James MacLeod MSc Dissertation 2014/15

Robot Crime Investigators: Non-Verbal Cues and the Misinformation Effect

Supervised by A.Sharkey

Abstract

This report presents research from a study to determine whether eyewitness memory was impacted by a robot interviewer presenting misleading post-event information with non-verbal cues (body language) disabled versus non-verbal cues enabled.

The study was conducted with 20 participants who viewed video footage of a crime, then were interviewed by a NAO robot with cues enabled or disabled, presenting control or misleading information about the event. Following this, the participant completed a written memory test to analyse their accuracy of recall of the video’s events, and a questionnaire on how they perceived the robot.

The results found no significant difference in the overall recall accuracy between the cues disabled (p~0.61) and cues enabled groups (p~0.19), nor in the recall accuracy for critical questions with cues disabled (p~0.74) and enabled (p~0.15). Though not significant, the results suggest there might be a link between the robot interviewer having cues enabled and its ability to misinform. If any link is due to cues then their impact might be subconscious, as participants did not consciously notice any difference in their perception of the robot. Further research using a larger, stratified sample is required to confirm this result.