Sironi Emanuele, Pinchi Vilma, Pradella Francesco, Focardi Martina, Bozza Silvia, Taroni Franco
School of Criminal Justice, Building Batochime, University of Lausanne, 1015 Lausanne-Dorigny, Switzerland.
Section of Forensic Medical Sciences, Department of Health Sciences, University of Florence, Largo Brambilla 3, 50134 Firenze, Italy.
J Forensic Leg Med. 2018 Apr;55:23-32. doi: 10.1016/j.jflm.2018.02.005. Epub 2018 Feb 9.
Not only does the Bayesian approach offer a rational and logical environment for evidence evaluation in a forensic framework, but it also allows scientists to coherently deal with uncertainty related to a collection of multiple items of evidence, due to its flexible nature. Such flexibility might come at the expense of elevated computational complexity, which can be handled by using specific probabilistic graphical tools, namely Bayesian networks. In the current work, such probabilistic tools are used for evaluating dental evidence related to the development of third molars. A set of relevant properties characterizing the graphical models are discussed and Bayesian networks are implemented to deal with the inferential process laying beyond the estimation procedure, as well as to provide age estimates. Such properties include operationality, flexibility, coherence, transparence and sensitivity. A data sample composed of Italian subjects was employed for the analysis; results were in agreement with previous studies in terms of point estimate and age classification. The influence of the prior probability elicitation in terms of Bayesian estimate and classifies was also analyzed. Findings also supported the opportunity to take into consideration multiple teeth in the evaluative procedure, since it can be shown this results in an increased robustness towards the prior probability elicitation process, as well as in more favorable outcomes from a forensic perspective.
贝叶斯方法不仅为法医框架下的证据评估提供了一个合理且符合逻辑的环境，而且由于其灵活性，还允许科学家连贯地处理与多个证据项目集合相关的不确定性。这种灵活性可能是以更高的计算复杂性为代价的，而这可以通过使用特定的概率图形工具，即贝叶斯网络来处理。在当前的工作中，此类概率工具被用于评估与第三磨牙发育相关的牙科证据。讨论了一组表征图形模型的相关属性，并实施了贝叶斯网络来处理超出估计程序的推理过程，以及提供年龄估计。这些属性包括可操作性、灵活性、连贯性、透明度和敏感性。分析采用了由意大利受试者组成的数据样本；结果在点估计和年龄分类方面与先前的研究一致。还分析了先验概率引出在贝叶斯估计和分类方面的影响。研究结果还支持在评估过程中考虑多颗牙齿的机会，因为可以证明这会提高对先验概率引出过程的稳健性，以及从法医角度来看会产生更有利的结果。
