The RSS’s Statistics and Law section has highlighted the need for better statistical expertise in the justice system in two important inquiries relating to the role of statistics in UK law.
In a submission to the Law Society’s inquiry into Algorithms in the Justice System (PDF), the RSS draws attention to ‘dangerously misleading statistical arguments in the current literature, and the consequent need for expert statistical input.’
Our submission identifies a number of misleading statistical analyses of algorithms currently in use. It states: 'A proper understanding of the statistical properties of algorithms, which is vital to their use and usefulness, requires the application of expert statistical knowledge and understanding,’ adding that the Royal Statistical Society would be pleased to act as a source of advice in this area.
In our submission to the House of Lords Science and Technology inquiry on Forensic Science, we also express concern at the ‘insufficient appreciation of the extent to which the main current forms of forensic evidence […] are founded on statistical methods.’
Our submission identifies a number of weaknesses in forensic science in the justice system, including inadequate vetting of experts, the ‘commoditisation’ of forensic science and poor quality of probabilistic reasoning and statistical evidence in this area. We suggest that a central register for experts might be useful.
We also point out the importance of statistics in communicating uncertainty and that in general, solicitors, judges would benefit from some statistical training. The RSS Statistics and the Law section is already planning to offer training courses for lawyers and statistical experts on the use of statistical information in personal injury litigation. ‘This is an objective for the medium term as it will take considerable resources to design and deliver such courses,’ our submission notes, adding that ‘funding will be essential’.
Jane Hutton, who chairs the RSS Statistics and Law Section, said: 'The precise questions posed to the scientist or statistician are very important and the assessment of data quality is essential.'