The next big challenge for the A Coruña engineer Miguel Anxo Pérez Vila is for his creation to learn to speak perfect Spanish. He has just developed an artificial intelligence system capable of detecting suicidal tendencies by analyzing activity on social networks, work that has already earned him an award for the Best Doctoral Thesis of 2024, granted by the Spanish Society for Natural Language Processing. Now, the challenge is to adapt the model, trained mostly with Anglo-Saxon data, so that it works with the same precision in the Spanish-speaking world.
In fact, the key to its operation is far from simply hunting for words like “sadness” or “depression.” The technology dives into the principles of psycholinguistics to find subtle but identifiable traces that mental disorders leave in the way we express ourselves. It analyzes the structure of sentences and the language that a person uses in writing to identify patterns that may reveal incipient depression.
In this sense, the system does not raise the alarm due to an isolated publication, a nuance that defines its reliability. Its true strength lies in the accumulation of evidence over timeallowing you to build a risk profile on a solid, ongoing foundation, rather than reacting to a one-off comment that could be taken out of context.
A digital safety net for mental health
Therefore, the final objective is not surveillance, but pure and simple prevention. Once the tool detects a possible case of risk, its function is to activate protocols to proactively offer help to the person. It is not about exposing anyone, but rather extending a helping hand in the digital environment, discreetly suggesting helplines or access to support chats with professionals.
Ultimately, technology becomes a discreet and anonymous bridge that connects those who suffer in silence with those who can help them. Pérez Vila’s proposal seeks to use artificial intelligence not as an end in itself, but as a means to facilitate access to vital resources that, at the right time, can make a difference.