The Open University
Modelling Eyewitness Accuracy using Predictive Analytics: Towards Reduced Wrongful Convictions
Abstract
dc:description.abstractWrongful convictions significantly impact lives. These can be caused by juries misinterpreting Eyewitness evidence. This thesis argues for the need to provide juries with probabilistic information on likely eyewitness identification accuracy, factoring in the complexity of the witnessing environment, and proposes that predictive analytics may provide a viable mechanism for quantifying this probabilistic information.<br></br><br></br> A series of experiments evaluated the effects of separation distance, retention interval, and exposure time on identification accuracy.<br></br><br></br> Higher accuracy was associated with lower separation distance, shorter retention interval, and longer exposure times, although the relationships were non-linear. Confidence was predictive of accuracy in both target-present and target-absent lineups. A predictive model developed using the experimental data demonstrated high calibration accuracy by closely matching participant performance. Integrated data analysis (IDA) was used to combine data from the individual experiments into the model, improving calibration, applicability across a wider range of witnessing conditions, and increasing precision. Fairness checks verified that the model's predictions were unbiased across participant demographics.<br></br><br></br> The thesis concludes that the predictive analytics approach could potentially be used for building a reliable courtroom tool that provides a probabilistic measure of eyewitness accuracy despite complex witnessing conditions.<br></br>
Degree
thesis:*- Name dc:type.qualificationname
- phd
- Level dc:type.qualificationlevel
- doctoral
- Grantor dc:publisher.institution
- The Open University
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Farmer, Stephen J.
Rights
- Language dc:language
- en