Behavioral Sci The Law - 2007 - Yokota - Application of the behavioral investigative support system for profiling

Application of the Behavioral Investigative Support System for Profiling Perpetrators of Serial Sexual Assaults

• Authors: Kaeko Yokota, Ph.D., Goro Fujita, Ph.D., Kazumi Watanabe, B.A., Kaori Yoshimoto, B.A., Taeko Wachi, M.Phil.

• Purpose: Developed a data-based profiling system to support offender profiling by storing incident records of prior offenders.

  • Input unsolved incident details, assigning probability scores indicating behavioral similarity with known offenders.

  • Ranked offenders based on scores to prioritize possible suspects.

  • Inferred characteristics of unknown offenders from high-ranked offenders.

  • Achieved a high retrieval accuracy (45 out of 81 trials retrieved target offenders ranked as 1).

Background on Offender Profiling

• History: Offender profiling techniques established by FBI agents, widely adopted internationally since the 1990s.

• Japan's Development: The National Research Institute of Police Science began profiling research in 1996.

  • Established police units focused on behavioral crime analysis to support investigations by 2000.

Behavioral Consistency in Offender Profiling

• Theory: For effective profiling, consistency in offender behavior across contexts is necessary, despite natural human variability (Alison et al., 2002).

• Related Studies:

  • Modus Operandi Approach: Offenders develop personal methods based on previous successes (O’Hara, 1956).

  • Psychological Signature: A consistent psychological imprint left at crime scenes by offenders (Douglas et al., 1992).

Methodology of Profiling System

Data Collection

• Incidents analyzed: 1,252 records from 868 offenses (rape and indecency through compulsion).

  • Variables included victim type, location, time, method of approach, weapon, and behavior.

• Exclusions: Different types of sexual offenses and duplicate incidents from the same offenders.

Calculation of Behavioral Similarity

• Probability Score Calculation: A probability score indicates an offender's behavioral similarity to an unsolved incident.

  • Multinomial distribution was used to determine choice probabilities for actions performed during crimes.

Simulation Experiments

Linking Crimes to Offenders

• A collection of simulations conducted to test the effectiveness of linking incidents with specific offenders.

  • Hit rate of 29.6%, indicating successful identification in approximately 30% of trials.

  • Median rank score of 4 suggests that half of the offenders were retrieved within a small range of rank scores.

Predicting Offender Profiles

• Examined sex crime history of offenders among the top 5% most similar to target incidents.

• Results suggested that offenders with similar criminal styles often share a repeat offender profile.

Results and Findings

Linking Crimes

• Inclusion of crime location as a variable improved hit rate from 29.6% to 55.6%.

  • Indicates the importance of geographical context in profiling.

Predicting Offender Characteristics

• The proportion of offenders with past sex crime history was notably higher among repeat offenders compared to first-time offenders, confirming profiling effectiveness.

  • Reliable predictors of offending behavior could infer the likelihood of further crimes based on past actions.

Discussion

• System Implications: The Behavioral Investigative Support System holds promise for accurately linking crimes to offenders, suggesting a systematic approach can yield positive investigative results.

• Limitations: Need for a robust database for comprehensive accuracy in profiling efforts; limitations due to variability in original data collection processes.

• Future Directions: Continued developments in profiling systems to enhance predictive accuracy and operational readiness for police forces in Japan.