CENTER FOR RESEARCH IN
COGNITION & NEUROSCIENCES/ConnexConnex

JOIN US ONLINE: https://bit.ly/connex-seminar-ik

Beliefs constitute subjective representations of the world, shaped by both perception and the attribution of personal meaning. Far from being static, they are dynamic constructs that the brain continuously updates in light of new information. Through this process, internal models of reality are refined to guide behavior and maintain coherence between expectations and experience. Context plays a crucial role in this dynamic, influencing how beliefs are formed, maintained, and revised as new evidence emerge.

Here, we leveraged two distinct forms of contextual cues to examine how they shape beliefs and consequently decision-making: experimental verbal suggestions about the effect of a placebo drink, and real-world uncertainty exemplified by the COVID-19 pandemic. Together, these approaches allowed us to assess how context-dependent information influences internal representations and, in turn, the cognitive and neural mechanisms supporting adaptive decision-making.

We hypothesized that context shapes both the valuation of incoming information in light of prior beliefs and its integration into updated representations, allowing the brain to minimize mismatches between what is expected and what is experienced.

Combining behavioral, neuroimaging, and computational approaches we showed that:

Suggestions about hunger shaped both subjective hunger sensations and food wanting during dietary decision-making. These effects were underpinned by increased activation in valuation and attentional brain regions, and a stronger weighting of taste attribute in participants expecting to be hungrier. In contrast, expecting to be less hungry was associated with a greater weighting of health attribute in decisions and modulation of valuation regions by cognitive control areas associated with the relative influence of health versus taste information.

The COVID-19 pandemic disrupted the typical optimism bias that characterizes belief updating about future adverse life events under normal circumstances. Outside the pandemic, participants integrated positive information more readily than negative information and showed a reinforcement learning–like pattern of belief updating. During the pandemic, this valence-dependent asymmetry disappeared, and belief updating became more Bayesian.

Collectively, our findings demonstrate that belief systems are inherently dynamic, as contextual information continuously shapes decision-making processes. Belief formation can be influenced by contextual cues such as verbal suggestions, which generate expectations that shape the valuation stage of decision-making. Likewise, belief updating is modulated by contextual uncertainty, affecting the learning processes in decision-making. These results underscore the bidirectional relationship between beliefs and context: prior beliefs shape the interpretation of contextual cues, and

context, in turn, influences belief revision — both processes working jointly to minimize mismatches between expectation and experience. This flexibility of the belief system enables adaptive decision-making in an ever-changing environment.