Final Report Summary - NEUROCODEC (Neurobiological basis of collective decision making in the human brain)
NEUROCODEC project was conducted to better understand the cognitive and neurobiological basis of collective decision making in humans. The projects output can be categorized in 4 categories that are conceptually related.
(1) Communication of uncertainty: We found that even though collective decisions are best made when disagreeing opinions are weighted by their respective reliabilities, human agents rarely follow this optimal path (Mahmoodi et al 2015; Hertz et al 2016). Instead, we saw that human mind resorts to a simplifying heuristic called Equality Bias: when making joint decisions, we ignore inter-individual differences in reliability and behave as if everyone’s is equally reliable. This is useful when decision partners have similar levels of expertise. It is unhelpful when partners have very different expertise: overconfidence of the lesser able partner damages joint decisions. This bias is universal among cultures and robust to performance feedback and financial incentives. A large portion of between-individual variability in confidence reports is explained by how people combine subjective probability and environmental noise to report their confidence. We found that this combination is like a computational fingerprint: unique to each participant and stable across time and cognitive domains. These findings could pave the way for developing bespoke interventions tailored to individual agents that help improve their collective performance.
(2) Social Influence: Our game theoretic analyses indicated that overconfidence does promote individual’s interests at the expense of others. We found that in client-advisor relations, advisors’ overconfidence depends on her social status (is the client ignoring the advisor?) and self-image (does advisor believe she is better than her rival?). The contribution of objective facts and strategic spin to confidence can be disentangled in behaviour and brain activity. These findings are the first of their kind in the literature of social influence and help us understand the complex mental states of those who wish to manipulate us.
(3) Pharmacology of collective decisions: We found that testosterone had a clear impact on collective decision making under uncertainty. Collective decisions were less sensitive and individuals were more egocentric under exogenous testosterone. Under Oxytocin, collective decisions making was not impaired. These results showed that hormonal states have a limited impact on the success and failure of collective decisions and one has to exercise caution when evaluating their impact on social behaviour.
(4) Large scale social interaction: We found new ways to organize group interactions to benefit from “wisdom of crowds” without suffering from the group-level consequences of overconfidence such as herding and polarization. Locally structured deliberation decreased the diversity of opinions within each small group, similar to “herding” and increased between-group distance, similar to what is seen in polarization. The aggregated impact of these two effects led to the almost-miraculous observation that putting together the consensus opinions of as few as 4 groups of 5 people could beat the wisdom of a crowd of 5000 individuals.
(1) Communication of uncertainty: We found that even though collective decisions are best made when disagreeing opinions are weighted by their respective reliabilities, human agents rarely follow this optimal path (Mahmoodi et al 2015; Hertz et al 2016). Instead, we saw that human mind resorts to a simplifying heuristic called Equality Bias: when making joint decisions, we ignore inter-individual differences in reliability and behave as if everyone’s is equally reliable. This is useful when decision partners have similar levels of expertise. It is unhelpful when partners have very different expertise: overconfidence of the lesser able partner damages joint decisions. This bias is universal among cultures and robust to performance feedback and financial incentives. A large portion of between-individual variability in confidence reports is explained by how people combine subjective probability and environmental noise to report their confidence. We found that this combination is like a computational fingerprint: unique to each participant and stable across time and cognitive domains. These findings could pave the way for developing bespoke interventions tailored to individual agents that help improve their collective performance.
(2) Social Influence: Our game theoretic analyses indicated that overconfidence does promote individual’s interests at the expense of others. We found that in client-advisor relations, advisors’ overconfidence depends on her social status (is the client ignoring the advisor?) and self-image (does advisor believe she is better than her rival?). The contribution of objective facts and strategic spin to confidence can be disentangled in behaviour and brain activity. These findings are the first of their kind in the literature of social influence and help us understand the complex mental states of those who wish to manipulate us.
(3) Pharmacology of collective decisions: We found that testosterone had a clear impact on collective decision making under uncertainty. Collective decisions were less sensitive and individuals were more egocentric under exogenous testosterone. Under Oxytocin, collective decisions making was not impaired. These results showed that hormonal states have a limited impact on the success and failure of collective decisions and one has to exercise caution when evaluating their impact on social behaviour.
(4) Large scale social interaction: We found new ways to organize group interactions to benefit from “wisdom of crowds” without suffering from the group-level consequences of overconfidence such as herding and polarization. Locally structured deliberation decreased the diversity of opinions within each small group, similar to “herding” and increased between-group distance, similar to what is seen in polarization. The aggregated impact of these two effects led to the almost-miraculous observation that putting together the consensus opinions of as few as 4 groups of 5 people could beat the wisdom of a crowd of 5000 individuals.