Description du projet
Éliminer la confiance excessive dans la prise de décision collective
Le biais de confiance excessive est une tendance bien ancrée selon laquelle la confiance subjective d’une personne dans ses décisions est largement supérieure à l’exactitude objective desdites décisions, en particulier lorsque le niveau de confiance est relativement élevé. Cette tendance très répandue et coûteuse est à l’origine de nombreuses erreurs humaines. Les tentatives précédentes, qui s’étaient soldées par un échec, visant à réduire la confiance excessive portaient sur des décisions individuelles. Le projet rid-O, financé par l’UE, vise à réduire cette tendance au niveau social de la prise de décision en déterminant les processus mentaux, neuronaux et sociaux sous-jacents qui interviennent dans le biais de confiance excessive. Le projet, qui se concentre sur trois formes principales de décisions sociales, cherche à développer une nouvelle intervention causale qui crée un consensus.
Objectif
“What would I remove if I had a magic wand? Overconfidence” Daniel Kahneman’s famous fairy-tale wish (The Guardian 18 Jul 15) conveyed a deeply seated pessimism in cognitive scientists that overconfidence is hardwired into human cognition. This bias is pervasive, costly and the root cause of many human failures. Previous failed attempts at reducing overconfidence targeted individual decisions. I propose to reduce this bias at social level of decision making by determining the underlying mental, neural and social processes involved in overconfidence and testing these models by causal interventions. I focus on 3 common forms of social decisions.
1.In honest communication of uncertainty (e.g. 2 doctors disagreeing over a diagnosis), overconfidence impairs joint decisions. Combining computational analysis of behavior and brain response, I develop a real-time feedback loop that allows each disagreeing agent to weight her opinion by an individually-tailored signature of her own uncertainty, freeing joint decisions from overconfidence.
2.Focusing on advising and consulting, I use a novel laboratory model (i.e. Advising Game) to develop a theoretical and empirical understanding of overconfidence in the presence of conflict of interest to understand the mental and neural processes underlying strategic manipulation of others. Plus, by connecting the social use of overconfidence to self-esteem and self-worth, I translate this research to a Mental Health application looking at social dysfunctions in Depression.
3.Overconfidence impairs group processes (e.g. a panel selecting among grants) by promoting herding (blindly following others) and polarization to extreme viewpoints. Inspired by a recent discovery from my lab, I develop a novel causal intervention that puts together seeking consensus within each group and aggregating consensus opinions across groups to remove herding and polarization.
Collective decisions can be made better and rid-O could help this wish come true.
Champ scientifique
Programme(s)
Régime de financement
ERC-COG - Consolidator GrantInstitution d’accueil
80539 MUNCHEN
Allemagne