AI and clinical data to prevent violence against women

These Italian initiatives are part of a larger ecosystem of international projects experimenting with complementary approaches. In Europe, for example, the EU Commission-funded Shield project acts mainly in the digital environment frequented by adolescents, using chatbots and language analysis systems to recognise signs of control, manipulation and abuse in everyday communications.

Outside Europe, the experiences are equally significant. In Canada, the Institute for clinical evaluative sciences (Ices) uses information from large health databases to identify correlations between repeat visits, types of trauma and subsequent interventions by anti-violence services. Ices analyses have returned spatial maps and useful indicators to identify areas and population groups where violence tends to emerge later than the first clinical signs.

In the United Kingdom, in the south east of the country, health services and law enforcement experiment with models that integrate clinical, historical and social data to estimate the likelihood of a known situation degenerating. They do not replace human assessment, but refine the ability to allocate resources and direct interventions when necessary.

In the United States, healthcare networks such as Kaiser Permanente employ advanced machine-learning models on electronic medical records to identify recurring anomalies: apparently accidental injuries, patients' accesses close in time, discontinuous prescriptions of analgesics or anxiolytics. In several publications, these models have outperformed traditional manual tools in their predictive ability on severe outcomes when used under close clinical supervision.

The operational logic is similar everywhere: information is anonymised and processed with semantic extraction techniques, context variables such as frequency of trauma, incongruent explanations, temporal patterns are identified, the model is trained on already known cases and, after a lengthy verification phase with human experts, returns a statistical indicator. However, the responsibility for interpretation always remains in the hands of the practitioner.

The value of these instruments is evident when one considers that domestic violence rarely manifests itself with a single serious episode, since it is often preceded by a sequence of signals distributed over time. Reducing the gap between those first signals and the initiation of a protection course can prevent the intensification of violent acts that become irreversible. This anticipatory capacity represents the concrete contribution of predictive systems.

The economic aspect is equally relevant. Gender-based violence produces very high costs for public systems in terms of repeated hospital admissions, hospitalisations, work absences, psychological treatment, court costs. Prevention, especially when possible at an early stage, significantly reduces these costs. Investing in predictive technologies and the ability of services to respond in a timely manner can generate structural savings as well as health and social benefits.

On a regulatory level, these technologies are in line with international obligations that require states to strengthen early detection mechanisms, as required by the Istanbul Convention. At the same time, they must comply with the principles established by the European AI Act (the first EU regulation establishing a legal framework for artificial intelligence, ed), which requires transparency, verifiability and full human control. It is a balance that requires investment not only in technology, but in governance, training and independent supervision.

The president of the Fondazione CRT, Anna Maria Poggi, sums up this vision: 'Technology, when guided by responsibility and an ethical vision, can become a valuable ally in protecting the most vulnerable people. This is why Fondazione CRT has supported the ViDeS project from the outset. ViDeS puts artificial intelligence at the service of the prevention of gender-based violence, helping to recognise often invisible signs more quickly,' he told Il Sole 24 Ore, 'ViDeS represents a concrete step forward in supporting the work of health professionals and strengthening the protection network around women. The project is the result of a scientific collaboration of great value, combining different competences, a precious capital of our territory'.

Predictive projects actually offer women concrete benefits. They make it possible to recognise signs of risk at an early stage and to activate personalised protection pathways, with access to anti-violence centres and targeted support, through spaces where victims can come forward with their experiences in a safe and judgement-free context. Thanks to the integration of hospitals, general practitioners and territorial services, these projects promote faster and more coordinated access to available resources, ensuring continuity of care and reducing the risk of women being excluded from protection pathways.

In this way, technology becomes a concrete tool to increase protection and support for those experiencing violence. Technology does not replace the caring relationship, it does not decide and it does not judge. It does, however, make it possible to identify earlier what today is still in danger of being addressed too late. In an area where the time factor can lead to dramatic differences in outcomes, this type of innovation represents a concrete opportunity to intervene more promptly and build public policies based on real and verifiable data.