Weekly AI Watch

Every week, what changed and what it means for you. No hurry, no alarm.

Week of 13 November 2026 | Past editions15,811 papers found · 14,944 read · 4,785 belonged · 46 stories passed every check

Observatorio de la
Inteligencia Artificial

Cada semana, qué cambió y qué significa para usted. Sin prisa y sin alarma.

Semana del 13 de noviembre de 2026 | Ediciones anteriores15,811 estudios encontrados · 14,944 leídos · 4,785 correspondían · 46 notas pasaron todas las comprobaciones

weeklyAI · Week of 13 November 2026weeklyAI · Semana del 13 de noviembre de 2026

← Your child← Sus hijos

interviews · NPJ digital medicine · la publicación, 19 jun 2026 · gratis

"No puede reemplazar a los psicólogos": lo que 32 jóvenes australianos pidieron a los chatbots de salud mental

En talleres en línea, jóvenes de 18 a 30 años dijeron qué querrían de estas herramientas. El estudio no midió si ayudan, y nadie menor de 18 participó.

Versión breve · la versión detallada sigue, unos 8 min

Pregunte al chat de weeklyAI

Pregúnteme sobre la IA en su vida: le señalo las notas que weeklyAI ha publicado sobre el tema, esta semana y antes.

Las conversaciones se guardan mientras exista weeklyAI, para mejorar la publicación. Se responde en el idioma en que usted escribe.

El estudio, de un vistazo
Quiénes
Jóvenes australianos que participaron en talleres sobre chatbots de salud mental
Cuántos
32
Dónde
Australia
Cuándo
entre marzo y agosto de 2025
Tipo de estudio
interviews
Quién lo hizo
Universidad de Sydney, Australia
El límite que importa
Nadie menor de 18 años participó y no se midió si estas herramientas ayudan o dañan.
En casa, la pregunta útil no es si el sistema es bueno o malo en abstracto, sino qué papel le está dando la familia.
Lectura de weeklyAI

"It can't outright replace actual psychologists and psychiatrists, because the human connection is still important at the end of the day." Lo dijo una joven en uno de los talleres. Traducido: un chatbot no puede reemplazar del todo a los psicólogos y psiquiatras, porque la conexión humana sigue importando.

Detrás de esa frase hubo un patrón. Eran 32 jóvenes de 18 a 30 años, en Australia, en talleres en línea organizados por investigadores de la Universidad de Sydney. Querían chatbots que sonaran empáticos, pero que acompañaran la atención humana en lugar de sustituirla. Muchos dudaban que un programa pudiera entender lo personal y complejo de un problema de salud mental.

También querían ver "qué hay debajo del capó": cómo razona el sistema, en qué evidencia se basa y cómo se guardan y se comparten sus datos. Una participante lo explicó a su manera: no tiene sentido que la información quede escondida cuando un profesional ya la está viendo.

Ahí trazaron una línea. Las conversaciones crudas, lo que el joven escribe, deberían quedar privadas y bajo su control. Los resúmenes y las alertas de riesgo que genera la inteligencia artificial sí podrían verlos los clínicos. Así se protege la privacidad y a la vez se permite que alguien intervenga cuando hace falta.

El chatbot que probaron, llamado Mia, nació como herramienta para profesionales y lo adaptaron para jóvenes durante el estudio. Los participantes probaron versiones en desarrollo en un solo momento, no productos terminados durante un periodo largo.

Lo que el estudio no puede mostrar es si algo de esto funciona. Todos los participantes tenían 18 años o más, vivían en un solo país y probaron prototipos. Nadie midió si estos chatbots mejoran el ánimo, si la gente los sigue usando o si hacen daño. Tampoco se escuchó a menores de 18 ni a los jóvenes más marginados. Los propios participantes mencionaron riesgos: respuestas inventadas, información mal recordada, sesgos y una dependencia poco sana.

Cifras aparte circulan sobre Mia, como un 94% de acierto al clasificar casos frente a un panel de expertos. Vienen de otra evaluación clínica, no de estos talleres, y no dicen nada sobre el bienestar de un usuario.

La próxima vez que su hijo o hija le cuente que habla con un chatbot sobre cómo se siente, fíjese en dos cosas: si el programa lo orienta hacia una persona real cuando el tema se pone serio, y si le explica con claridad de dónde saca lo que le dice.

