Internet crime reporting: design and evaluation of a computer investigative interview system
Gondy Leroy, Alicia Iriberri · 2009
This dissertation presents an investigation that designed and evaluated an Internet crime reporting and investigative interviewing system, i-recall. The investigation followed the Information Systems Design Research Framework by Hevner, March, Park and Ram (2004). Knowledge gained from cognitive and social psychology, criminology, computer science, information systems and technology, and from two exploratory studies on witness reporting preferences and police department information requirements was used in the design and development of i-recall. i-recall emulates a police officer conducting a cognitive interview (CI). It incorporates CI techniques to enhance witness memory retrieval and leverages natural language processing technology (NLP) to support understanding of witness narratives and generation of logical sequences of questions based on witness preceding responses. Two evaluations of i-recall were conducted. The first evaluation verified the efficacy of i-recall's information extraction (IE) algorithm. Written narratives from individuals describing suspects depicted in mug shots were collected. The narratives were processed with the IE algorithm. The results indicate that the algorithm extracts 90 percent of information from witness-written narratives with 96 percent precision. The second evaluation validated i-recall's interviewing performance. In a controlled user study, the system was compared to a face-to-face (human) cognitive interview (ci-interview) and a non-interactive textbox computer system (i-textbox). College sophomores participated as witnesses to a videotaped crime and reported what they saw using one of the three methods. Significant differences in performance among the methods were observed. The results show that i-recall elicited 51 correct descriptive items from witnesses' memory with 94 percent accuracy; the face-to-face human cognitive interview elicited 69 correct items with 97 percent accuracy; and the textbox system elicited 21 correct items with 96 percent accuracy. This indicate that i-recall elicits 26 percent less amount of information than the ci-interview and 143 percent more than the i-textbox and that 94 percent of the information it elicits is correct. That is, i-recall is a capable Internet reporting alternative. This dissertation contributes a computer-based application of the CI procedure to use when CIs are not feasible. It also represents an application of NLP based-interfaces in e-government services. Future work will evaluate i-recall in natural settings.