In the first article, I examine the "reverberation effect" of face-to-face communication in public goods games, exploring whether its positive impacts on cooperation persist after being removed. Participants engaged in the classical voluntary contribution mechanisms (VCM) with and without pre-play communication.
Persistence of Communication Effects: Contributions remained high after communication was removed, suggesting a lasting but diminishing positive impact of communication. Over time, cooperation declined, aligning with pre-communication levels, indicating that the effect is not permanent.
End-Game Behavior: Despite communication, the "end-game effect" — a significant drop in contributions in final rounds — persisted, highlighting inherent challenges in sustaining cooperation long-term.
Mechanism Insights: Communication fostered an understanding of mutual benefits, temporarily enhancing cooperation. However, group reshuffling and absent institutional reinforcements weakened these effects.
In total, while communication improves efficiency, repeated interventions may be necessary to maintain cooperation in the absence of institutional measures. Further, a brief machine learning analysis of facial expressions and communication content revealed predictors of cooperative behavior, emphasizing the value of discussing end-game strategies.
The second article dives deeper into neuro-science. Together with two neuro-information scientists, we investigate how exactly non-verbal communication predicts group cooperation in public goods games. Using three-minute face-to-face communication (FFC) videos, we analyzed facial expressions and verbal content to predict whether groups would contribute fully by the game’s end. Using Random Forest classification (RF), we processed facial activity descriptors to identify behavioral patterns. Results showed that late-stage communication was more indicative of cooperation than earlier phases.
FFC significantly increases cooperation, with the final minutes providing the most predictive cues.
RF achieved better-than-chance predictions but requires larger datasets for accuracy.
Verbal discussions about "end-game" strategies enhanced contributions.
This interdisciplinary approach links behavioral economics and computer vision, highlighting the role of communication in fostering cooperation and proposing methods to predict and improve group dynamics in social dilemmas.