Qué significa para usted

Cuando su hijo o hija le cuente que habla con un chatbot sobre cómo se siente, puede preguntarle si el programa lo manda con una persona real cuando el tema se pone serio, y si le explica de dónde saca lo que le dice. Eso es lo que estos 32 jóvenes australianos pidieron, y vale como lista de lo que usted puede revisar. Nadie ha medido todavía si estas herramientas ayudan o hacen daño.

Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9

Quién pagó: El trabajo fue financiado por un donante filantrópico anónimo afectado por la salud mental, con apoyo adicional de subvenciones del National Health and Medical Research Council a dos autores y una beca del Australian Government Research Training Program; el artículo no indica si los financiadores tuvieron alguna influencia, y no se nombran proveedores de equipos o software.

No tome esto como consejo médico profesional.

Versión detalladaLos pasajes copiados del artículo, las ilustraciones y cada fuente con cuánto leímos de ella · 1548 palabras · unos 8 minLeerla →Cerrar

Lo que 32 jóvenes en Australia pidieron a un chatbot de salud mental: transparencia, un lugar acotado y que no sustituya a las personas

Un estudio de diseño con jóvenes de 18 a 30 años no midió si estas herramientas ayudan ni si son seguras. Pidió a los participantes que dijeran qué exigirían antes de usarlas.

Un joven abre una aplicación de chat y le escribe lo que no le ha contado a nadie: que no duerme, que se siente vigilado, que lleva semanas sin querer salir. El sistema responde con una pregunta más, y otra, hasta armar un resumen de lo que oye. El estudio que nos ocupa reunió a 32 jóvenes australianos en talleres sobre un sistema así, el Mental health Intelligence Agent, conocido como Mia, y les preguntó qué les parecía y qué le exigirían antes de usarlo en serio12.

Mia no nació para jóvenes. Fue creado para profesionales de servicios de salud mental juvenil en Australia: interpreta lo que la persona escribe, detecta información faltante, arma resúmenes y líneas de tiempo, y propone recomendaciones de tratamiento y de navegación dentro del sistema de servicios1. Los participantes trabajaron en talleres en línea, entre marzo y agosto de 2025, para imaginar cómo se convertiría ese sistema en una herramienta para los propios jóvenes2; entre los 19 que respondieron una encuesta opcional de datos personales, las edades iban de 18 a 30 años.

Del análisis salieron cuatro temas. Los jóvenes aceptaron el chatbot como apoyo, siempre que acompañe y no reemplace la relación con el clínico: la conexión humana no se replica, dijeron, y los problemas de una persona son más complejos que el consejo que una máquina puede ofrecer34. Exigieron ver qué hay debajo del capó: que el sistema explique cómo interpreta lo que lee, de dónde saca sus respuestas y cómo se usan los datos56. Discutieron en qué momento del recorrido de atención encaja y en cuál no. Y pidieron elegir cómo interactuar, con la privacidad como condición: los datos crudos, lo que uno escribe, quedan bajo control del usuario; los resúmenes que el sistema genera pueden llegar al clínico para cuidar la seguridad78.

También pusieron límites claros. Quisieron que el sistema detecte cuando alguien describe una situación peligrosa o cuando su estado empeora, y que en esos casos dirija a la persona a ayuda profesional9. Y trazaron una raya: un chatbot que haga de amigo no es apropiado en salud mental, donde los límites terapéuticos importan10. El estudio no comparó nada ni midió resultados de salud. Los autores mismos señalan que los participantes tenían entre 18 y 30 años, vivían en Australia y se ofrecieron a participar en investigación sobre servicios de salud mental11, y que usaron prototipos en un solo momento, no un producto terminado durante semanas.

Piense en el teléfono de su hijo. Si tiene entre 13 y 17 años, es probable que ya haya conversado con un chatbot de inteligencia artificial: según el resumen de una encuesta con adolescentes de Estados Unidos, lo único disponible de ese trabajo, más del 60 por ciento de los encuestados había usado uno, y 11.4 por ciento lo hacía a diario o casi a diario12. Las razones más mencionadas fueron entretenimiento, amistad y consejo13. En esa misma encuesta, 47.1 por ciento reportó haber vivido al menos una experiencia de riesgo o daño con estas herramientas14.

