Task Engagement Inference Within Distributed Multiparty Human-Machine Teaming via Topic Modeling

Abstract

Research in Intelligent Awareness Systems (IAS) focuses on designing systems that are aware of their current environment by monitoring human interactions and making inferences on when to engage with human counterparts. A potential gap is task engagement inference for distributed human-machine teaming. The objective of this paper is a proposed intelligent awareness system via task topic modeling for task engagement inference within these domains. If the system has information on “what” teammates are discussing or the task topic, it is better informed prior to engaging. The proposed task topic model is applied to two simulated multiparty, distributed teaming interactions and evaluated on its ability to infer the current task topic. For both tasks, the model performs well over the random baseline, however the performance is degraded for interactions with more robust dialogue. This work has the potential of informing the development of intelligent awareness systems within distributed multiparty teaming and collaborative endeavors.