Age estimation

While essential for a wide range of legal questions, estimating chronological age in living individuals has turned out to be a challenging task. The Age Estimation (AE) project entails three parallel data collections pertaining to: a) DNA methylation and protein profiling, b) visual age assessments and c) interviewing.

The DNA research (a) aims to evaluate the use of DNA-methylation levels as a tool for estimation of age in young individuals. DNA methylation is important for gene regulation and the DNA methylation levels change with chronological age. A subset of 5-10 markers which are highly correlated with age will be used to predict age using Pyrosequencing technology and interpreted using artificial neural networks (Level 1).

The research into visual age assessments (b) is motivated by, for example, that Courts such as the Special Court of Sierra Leone (SCSL) and the International Criminal Court (ICC) have relied heavily, and sometimes solely, on the physical appearance of alleged child soldiers. In this research we will compare accuracy levels in the age estimations of legal actors in e.g. Sweden, UK and the Netherlands (visual age experiments) to automated approaches to age prediction based on facial images (automatic age prediction). This enables evaluation of whether and how automated approaches (machine learning, image processing) can assist or inform decision making in criminal cases (Level 1).

The interviewing research (c) aims to test whether language and behavioral cues obtained in an interview significantly aids the age estimation process, on top of visual information.Coordinated evaluations of these age estimation methods (a-c) enables comparisons of accuracy levels, which will be directly informative of best practices, and facilitate prioritisation between contradicting age evidence (Levels 2 & 3).