Learning about Bayesian networks for forensic interpretation: an example based on the 'the problem of multiple...

Area of Science:

• Forensic Science
• Probability Theory
• Computer Science

Background:

• Probabilistic evaluation and Bayesian networks are increasingly used in forensic science.
• A knowledge gap exists in foundational resources for scientists new to these topics.
• Master's students in forensic science require accessible learning materials on these complex subjects.

Purpose of the Study:

• To report on the learning experiences of master's students using Bayesian networks for likelihood ratio-based probabilistic inference.
• To address the gap in foundational literature for applying Bayesian networks in forensic science.
• To demonstrate the utility of Bayesian networks in handling complex forensic casework with multiple propositions.

Main Methods:

• Utilized a casework scenario involving a questioned signature from published literature.
• Employed generic Bayesian network fragments from existing literature for teaching.
• Students analyzed the scenario to calculate likelihood ratios and posterior probabilities.
• Students derived an alternative Bayesian network structure with equivalent computational output.

Main Results:

• Students successfully tracked the probabilistic underpinnings of the scenario using Bayesian networks.
• Participants correctly calculated likelihood ratios and posterior probabilities.
• An alternative Bayesian network structure was developed by students, yielding equivalent results.
• The practical exercise enhanced students' understanding of probabilistic procedures.

Conclusions:

• Bayesian networks can effectively support and clarify foundational principles of probabilistic evaluation in forensic science.
• Teaching probabilistic inference through Bayesian networks is feasible and beneficial for forensic science students.
• The study highlights the potential of Bayesian networks to simplify complex forensic casework analysis.