Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster D in Poster Session D: Wednesday, August 5, 2:00 – 3:45 pm, Kimmel Center, Shorin & Rosenthal Rooms

How do we learn what to believe and which sources to trust?

Prashanti Ganesh¹, Frederike H. Petzschner¹, David Levari¹, Noham Wolpe², Matthew Nassar¹; ¹Brown University, ²Tel Aviv University, University of Cambridge

Presenter: Prashanti Ganesh

In daily life, media exposes us to information ranging from personal matters to global issues. Encountering a news piece may prompt belief updates about the source ("Is it credible?"), the topic ("Is the new law effective?"), or both. These updates may not be independent: a source's credibility may shape how we interpret its claims, and those claims could color our assessment of the source. To capture this interdependence, we developed a Bayesian model that jointly updates both sets of beliefs. Simulations show that belief-aligned information boosts perceived credibility and trust across topics, illustrating how credibility transfers and shapes belief updating about other topics. These simulations gave two predictions: (1) belief-consistent information increases perceived credibility regardless of true reliability, while inconsistent information reduces it (H1); and (2) increased perceived credibility amplifies belief updating, while reduced credibility attenuates learning (H2). As mechanisms of real-world belief updating remain relatively unknown, we developed a news-based inference task using authentic media content to track participants' beliefs about topics and sources. We use high and low levels actual credibility of a source and alignment. Our preliminary results (N = 30) show that sources aligning with participants' initial beliefs were perceived to be more credible in comparison to a low alignment source, irrespective of its actual credibility (H1). Next, people updated their beliefs to a lesser extent when they were being informed by a low alignment source than a high alignment source (H2), although the difference was not significant. Overall, we introduce a modeling framework that, while grounded in rational principles, captures the loop that arises when individuals jointly infer source credibility and topic-related beliefs, potentially resulting in biased belief formation, and our preliminary findings suggest that belief updating reflects an interaction between content and perceived source credibility, with alignment potentially influencing how both beliefs are updated.

Topic Area: Memory, Learning & Knowledge Structures