Using solved cases and expert knowledge to prioritize the search of disappeared people

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

Ciudad Autónoma de Buenos Aires

Workshop; Latin American School and Workshop on Data Analysis and Mathematical Modeling of Social Science; 2016

Centro latinoamericno de formación interdisciplinaria (CELFI)

During the last 32 years, the Argentine Forensic Anthropology Team (EAAF) has been using a multidisciplinary approach to recover and identify the remains of thousands of disappeared people during the last military dictatorship in Argentina (1976-1983).The idea of this work is to learn from the already-resolved cases and from the expert knowledge of EAAF researchers to carry out new searches. We have mathematically systematized non-genetic variables and information obtained by already solved identifications in a particular event to generate a probabilistic ranking of candidates for the unidentified related cases of the same event, in a Bayesian framework. A particular event can be defined as a fact in which a set of people whose number is known were killed on a known day in a specific place. An example of an event is the Fatima's slaughter in the Province of Buenos Aires, where 30 people (20 men and 10 women) were killed on 20/8/1976 in this location. So far, 24 of the victims have been identified at different stages and 6 of them remain unidentified. The variables used are geographical and temporal (i.e. the place and date of individuals´ kidnapping) among others. We have constructed a specific geo-referential frame to address geographical variables that is related to the historical process in Argentina. The results have been evaluated using reserved data through cross-validation. Measures of predictability and efficiency have allowed us to determine the adequate parameter of the model for a particular event. Once the ranking has been constructed, people are represented on a network where the connections are established among them based on the individuals' attributes while they were alive. Those candidates bearing low probability values but strongly connected to other candidates bearing high probability values improve their position on the ranking. Finally, the information related to every skeleton is used to construct a specific ranking for each one.  Prioritizing the searches helps the work of constructing new hypotheses of identity, which are later evaluated with genetic evidence. The importance of establishing a ranking lies in the fact that the EAAF has no blood samples of the family members of all of the still-unidentified missing people. For this reason, the rankings can prioritize which families to solicit blood samples from. The methodology has been applied to different events, generating specific rankings for several cases of unidentified remains.