To cooperate or compete: how does our brain decide?

Published by Redbran - 28 days ago - Other Languages: FR, DE, ES, PT

Friends or foes? During social interactions, the ability to infer others' intentions is crucial for predicting their actions and determining whether to cooperate or compete. A new study highlights the brain processes and computational mechanisms involved in this function, known as theory of mind (ToM).

The findings, published in Nature Communications, demonstrate that the brain uses specific algorithms to make these inferences by dynamically arbitrating between intentions of cooperation and competition.

Image by Pixabay

In social interactions, the intentions of other agents (human or artificial) may fluctuate over time between competition and cooperation. This inherent uncertainty in the possible behavior of others makes it extremely difficult to predict social behavior during an interaction. Indeed, unlike most inanimate objects, the observable behaviors of other agents provide only partial information about their likely future behaviors.

Which computational mechanisms allow the brain to effectively predict others' social behavior based on observations of their past behavior? Understanding these mechanisms is important as it remains challenging, for instance, for a social robot to decode a person's intentions and respond accordingly for smooth interactions.

The study determined how the brain adapts to others' fluctuating intentions when neither the nature of the interactions (cooperation or competition) nor the change between these two types of interactions are explicitly signaled.

To achieve these results, scientists placed participants in a scanner to observe real-time brain reactions while they thought they were playing a networked game with another player. The game involved inferring which of two possible cards the other player would choose. In reality, they were playing against an algorithm (AI) that alternated between cooperation and competition strategies without notification (based on the player's past behavior). In the cooperative situation, one of the best strategies was to choose the same card predictably between trials, whereas in the competitive situation, the optimal strategy was to randomly choose between the two cards from trial to trial.

At the core of the brain, a mechanism arbitrating between two cooperative and competitive experts

Scientists used functional magnetic resonance imaging coupled with mathematical modeling to track the participants' reactions based on the AI's behavior. Comparisons between many mathematical models and the observed behavior revealed that one computational mechanism was superior. This mechanism involves arbitration between two experts: one competitive and one cooperative, each weighted by their relative reliabilities.

This model outperforms other learning models in predicting choice behavior. Two brain regions, named the ventral striatum and the ventromedial prefrontal cortex, follow the reliability difference between these experts. Thus, in interactions where others' intentions are fluctuating and not signaled, the brain weighs between a competitive and a cooperative expert by calculating the reliability difference between a competitive and cooperative interaction.

These results help understand the neurocomputational mechanisms by which the brain arbitrates in real-time between cooperation and competition intentions during adaptive social decision-making.

This study provides a mechanistic explanation of how the brain dynamically arbitrates between intentions of cooperation and competition during adaptive social decisions. These novel findings identify the algorithms and brain mechanisms involved in estimating and adapting to others' competitive and cooperative intentions.

This characterization is essential for understanding the brain mechanisms underlying our social interactions, enabling smooth transitions between changing cooperative or competitive contexts. These neurocomputational mechanisms provide the necessary foundation for our social interactions with other natural or artificial agents.

Philippe, R., Janet, R., Khalvati, K. et al. Neurocomputational mechanisms involved in adaptation to fluctuating intentions of others. Nat Commun 15, 3189 (2024).
Page generated in 1.926 second(s) - hosted by Contabo
About - Legal Notice - Contact
French version | German version | Spanish version | Portuguese version