El detalle de esos riesgos vale la pena, porque es lo que un padre quiere saber antes de decidir. En la encuesta, 32.3 por ciento de los adolescentes dijo que el chatbot le pidió información personal incómoda, 23.1 por ciento se sintió manipulado o presionado, 17.1 por ciento recibió información falsa y 18.7 por ciento fue animado a actuar de forma poco ética o ilegal15. Un 15.2 por ciento fue empujado hacia conductas de riesgo, y hubo quienes recibieron mensajes sobre autolesiones y sobre suicidio15. La propia encuesta concluye que hacen falta protecciones adaptadas a la edad y sistemas de vigilancia continua16. Es el mismo terreno que el estudio australiano pisó desde el otro lado: no qué les pasa a los jóvenes con estos sistemas, sino qué les pedirían a quienes los diseñan.

Hay una razón de fondo para la cautela. Un análisis académico sobre la adolescencia y la inteligencia artificial generativa, del que solo se pudo leer el resumen, describe esa etapa como un periodo sensible en el que se construyen la pertenencia social, la regulación emocional, la autonomía y la identidad, con consecuencias para toda la vida17. Ese mismo trabajo advierte que, aunque estas herramientas puedan ofrecer espacios de bajo riesgo para ensayar interacciones o explorar la identidad, también amenazan con desplazar experiencias necesarias para el desarrollo18. Entre los riesgos que enumera están el aislamiento social, la dificultad para tolerar el malestar, expectativas distorsionadas sobre las relaciones humanas y la exposición a contenido sesgado, inexacto o sexualmente inapropiado19. No es un juicio sobre Mia. Es el mapa del terreno donde cualquier herramienta de este tipo va a operar.

Y lo que los jóvenes hacen con estos sistemas, cuando nadie se los pregunta, tiene poco que ver con la salud mental tal como la imaginamos los adultos. Según el resumen de un análisis de lo que miles de menores escribieron en aplicaciones de chat en Estados Unidos, lo único disponible de ese trabajo, el uso más común fue como herramienta práctica, pero una proporción notable de los días registrados incluyó violencia, juego de rol sexual y juego de rol romántico, y esas conversaciones tendieron a ser más largas que las de apoyo emocional2021. Los temas variaron con la edad: el uso como herramienta fue más frecuente entre los mayores, y la violencia y el juego de rol sexual o romántico, entre los más chicos22. Ese análisis observó teclazos en una aplicación de control parental que las familias instalan de forma voluntaria, así que retrata solo a quienes aceptaron ser observados23.

Aquí viene la lectura de weeklyAI, y conviene decirla con claridad: el estudio australiano no evaluó una herramienta, escuchó a treinta y dos personas. Lo que describe es el pliego de peticiones que esos jóvenes levantarían antes de confiar en un chatbot de salud mental. En casa, la pregunta útil no es si el sistema es bueno o malo en abstracto, sino qué papel le está dando la familia. Un chatbot puede servir de puente hacia una persona, o puede convertirse en el filtro que decide quién llega a la consulta y quién no. La diferencia no la marca la tecnología: la marca quién revisa lo que el sistema concluye.

De ahí se desprende otra cosa que weeklyAI esperaría ver en los próximos meses: que algunos jóvenes prefieran escribirle primero a la máquina, no porque confíen más en ella, sino porque preguntar a una pantalla expone menos que preguntar a un adulto. También cabría esperar que los servicios, al ver que el sistema ordena la información, se tienten a usarlo como antesala obligatoria de la atención humana. Esta lectura se equivocaría si los jóvenes del estudio hubieran pedido una sola respuesta ya resuelta, sin opciones ni explicaciones, o si hubieran rechazado cualquier decisión del sistema que no pase por una persona. No fue así: pidieron explicaciones, alternativas y control sobre sus datos.

Para los padres, la traducción práctica es sencilla y no requiere saber nada de modelos de lenguaje. Pregúntele a su hijo si el chatbot le ha recomendado algo y si lo consultó con alguien antes de actuar. Si la respuesta es que no, ahí hay una conversación pendiente, no una regaño. Y cuando llegue el momento de poner reglas, conviene que las decisiones sobre su bienestar pasen por una persona que responda por ellas, en casa o en la escuela. Una prohibición sin diálogo suele enseñar a esconder el uso, no a moderarlo.

