A recent report entitled "Child Maltreatment 2011" issued by the U.S. Department of Health & Human Services Administration for Children and Families estimated that more than 3.7 million children were the subjects of at least one report of maltreatment, and that over 681,000 children were found to be unique victims of child maltreatment in the United States. This multidisciplinary research will compare the effectiveness of robot vs. human interviewers for gathering sensitive information from children, using situations in which this would commonly occur: cases of child eyewitness memory and child reports of bullying. The PI argues that the use of a robot as an intermediary during these so-called forensic interviews could reduce or eliminate unintentional cues observed in human interviewers that result in inaccurate reports by children. To validate her hypothesis the PI and her team will develop a systems architecture, an interactive user interface and an interactive robotic toolkit for interviewers, and perform a series of six studies involving children ages 8-11. The interdisciplinary research team is comprised of experts in human-computer interaction, human-robot interaction, robotics, psychology, sociology, and social work, and the project will make contributions to each of these domains. The team further includes a member of the legal profession as a consultant, who will iteratively evaluate the potential for extending the research findings to real-world legal proceedings and investigations. Preliminary research conducted by the PI has attracted attention from the law enforcement and legal communities, so if successful this project has the potential to transform information gathering for investigative purposes. The PI and her colleagues have been actively involved in community outreach in local middle schools and Boys and Girls Clubs with respect to the use of robots for eliciting information related to bullying, and this outreach will be extended to elementary school children involved in the current research.
The research goals for this project will be accomplished through the development of an integrated robotic toolkit based on a novel Interactive Social Engagement Architecture (ISEA) and a unique interactive user interface. ISEA provides a framework for the autonomous generation of robot behaviors for self-preservation and to convey social intelligence. The toolkit will be designed to integrate behavior-based robotics, human behavior models, cognitive architectures, and user input to increase social engagement between a human and system (robot, avatar, etc.). The interactive user interface will provide interviewers with the ability to use the robot as an intermediary for gathering sensitive information. ISEA has three primary parallel paths for processing robot behaviors: (1) verbal behaviors based on expert user input from the interactive user interface; (2) autonomous self-preservation behaviors if the robot is threatened that consist of both verbal and non-verbal responses; and (3) non-verbal autonomous behaviors generated from sensor data coming from the environment, the current internal state of the robot, user input, and prior knowledge from the knowledge base/long-term memory. As part of the research, six human studies will be conducted that use typical situations in which gathering sensitive information from children might occur. Three of these experiments will examine whether child eyewitness memory is more accurate when a robot rather than a human presents misleading information during an interview, while the other three will examine whether children who have been victimized by bullying will be more likely to disclose that victimization to a robot as opposed to a human interviewer. Some of these experiments will examine the role of gender both for humanoid robots and adult interviewers, using established forensic interview protocols, while others will examine whether interviewers high in social intelligence elicit more accurate child eyewitness memory and reports of bullying than those low in social intelligence (where social intelligence is defined by the use of gestural and facial non-verbal behaviors).
