Innovation Day – Saving lives: Leveraging AI & Data to prevent violence

“Innovation Day exposes United Nations employees to new ideas, practices, behaviours, and concepts. These briefings, which cover a wide variety of innovative and creative topics, are generally held twice a month. Since 2019, the ideas and stories shared through Innovation Day have helped employees to: introduce new methodologies and approaches in their day-to-day work, find ways to apply innovation and creativity in mandate delivery, connect with others outside their network, expanding their social capital, and scale ideas across the global Secretariat.”

On this occasion, the UNODC-INEGI Center of Excellence (CdE) participated with the presentation of the initiative “Saving lives: harnessing Artificial Intelligence and data to prevent violence.” The objective of this initiative is to discover patterns in the reports of emergency calls to 9-1-1 to detect cases of violence against women that, due to the nature of the classification assigned by the operators to attend to the emergency, could not be be identified.

In this sense, the Global Estimates on Femicide published by UNODC and UN Women in 2022 are presented, indicates that homicides of women globally 55% are committed by their partners and members of their families, compared to 12% in the case of homicides of men.

During the presentation, it was highlighted that the detection of cases and statistics are based on traditional sources of information, such as administrative records and surveys, which have certain advantages and disadvantages. Therefore, this initiative is based on the use of non-traditional sources, such as emergency call records, which have the characteristic of being an unexplored data source due to the complexity of unstructured texts, but which contain valuable information, for example, the identification of cases of violence against women.

In general terms, the methodology used is described, which is developed in five phases that include context analysis, data review and cleaning, and the design and training of models based on Artificial Neural Networks for Natural Language Processing.

Finally, Adriana Oropeza, coordinator of the CoE, pointed out the main challenges that the CoE has faced when trying to replicate the exercise carried out due to access, availability, privacy and security of data; to the legal frameworks and the need to implement statistical processes that make the phenomena of interest visible.