Recommendation of Collaboration Patterns for Human-Machine Collective Intelligence

The problem of supporting collaborative processes is especially acute in scenarios, where teams are formed dynamically and team members have little experience in working with each other. The paper proposes an approach to increase the efficiency of the collective activities in the context of a human-machine collective intelligence environment for decision support. The approach consists of two stages. In the first stage, non-productive situations of the collective work are identified by a set of rules, applied to the ontological representation of the status of the team activities, and a set of candidate recommendations of team behaviors (collaboration patterns) is formed. In the second stage, the candidate list of recommendations is ranked and filtered according to teams feedback by a contextual bandits algorithm. Gradually, it allows to validate the set of rules and increase the relevance of the recommendations.