You are here


A cognitive approach for modelling and reasoning over shared experiences
Sep 2014 - Jan 2017
Main researcher
IIIA Collaborators
Project description

Our dependance on intelligent agents in our everyday lives is steadily increasing, as we become more and more reliant on the appliances and gadgets surrounding us. This demands an efficient human-agent collaboration. CollectiveMind presents an original and bold approach for human-agent collaboration. It proposes the creation of a collective mind, which maintains past shared experiences and has a reasoner that builds a model of the group based on those experiences. This accelerates negotiation and argumentation to reach mutual agreements based on the accumulation of shared experiences. CollectiveMind follows a cognitive approach that is more intuitive for humans to comprehend, which we believe is a must in mixed human-agent societies. The project will first investigate and introduce the notion of experience to the epistemic state of individual agents. This situates agents in their environments and closes the BDI circle: experiences shape beliefs, which influence desires and intentions, leading to new experiences. Using the notion of shared experience as a primitive construct, we will then develop a novel formal model of shared intention which we believe more adequately describes and motivates social behaviour than traditional BDI logics that focus on modelling individual agents. Whilst many philosophers have strongly argued that collective intentionality cannot always be equated to the collection of the individual agents, there has been no AI model that has gone beyond the simple aggregation of individuals agents. We hope that our proposal will shed light on the collective mind concept from an AI perspective. This project will investigate the concepts of experience and collective intentionality. Some of the expected results include a formal model of shared experience to underpin collective intentionality, an experience-based BDI logic for reasoning over collective intentionality, and a corresponding experience-based BDI reasoner.