patterns through Big Data reveal clinical evidence of developing depression
En una época en que las redes sociales se han convertido en el diario íntimo de millones de personas, un ingeniero gallego ha encontrado en esa huella digital una señal clínica: los patrones con que compartimos música, imágenes y palabras pueden revelar el inicio silencioso de una depresión. Miguel Anxo Pérez ha desarrollado un sistema de inteligencia artificial que traduce comportamientos cotidianos en línea en evidencia médica temprana, apostando por la detección antes de que la enfermedad arraigue. Su trabajo nos recuerda que lo que elegimos mostrar al mundo —incluso inconscientemente— habla de lo que llevamos por dentro.
- La depresión avanza a menudo en silencio durante meses antes de ser diagnosticada, y ese intervalo invisible es precisamente donde este sistema quiere intervenir.
- El algoritmo detecta señales de alerta en comportamientos tan cotidianos como acompañar una foto con música melancólica o reducir la frecuencia de publicaciones, convirtiendo gestos digitales en marcadores clínicos.
- La tecnología fue reconocida como la mejor tesis doctoral de 2024, lo que le otorga respaldo académico, pero su aplicación real —en manos de individuos, sistemas sanitarios o plataformas— aún está por definirse.
- El potencial de intervención temprana choca con preguntas sin resolver sobre privacidad y consentimiento, tensiones que determinarán si la herramienta llega a quienes más la necesitan.
Miguel Anxo Pérez, ingeniero informático gallego, ha construido un sistema de inteligencia artificial capaz de identificar los primeros síntomas de depresión analizando el comportamiento de las personas en redes sociales. El sistema no busca confesiones explícitas, sino patrones: la música que acompaña a una foto, la frecuencia de las publicaciones, el tono emocional del contenido compartido. A partir de esos rastros digitales, extrae lo que Pérez denomina evidencia clínica de riesgo depresivo.
Detrás del proyecto hay una observación tan sencilla como reveladora: las personas de hoy externalizan su vida emocional de formas que generaciones anteriores nunca hicieron. Una joven entrevistada lo ilustra con naturalidad: cuando está triste, sube fotos con música melancólica; cuando está bien, comparte momentos alegres con amigos. El algoritmo aprende esas diferencias y las convierte en señales medibles.
El sistema apunta a un vacío real en la atención a la salud mental. La depresión suele desarrollarse de forma gradual, con señales tempranas que pasan desapercibidas hasta que la enfermedad ya ha echado raíces. Pérez concibe las redes sociales no como ruido, sino como una fuente de información clínica que puede anticipar ese deterioro.
El trabajo fue reconocido como la mejor tesis doctoral de 2024, un aval que confirma su solidez académica. Lo que queda abierto es cómo se desplegará en la práctica: si será una herramienta de automonitoreo, un recurso para sistemas sanitarios o un debate sobre los límites de la privacidad y el consentimiento. Por ahora, la premisa central se sostiene: la manera en que nos presentamos en línea contiene información real sobre nuestro estado mental, y esa información puede ser leída.
Miguel Anxo Pérez, a computer engineer from Galicia, has built an artificial intelligence system that watches what people post on social media and learns to recognize the early signs of depression. The system works by identifying patterns in how we behave online—the songs we pair with our photos, the frequency of our posts, the tone of what we share—and translating those patterns into clinical evidence that someone may be developing depression.
The insight behind the work is straightforward but powerful: people now broadcast their emotional lives in ways previous generations never did. Where someone struggling decades ago might have confided only in a close friend, today that same person is likely to post about it. A young woman interviewed for the story describes the phenomenon plainly: when she is sad, she uploads photos paired with melancholy music; when she is happy, she shares moments with friends. The algorithm learns these distinctions. It extracts what Pérez calls "patterns through Big Data and artificial intelligence algorithms" that can surface clinical evidence of depressive symptoms—not guesses, but measurable behavioral markers that suggest someone is at risk.
The technology addresses a genuine gap in mental health care. Depression often develops gradually, with early warning signs that go unnoticed until the condition has taken hold. By analyzing the digital traces people leave behind voluntarily, the system aims to catch those warning signs before they deepen. Pérez emphasizes that the goal is early detection, "taking advantage of all this new information that social networks give us and the evidence that we can manifest certain symptoms of depression." The approach treats social media not as a distraction but as a source of clinical insight.
Interestingly, the research also touches on how older people use these platforms differently. For some, social media serves a genuine purpose—a way to feel less isolated, to browse products, to stay connected to friends. One older woman explains that scrolling through social networks helps her when she is bored, giving her something to do and a way to maintain contact. The technology does not judge these uses; it simply reads what is already there.
Pérez's work has been recognized as the best doctoral thesis of 2024, a validation that the approach has merit within the academic and research community. What remains to be seen is how the technology will be deployed in practice—whether it becomes a tool that individuals can use to monitor their own mental health, whether healthcare systems adopt it for screening, or whether it raises questions about privacy and consent that need to be resolved first. For now, the core insight stands: the way we present ourselves online contains real information about our mental state, and that information can be read.
Citações Notáveis
Algorithms extract patterns through Big Data and artificial intelligence that can provide clinical evidence of symptoms to detect if certain people are developing depression— Miguel Anxo Pérez, computer engineer
When people are sad, they post photos with sad music; when happy, they share moments with friends— Young woman interviewed about social media behavior
A Conversa do Hearth Outra perspectiva sobre a história
So the AI isn't diagnosing depression—it's spotting patterns that suggest someone might be developing it. How confident is the system?
It extracts clinical evidence through behavioral markers. The patterns are real—how people post, what music they pair with photos, the rhythm of their activity. But this is early-stage research. The thesis was recognized as excellent, but deployment is another question entirely.
What happens when the system flags someone? Does it alert them? A doctor? Their family?
That's not addressed in what we know. The research focuses on detection capability. The practical question of what to do with that detection—who gets told, how, whether it's voluntary—those are the hard problems still ahead.
Doesn't this feel invasive? Analyzing someone's posts to predict their mental state?
It could be. But people are already publishing this information. The question is whether that data gets read clinically or just algorithmically for engagement. There's a difference between a system designed to sell ads and one designed to catch depression early. The ethics depend entirely on how it's used.
The older woman in the story—she uses social media to fight loneliness. Could the system misread that as depression?
That's a real risk. Activity itself isn't the marker. It's the pattern of what you share, how you share it, the emotional content. But yes, distinguishing between someone scrolling for connection and someone spiraling would require careful calibration.
What's the next step for this technology?
Clinical trials, probably. Testing it against actual patient populations, seeing if early detection actually leads to better outcomes. And then the harder conversation: privacy frameworks, consent, who owns the data, who can access it.