Presentation
Automated identification of insight seeking behaviours
SessionvPOS1: Poster Gallery
Event Type
Poster
Virtual
TimeTuesday, October 26th10:00am - 10:50am EDT
LocationVirtual 1
DescriptionIn an investigation, analysts query information, gain insights, and make claims. Whilst this is a fluid and informal process, we propose that by capturing interactions with a question-answer system we can model and predict meaningful strategies employed by analysts to gain insights. In this paper, we analyse data from a study with 11 intelligence analysts applying a mix of qualitative and quantitative methods. Each participant interacted with a prototype system to conduct an investigation, based upon a real-world scenario. We recorded, transcribed and analysed the audio data to identify markers for insight seeking behaviours. We then developed a method to automatically label and to classify when these behaviours occurred. We applied quantitative methods to identify what kinds of rules and strategies were used by analysts when seeking insights and making claims. We propose that this understanding will aid us to design intelligent systems for knowledge discovery.