TL;DR
This study uses brain stimulation (cTBS) and computational modeling to show that the left temporoparietal junction (TPJ) is causally involved in how we incorporate others' dissenting opinions into our own decisions during goal-directed learning. Disrupting the left TPJ specifically impaired this social influence computation, while leaving other learning processes intact.
What Problem Does This Paper Address?
Understanding how our brains process social information is crucial for explaining human behavior, especially in learning scenarios where we might be influenced by others. While the temporoparietal junction (TPJ) is known to be involved in social cognition, previous research has often focused on the right TPJ or used correlational methods like fMRI. This paper aims to establish a causal link between specific parts of the TPJ (left vs. right) and the computation of social influence during goal-directed learning, particularly when others' choices contradict our own. They want to pinpoint which hemisphere of the TPJ plays a critical role and how this influence is integrated into our decision-making process.
Methods
The researchers employed a combination of techniques:
- Continuous Theta-Burst Stimulation (cTBS): This non-invasive brain stimulation technique was used to temporarily disrupt activity in specific brain regions. Participants received cTBS over the left TPJ, right TPJ, or a control region (vertex) in separate sessions.
- Social Learning Task: Participants engaged in a probabilistic reversal learning task. They made an initial choice, then observed the choices of four simulated "intelligent computer algorithms." Based on this social information, they could adjust their choice before receiving the outcome.
- Hierarchical Bayesian Computational Modeling: This sophisticated modeling approach was used to quantify how participants integrated different sources of information (their own learning and social learning) into their decisions. They specifically modeled parameters related to social influence and observational learning.
- Model-Agnostic Behavioral Analysis: Standard statistical models (linear mixed models) were used to analyze behavioral outcomes like choice switching probability, response time, and accuracy, looking for effects of cTBS.
Datasets
Not discussed in the paper. The study used data collected from 31 human participants.
Key Insights and Ideas
- Left TPJ is Key for Social Influence: The most striking finding is that disrupting the left TPJ, but not the right, significantly reduced participants' willingness to adjust their choices when others disagreed. This suggests a causal role for the left TPJ in processing and acting upon social dissent.
- Social Influence vs. Observational Learning: The computational modeling revealed a crucial distinction. Disrupting the left TPJ specifically impaired the computation of social influence (how much dissenting opinions swayed the decision), but it did not affect the integration of observational learning (learning from others' outcomes). This suggests the left TPJ is specialized for processing social disagreement, rather than general learning from others.
- Behavioral Manifestations: The behavioral impact was clear: reduced choice adjustment and slower response times when participants' choices were contradicted by the group, specifically after left TPJ disruption.
Key Results
- Behavioral:
- Disrupting the left TPJ significantly reduced choice switch probability, especially when participants' choices were contradicted by the majority (4:0 consensus).
- Disrupting the left TPJ also led to a trend of increased response time for choice adjustment in the 4:0 consensus condition.
- Choice accuracy was not significantly affected by TPJ stimulation.
- Computational:
- Disruption of the left TPJ significantly decreased the parameter representing the computation of social influence (β(Nagainst)) compared to the right TPJ and vertex.
- There was no significant difference in the parameter representing the integration of vicarious value computation (β(Vother)) across stimulation sites.
- The computational measure of social influence (β(Nagainst)) was a significant predictor of behavioral choice switching, and this predictive relationship was weakest when the left TPJ was disrupted.
Limitations and Future Work
- Potential Role of Attention: The authors acknowledge a potential caveat that TPJ stimulation might have affected attentional processing, which could confound the results. However, they argue against this by noting that their stimulation sites were identified based on "mentalizing" rather than "attention" tasks and that an attentional account wouldn't explain the differences between left and right TPJ.
- Artificial Agents: The study used simulated "intelligent computer algorithms" instead of real human participants. While these algorithms were designed to mimic human behavior, future work could explore these effects with actual human interactions.
- Combined TMS-fMRI: The authors suggest that future studies could combine TMS with fMRI to investigate the connectivity patterns between the TPJ and other brain regions (like vmPFC and ACC) that are involved in learning and decision-making. This could provide a more nuanced understanding of how the TPJ integrates social information.
- Measuring Social Information and Attention Simultaneously: For future brain stimulation studies of the TPJ, it would be beneficial to measure both social information processing and attentional processing concurrently to better understand potential interactions.
Summary updated