Evaluating how quantitative likelihood ratios can support the interpretation of facial comparison evidence in practice.

Forensic facial comparison is currently based on qualitative, perceptual observations made by trained examiners. While structured, these judgements can be difficult to validate and can vary between practitioners.

This project explored the use of automated facial recognition systems and statistical modelling to generate quantitative measures of similarity, expressed as likelihood ratios.

The aim was to assess whether these approaches could provide a more consistent and transparent basis for evaluating the strength of evidence, and how they compare to existing methods in real-world conditions.

This work sits at the intersection of biometric systems and forensic practice, focusing on how quantitative outputs can support, but also influence, the interpretation of evidence.