A New Complex Investigation Model for Searching, Mapping, and Identifying Disappeared Persons in Argentina

INÉS CARIDI; ENRIQUE ALVAREZ; CARLOS SOMIGLIANA; MERCEDES SALADO PUERTO

New Orleans

Congreso; 69 Annual Scientific Meeting: Our Future reflects our past: The Evolution of Forensic Sciences; 2017

AMERICAN ACADEMY OF FORENSIC SCIENCES

After attending this presentation, attendees will gain a new viewpoint of the problem related to the identification of human remains of disappeared people, in particular, the importance of applying innovative frameworks of research, as in this case, the combination of complex networks, Bayesian inference and statistical evaluation tests in the systematization and analysis of the data. This presentation will impact the forensic science community by providing a framework of complex networks and Bayesian inference, to be applied to the identification of human remains belonging to disappeared persons during the last military dictatorship in Argentina (1976-1983). During the last 32 years, the Argentine Forensic Anthropology Team (EAAF), had been using a multidisciplinary approach (archaeology, anthropology, odontology, pathology and genetics) to recover and identify the remains of thousands of disappeared in the country. Typically, after their kidnapping, people were taken to illegal detention centres, tortured and killed. Their unidentified bodies, were buried in individual and common graves in official cemeteries or clandestine mass graves at military/police compounds. The need to generate hypotheses of identity for the recovered  remains, triggered the interest on applying an alternative model for the analysis of the information. This new model has mathematically systematized non-genetic variables, resulting from information obtained by already solved identifications in particular events, in such a way that can be used in the search, GIS mapping and generation of new hypothesis of identity for unsolved related cases. In other words, this dynamic model based on complex networks and Bayesian inference is able to ?learn? from identified cases generating after a probabilistic ranking of candidates forunidentified related cases. The geographical and temporal systematized variables, using georeferenced data and image processing techniques, from the identified skeletal remains are analyzed within a Bayesian framework. In a second stage, the data are included on a complex network, where connections are established between related cases by using rules based on the individuals´attributes. Previous to this, the most effective rules in defining the network are evaluated. Once the network has been formalized, a ranking of candidates for unsolved cases (recovered skeletal remains) is produced. The advantage of this complex model is, among others, to minimize bias in the investigation of related cases, providing probabilistic values to connected cases. This model has a high applicability in the mapping and search of burial sites, identification of human remains and systematization of information, essential in the investigation of massive number of victims as in crimes again humanity, conflict or migration cases. The developed software, which includes processing of raw data, calculations and visualization uses R language, a free software and open source environment for statistical computing and graphics.