Human Computer Collaboration at the Edge: Enhancing Collective Situation Understanding with Controlled Natural Language

Effective coalition operations require support for dynamic information gathering, processing, and sharing at the network edge for Collective Situation Understanding (CSU). To enhance CSU and leverage the combined strengths of humans and machines, we propose a conversational interface using Controlled Natural Language (CNL), which is both human readable and machine processable, for shared information representation. We hypothesize that this approach facilitates rapid CSU when assembled dynamically with machine assistance, via social sensing, from local observations, with information rapidly disseminated among people at the network edge. We report a behavioural experiment wherein small groups of users attempted to build CSU via social sensing, interacting with the machine via Natural Language (NL) and CNL. To simulate a tactical environment, participants answered 36 questions (operationalized as CSU) by visiting various locations and describing their discoveries to a mobile conversational agent. To test our hypothesis, we compared the performance of groups of users between the: 1) Online Condition: CSU, the status of all questions, dynamically updated by the machine as users collect information. 2) Offline Condition: No dynamic machine-supported CSU, simulating unreliable connectivity at the edge. Each participant was restricted to their own information until the end of the experiment. Results indicated the Online Condition had greater agreement in CSU, but individual participants answered significantly fewer questions than the Offline Condition. In other words, the Offline Condition group provided more answers, but there was more consistency among the answers provided by the Online Condition group.