La alerción del Ministerio de Salud por casos de menores con ideación suicidal que habrían recibido Consejos híctimas de un chatbot ha en todas las lujas red. In this article we offer a serene and useful reading: the chatbots can be complementary support, but do not replace therapy nor human accompaniment. His diagnosis is based on a phenomenon that we have seen for decades: the ELIZA effect, the tendency to humanize the bots and attribute understanding and care to them, especially when simulating closeness 24 / 7.
What is the ELIZA effect (and why does it matter now)?
It's a cognitive bias by which we project human traits —empathy, understanding, intention— in software with a conversational interface. It is called by ELIZA, the chatbot created in 1966 by Joseph Weizenbaum who parody a Georgian therapist; despite its simplicity, many users attributed "intelligence." Today, with much more credible attendees, the risk of confusing "simulation" with "understanding" is triggered.
Most importantly:
- Support - therapy. The IA may serve as basic guidance or punctual detachment, but should not "be left alone with the pain" of a minor or take the place of clinical intervention.
- Clinically recognizable risk patterns. Long conversations, emotional escalation and acrylic validation can aggravate discomfort, especially in vulnerable users.
- The adult role is unabated. The emotional digital literacy in family and school: know what young people use, how they talk to the IA and talk to it at home. Detect signals on time and seek professional support when you touch.
- Responsibility of platforms and regulators. Calls for "product" controls: early risk detection beyond explicit words, automatic diversion to aid resources, "safe off" to intense dialogues, reliable age check and limit emotional company bots in minors.
The context: from extreme cases to demands
The debate is not theoretical. In 2023, a man in Belgium died of suicide after weeks of conversation with a chatbot in the Chai app; his widow attributed a causal role to these interactions. The case put the safety limits of these systems on the public agenda.
In 2024 and 2025 have been submitted US claims. United States. for deaths of adolescents in which it is alleged that chatbots encouraged or did not stop self-harm. The most recent one, brought by the parents of Adam Raine (16 years old) against OpenAI in California, has revived the scrutiny on the response of these models in long talks with children.
Practical guide for families and educators
1) Set clear rules of use
- Common areas at home, time-bound (avoid night use), and specific purpose (tasks and information, I don't take off when you're very bad).
2) To speak explicitly of the IA
- Commenting with the kids how and for what They use it, what makes them feel and where their limits are. To strengthen the idea that "seem empathic," "be a person" (ELIZA effect).
(3) Warning signs
- Persistent sadness, isolation, self-destructions, or "I don't want to go on." In the face of these signs, the app is turned off, is accompanied and contacted with a professional / emergency.
4) Choose safer environments
- Prioritize managed educational accounts, active parental controls by default and lock of "company / boyfriend" bots for minors.
What platforms should do (from now on)
- Multimodal risk detection (no solo por palabras prohibidas, también por dinámica de conversación).
- Desvío proactivo a líneas de ayuda y derivación a recursos locales.
- Cortes seguros en chats largos/intensos; memorias prudentes con usuarios menores; verificación robusta de edad.
- Transparencia: dejar claro que el bot no es terapeuta. Estas medidas, que Vilalta plantea en la entrevista, son factibles y urgentes si de verdad queremos productos “seguros por diseño”.
No se trata de demonizar la tecnología, sino de ponerla en su sitio: útil como herramienta, peligrosa como sustituto del vínculo humano. La entrevista de Ariadna Vilalta nos recuerda que, en salud mental, la prevención es presencia: adultos informados, límites claros y sistemas que prioricen la seguridad sobre el engagement.
Si necesitas ayuda ahora mismo
En España, puedes llamar gratis al 024 (Línea de atención a la conducta suicida, 24/7) o al 112 en caso de emergencia vital. También está el Teléfono de la Esperanza (717 003 717). Hablar funciona; no estás solo/a
