Given the need for digitalization to advance in the construction of a standardized care system for victims among the various areas of government, along with the Secretariat for Substantive Equality between Women and Men (Secretaría de Igualdad Sustantiva entre Mujeres y Hombres, SISEMH), we identified an opportunity to improve the first-contact points with women victims of violence by digitally storing their complaints. It allows escalating reports between areas without victims needing to narrate the events repeatedly and speeds up the possible submission of a complaint by shortening the bureaucratic gaps.
Women who seek direct attention have to repeatedly narrate the aggressions they experienced as they are referred to different government bodies and care institutions depending on the type of violence they experienced, generating a feeling of fatigue, revictimization, and poor perception of care and trust in democratic institutions.
Public officials collect the information by hand and transcribe it into a computer, which implies a significant loss of time and a window to greater human errors such as interpretation, paraphrasing, or omission of important information.
Women who are victims of violence in the state of Jalisco seeking counsel at the Metropolitan Unit for Comprehensive Care of Women and Children (Unidad Metropolitana de Atención Integral a Mujeres y Niñez, UREA).
Public officials of first contact attention, psychologists, systematizers, decision-makers, and public policy designers.
Victims come to the SISEMH/DAMVV physically seeking advice, support, and guidance to solve their cases. A social worker is appointed to interview them, who manually fills out a Single Registration Card (Cédula de Registro Único, CRU).
To implement a "speech to text" model program for the automation and digitalization of files and reports of (adult) women victims assisted in the Units of SISEMH in the State of Jalisco.
The solution proposes introducing a device that records the story of women victims of gender-based violence and using audio-to-text conversion AI models to digitally store the victim's story as narrated.
All the considerations related to the storage and use of sensitive data (victims' stories) were reviewed since the system currently used to store the databases is very vulnerable. Therefore, it is necessary to determine who will be the custodian of the data and responsible to produce specific legislation to protect victims' personal data.
It is also necessary to explicitly obtain the victims' consent to process their data.
With the support of Tec de Monterrey, the SISEMH of Jalisco assessed the current situation in the process of care for women victims of violence in Jalisco to analyze how AI could support the institution in its care processes.
To ensure the machine does not misunderstand the vocabulary used in the narratives (local expressions, idioms, slang); also, infrastructure, labeled training data, and test data are not currently available, and these problems may take a long time to solve.
The "story summarization model" will require intensive training, and the team of the Secretariat noted that this model might be difficult to implement due to time, data, infrastructure, and resources issues, and it is also unclear how to evaluate the impact of the model. Likewise, it is also unclear how to assess whether the summaries are complete and how such a summary would help the process.
Similar to implementation challenges related to the use of language.
Government of Jalisco - Tec de Monterrey
Colombia lanzó la Misión de Expertos en IA para evaluar las potencialidades de la IA y dar recomendaciones concretas en el corto, mediano y largo plazo.