WASHINGTON (Oct. 24, 2022) – Learning about how we learn about other people can be as complicated as, well, people. Existing social learning models often fall short because they typically rely on one-dimensional feedback to describe how people update their notions about others over time. Now, a new study published in Nature Communications by an international research team from the George Washington University and the University of Heidelberg introduces a new computational modeling framework that describes how people learn about others and how they rely on previous knowledge during learning, depending on whether people are more or less similar to them.
The research team conducted a series of behavioral experiments in which participants rated the personality traits of several people they never met. The participants then received feedback about the unknown people’s self-ratings on these personality traits, thereby learning about them as the task went on.
The researchers used a set of mathematically specified models to describe how participants learned during the task. Some models took preexisting social knowledge structures into account. These knowledge structures could differ based on two components: granularity and reference points. Granularity refers to how coarse or fine-grained associations are between different types of knowledge. The better we know someone or the more similar someone is to people we know, the more fine-grained our representations become. Reference points are snippets of prior knowledge, initial estimates based on us or our peer group. We use these as a starting point to learn about the new people we meet. When participants learned about new people, they relied on both reference points and more or less granular social knowledge representations. Over time, these concepts guided learning to allow participants to generalize from specific traits to holistic impressions of people.
FROM THE RESEARCHERS
“In this paper, we introduce a robust and quantifiable framework for social learning that describes when and why people actively learn about another person vs. fall back on their knowledge about similar others or even stereotypes. While in this study we focused on personality traits, the framework can be broadly applied to shed light on how people learn about other types of information, such as preferences, beliefs, or political and social views.” – Gabriela Rosenblau, assistant professor of cognitive neuroscience, George Washington University
“First, we find that when learning about others’ personalities humans use a rather complex set of strategies that rely on prior information (e.g., self-ratings or stereotypes) and the innate structures within this information. Second, these results stem from computational models that cast a new look on human learning. In general, we believe that these models can offer a much broader look on learning in general.” – Koen Frolichs, doctoral student of social neuroscience, University of Heidelberg
“We identified a set of models that describe social learning at different levels of complexity. We are now investigating whether these models help us to better understand the social deficits and difficulties of people suffering from psychiatric disorders.” – Christoph Korn, assistant professor of social neuroscience, University of Heidelberg
The study, “Incorporating social knowledge structures into computational models,” was conducted at the University Medical Center Hamburg-Eppendorf and the University of Heidelberg. It was funded by the German Research Foundation. To schedule an interview with Dr. Rosenblau, please contact Danny Parra at [email protected].