Implementing Artificial Intelligence into Team Decision Making

The problem this research aims at is the suboptimality of team decision making. This thesis performs an attempt at implementing an artificial agent into a conversation to facilitate team decision making. Two methods are experimented with to attempt achieving this goal. The agent records and analyzes under and over speaking within participants of the team. The agent also tracks conversation topics and generates text to add relevant information into the conversation based on what the team is currently discussing. The thesis refers to the first method as ‘moderating speaking equity’ and to the second method as ‘inserting conversation relevant information’. Moderating speaking equity is performed through a python script and inserting conversation relevant information is performed through two separate Recurrent Neural Networks (RNNs) with Long Short-Term Memory (LSTM) layers and Generative Pretrained Transformer J (GPTJ).