DScent Final Report
- Publisher: Leeds Metropolitan University
DScent was a joint project between five UK universities combining research theories in the disciplines of computational inference, forensic psychology and expert decision-making in the area of counter-terrorism. This document discusses the work carried out by Leeds Metropolitan University which covers the research, design and development work of an investigator support system in the area of deception using artificial intelligence. For the purposes of data generation along with system and hypothesis testing the project team devised two closed world games, the Cutting Corners Board Game and the Location Based Game. DScentTrail presents the investigator with a ‘scent trail’ of a suspect’s behaviour over time, allowing the investigator to present multiple challenges to a suspect from which they may prove the suspect guilty outright or receive cognitive or emotional clues of deception (Ekman 2002; Ekman & Frank 1993; Ekman & Yuille 1989; Hocking & Leathers 1980; Knapp & Comadena 1979). A scent trail is a collection of ordered, relevant behavioural information over time for a suspect. There are links into a neural network, which attempts to identify deceptive behavioural patterns of individuals. Preliminary work was carried out on a behavioural based AI module which would work separately alongside the neural network, with both identifying deception before integrating their results to update DScentTrail. Unfortunately the data that was necessary to design such a system was not provided and therefore, this section of research only reached its preliminary stages. To date research has shown that there are no specific patterns of deceptive behaviour that are consistent in all people, across all situations (Zuckerman 1981). DScentTrail is a decision support system, incorporating artificial intelligence (AI), which is intended to be used by investigators and attempts to find ways around the problem stated by Zuckerman above.