suggestions, using it with a degree of caution. This approach helps them avoid the costs associated with extensive exploration. The findings suggest new ways to implement similar learning methods in artificial intelligence (AI).
Picture yourself in a new city, ready for dinner. How do you select a restaurant? You might check online reviews and opt for the highest-rated place, but how can you be certain that those reviewers have similar tastes, spice preferences, or budget constraints as you? Moreover, how do humans generally learn from the opinions of others when individual preferences can differ significantly?
Utilizing Social Information
Previous research on how individuals learn from others primarily examined situations where everyone had the same goals and preferences. However, in reality, this is seldom the case. A recent study, published in the scientific journal Proceedings of the National Academy of Sciences (PNAS), addresses this gap by analyzing how individuals use social information to make decisions when preferences among peers are not perfectly aligned. This research was conducted by scientists from two prominent German research clusters: the Cluster Machine Learning at the University of Tübingen and the Cluster Collective Behaviour at the University of Konstanz, along with collaborators from RIKEN (Japan) and the University of St Andrews (UK).
The researchers designed an online experiment resembling a video game to explore this concept. The game simulated everyday decision-making scenarios. Participants played in groups of four, each with an individual goal that was unique yet similar to the others. Throughout the game, they could observe their peers’ progress.
Using Social Information as a Decision-Making Resource
The study’s findings indicate that, even in this context, individuals employ social information to influence their decisions while maintaining a degree of skepticism. Participants regarded social cues as less trustworthy than their own gathered information, yet they adeptly adjusted them to fit their individual situations. To elucidate this behavior, the researchers proposed a new model for social generalization, which proved more effective than numerous earlier models in predicting behavior. “Our model suggests that social information should be integrated similarly to individual knowledge rather than being copied uncritically,” explains lead author and PhD student Alexandra Witt.
Using this model, the researchers demonstrated that people leverage social information as an exploration strategy. Individual exploration can be cognitively demanding and risky. When participants had access to social information, they utilized it to inform their choices, thereby avoiding the expensive process of individual exploration. “The concept that social learning can serve as a guide for exploration is not novel,” states Wataru Toyokawa, a co-author of the study and former researcher at the University of Konstanz, now leading a group at RIKEN. “However, our results not only reinforce this idea but also expand and generalize the theory for diverse human societies.”
The Importance of This Research
“While recent advancements show the capabilities of Artificial Intelligence, it still falls short of learning socially as effectively as humans do,” notes senior author Charley Wu, leader of the Human and Machine Cognition Lab at the University of Tübingen. “Our ability for social and cultural learning has been crucial to the success of humankind. Gaining a better understanding of this capacity could enable us to apply similar concepts to AI, in areas like virtual assistants or recommendation systems.” In summary, social learning is one of humanity’s most significant tools, and this research brings us a step closer to comprehending this remarkable capability.
