IRG lab member Joseph Kim's recently developed an intelligent machine that listens to a team’s conversation and infers whether there is consensus. The machine then provides feedback, alerting the team to review "weak" discussion points that could later result in confusion. Joseph designed a computational model that processed the team’s utterances to extract a set of dialogue features that are indicative of a team’s shared consensus, according to well-established cognitive models. The machine used this model to participate in live team meetings, improving the team’s consistency in understanding by up to 17%. The article is published in the IEEE Transactions on Human-Machine Systems (THMS). You can read the full paper here. You can also check out the system demonstration video here.
June 30, 2016