Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Case Study STATISTICAL CRIME FIGHTERS: IMPROVING FORENSIC SCIENCE Scientific evidence has a crucial role to play in the administration of justice. It is important that it is evaluated objectively and interpreted clearly for the courts, since failure to do so can lead to appeals or even miscarriages of justice. THE COMPARISON PROBLEM Professor Colin Aitken and his collaborators have developed Bayesian statistical methods to aid in the evaluation and interpretation of forensic scientific evidence. These new methods have reduced cost, increased accuracy and improved the interpretation of the value of evidence. ASSESSING UNCERTAINTY An understanding of the value of forensic evidence relies heavily on an assessment of uncertainty. Imagine, for example, that fragments of glass found at a crime scene are believed to come from a broken bottle found in possession of a suspect. The chemical composition of the glass in the fragments and in the bottle is analysed. What is the value of similarities between the composition of the two samples? In order to improve the evaluation of such evidence, Professor Aitken and other researchers from the Maxwell Institute for Mathematical Sciences developed new Bayesian statistical methods. These methods enable forensic scientists worldwide to interpret their data reliably. The glass fragments can come from the bottle found in possession of the suspect, or they may come from another bottle. To help the court estimate the relative likelihood of these two possibilities, Professor Aitken and collaborators calculate a so-called likelihood ratio (LR) that takes into account variations within glass bottles and between multiple bottles, first assuming the fragments came from the suspect’s bottle, and second assuming they came from another bottle. TRANSFORMATIVE IMPACT A senior forensic statistician at the Netherlands Forensic Institute says: ‘The groundbreaking work of Aitken and others has transformed the way we evaluate forensic evidence … the LR method is the next step in the evolution from forensic craft to forensic science.’ In comparison of evidence involving glass, a key question is how much support measurements give to the proposition that the fragments in the upper picture originated from the bottle in the lower picture as compared with the proposition that they did not. In another scenario, when large consignments of potentially incriminating material are seized (such as pills suspected of containing illegal drugs, or computer files suspected of containing illegal material), the police want to estimate the proportion that is illicit. PROBABILITIES IN THE BALANCE Examination of every item is time-consuming, costly and stressful. Professor Aitken developed procedures for determining the optimal size of samples that should be examined. Through a careful assessment of the uncertainties associated with an examination of only a fraction of a consignment, his research ensures that investigators can sample fewer items and still provide evidence that is fit for purpose in a criminal trial. Professor Aitken’s sampling protocols have been widely adopted. They have been recommended to forensic laboratories by the Crown Office in Scotland and in guidelines by the United Nations Office on Drugs and Crime. Sampling software based on Professor Aitken’s statistical methods is available through the European Network of Forensic Science Institutes. This software allows forensic scientists without a strong background in statistics to benefit from cutting-edge Bayesian statistical methods. RELEVANT RESEARCH PUBLICATIONS FROM THE AITKEN GROUP Evaluation of evidence: the left-hand scale shows the weight associated with the prosecution’s proposition Hp (that the suspect is guilty) as the logarithm of the probability of the evidence E if Hp true. The right-hand scale shows the weight associated with the defence proposition Hd (that the suspect is innocent) as the logarithm of the probability of the evidence E if Hd true. Aitken, CGG (1999) ‘Sampling – how big a sample?’ Journal of Forensic Sciences, 44, pp 750-760. Aitken, CGG and Lucy, D (2002) ‘Estimation of the quantity of a drug in a consignment from measurements on a sample’ Journal of Forensic Sciences, 47, pp 968-975. Aitken, CGG and Lucy, D (2004) ‘Evaluation of trace evidence in the form of multivariate data’ Applied Statistics, 53, pp 109-122, with corrigendum pp 665-666. Aitken, CGG, Zadora, G and Lucy, D (2007) ‘A two-level model for evidence evaluation’ Journal of Forensic Sciences, 52, pp 412-419. Aitken, CGG. and Taroni, F (2004, 2nd edition), ‘Statistics and the evaluation of evidence for forensic scientists’ John Wiley & Sons Ltd. CONTACT Professor Colin GG Aitken Maxwell Institute for Mathematical Sciences School of Mathematics +44 (0)131 650 5060 [email protected] www.maths.ed.ac.uk/~cgga If you require this document in an alternative format, such as large print, please contact: Communications and Marketing Tel: +44 (0)131 650 2252 Email: [email protected]