Bayesian networks and dissonant items of evidence: A case study

Area of Science:

• Forensic Science
• Probability Theory
• Computational Statistics

Background:

• Assessing diverse forensic evidence supporting different conclusions is complex.
• Real-world criminal cases often involve intricate evidential scenarios.

Purpose of the Study:

• To develop a model for assessing the joint probative value of multiple forensic analyses.
• To demonstrate the utility of Bayesian networks in handling complex forensic evidence.
• To investigate the role of conflict measurement and contamination events in forensic assessment.

Main Methods:

• Development of a probabilistic graphical model, specifically a Bayesian network.
• Integration of three distinct analyses of biological material.
• Inclusion of a contamination event to model potential inconsistencies.

Main Results:

• The Bayesian network effectively handles complex, multi-faceted forensic evidence.
• A conflict measure is crucial for identifying internal inconsistencies within the model.
• Modeling a contamination event helped explain and resolve identified conflicts.

Conclusions:

• Probabilistic models, like Bayesian networks, are valuable tools for forensic science.
• These models can assess the joint probative value of multiple evidence types.
• The approach provides insights into managing and resolving conflicting evidence in criminal investigations.