Evaluation of factors involved in calculating Bayesian likelihood ratio for forensic shoeprint comparison analysis

Abstract

The discipline of forensic science first came under scrutiny in 2009, where problems associated with the reliability, validity and accuracy of various feature comparison methods were identified and examined. Recommendations were made to resolve these problems. In 2016, the President's Council of Advisors on Science and Technology (PCAST) released a report challenging all feature comparison methods in forensic science, including shoeprint comparison analysis, arguing that that there is no "foundational scientific validity" for any of these methods and they should not be used continually as credible evidence in court. This has created a debate within the forensic communities as well as the associated groups to determine whether foundational scientific validity exists for feature comparison methods. The main purpose of this research is to evaluate all the available literature on shoeprint comparison analysis to see if foundational scientific validity exists. And if possible, attempt to derive a Bayesian likelihood ratio for shoeprint comparison Bayesian interpretation. Then, to identify the gaps existing in the current available knowledge of shoeprint comparison analysis in order to address the problems associated with the report, and to devise possible research directions for future studies to strength the discipline. An experiment was conducted to discover possible problems associated with carrying out a study on shoeprint comparison analysis and determine the factors associated with shoeprint comparison analysis. A pair of common canvas shoes were worn by the author for a period of 3 months, details of the time interval and activities of the author were recorded. Test prints and photographs of the left shoe outsole were taken for analysis. The results showed that the experiment could provide a baseline for future studies as the shoe was fairly common, of a common size, and traversed in a regular pattern with the author having a moderately active lifestyle. They study also provided Abstract iii insight into some of the considerations which were not made prior to the experiment taking place, such as the inclusion of partial prints and the length of time for the experiment to be completed. These are all important factors to contemplate when designing an experiment. Based on the literature review conducted, PCAST has not reviewed many of the studies they did not consider valid, although they could have provided valuable insight into the discipline. Some studies did contain the flaws that were addressed by PCAST but others were simply omitted without any regards, despite potentially being useful. The fact that PCAST decided to discard challenges to class characteristics was also questionable, since class characteristics do contribute to the calculation of error rates and the likelihood ratio of matching. Although PCAST criticised expert experience and skill as subjective, statistical data still has to be based on the skill of someone with specialized knowledge to analyse them. Further research topics to address gaps in knowledge of shoeprint comparison analysis were discussed in the thesis. In conclusion, I believe the PCAST report should only serve as a guideline for what could be achieved in feature comparison methods. While the scientific communities acknowledge the gaps existed in the discipline, it would be preposterous to disregard decades of existing research.