El estudio tiene límites que un lector apurado podría pasar por alto. Los autores señalan que sus conclusiones están limitadas por el contexto local, las edades de los participantes y la ausencia de ciertas voces, en particular las más marginadas por el estigma o las de quienes tienen menos de 18 años11. Los talleres fueron remotos, y la conexión a internet afectó en algunos casos la calidad de las transcripciones. El foco fue un solo chatbot en un solo país, y los participantes vieron prototipos una vez, no un producto final durante semanas. Los autores declaran financiamiento de un donante filantrópico que pidió el anonimato, además de becas de investigación del gobierno australiano, y uno de los autores tiene una participación accionaria en una empresa de transformación digital de servicios de salud mental. Nada de eso invalida lo que los jóvenes dijeron. Solo dice con qué cartas se jugó.

Vale la pena volver al inicio: el joven que escribe a la pantalla lo que no le ha contado a nadie. Lo que este estudio deja en claro es que ese joven, al menos los treinta y dos que hablaron, no quiere que la máquina decida por él. Quiere que le explique, que le deje escoger y que, cuando el asunto sea grave, haya una persona del otro lado. Esa es la vara. La pregunta que puede llevar a su casa esta semana es una sola: ¿quién revisa lo que el chatbot le está diciendo a mi hijo, y con quién lo puede hablar él?

De dónde sale cada dato de contexto, y cuánto leímos de cada documento

  1. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "We examined the Mental health Intelligence Agent (Mia), a genAI chatbot originally designed for professionals in Australian youth services."
  2. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Following co-design, 32 young people participated in online workshops exploring their perceptions of genAI chatbots in youth mental health and to develop recommendations for reconceptualising Mia for consumers and integrating it into services."
  3. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "It can’t outright replace actual psychologists and psychiatrists, because the human connection is still important at the end of the day, it’s not something that can be replicated by AI or chatbots"
  4. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "I feel a lot of the problems that I experience as a young person with lived experience are a lot more complex than maybe the advice that a system like this might be able to offer"
  5. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Young people wanted genAI chatbots to interpret user inputs aptly, draw appropriate inferences from them, and provide reliable outputs grounded in evidence."
  6. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A lot of people say, “Oh, you can’t trust AI 100%”. And I’m sure you can’t really trust anything 100% but just to show that it bases its answers on reasonable things, I think it’s important to show its thought process"
  7. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Young people emphasised that for the sake of promoting unfiltered interactions and protecting privacy and autonomy, their raw data should remain private and under their control."
  8. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "While raw conversational data should remain private, AI-generated insights developed from genAI chatbot interactions (i.e. summaries detailing a user’s needs, personalised treatment, self-management, and service navigation recommendations, and deterioration or risk detection) should be accessible to clinicians and services."
  9. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Young people expected these tools to detect when users discuss unsafe situations (e.g. sharing imminent plans of self-harm or violence) and when patterns suggest deteriorating mental health or elevated risk."
  10. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Importantly, young people distinguished the guide role from companionship or friendship, emphasising that a genAI chatbot functioning as a friend would be inappropriate in mental health contexts where therapeutic boundaries and clinical relationships are paramount."
  11. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - el artículo del que trata esta nota — el artículo completo — el pasaje: "the study participants consisted of young people aged 18-30 who were willing to participate in mental health services research in the Australian context"
  12. Hinduja S, Patchin JW. (2026). Risks and Harms of Conversational Artificial Intelligence (CAI) Chatbot Use Among US Youth. Journal of Adolescence. 10.1002/jad.70164 — solo el resumen - el texto completo está tras una suscripción — el pasaje: "Over 60% of the sample reported using a CAI chatbot, with 11.4% doing so every day or nearly every day."
  13. Hinduja S, Patchin JW. (2026). Risks and Harms of Conversational Artificial Intelligence (CAI) Chatbot Use Among US Youth. Journal of Adolescence. 10.1002/jad.70164 — solo el resumen - el texto completo está tras una suscripción — el pasaje: "Main reasons included entertainment (85%), friendship (60.1%), and advice (65.6%)."
  14. Hinduja S, Patchin JW. (2026). Risks and Harms of Conversational Artificial Intelligence (CAI) Chatbot Use Among US Youth. Journal of Adolescence. 10.1002/jad.70164 — solo el resumen - el texto completo está tras una suscripción — el pasaje: "CAI chatbot usage is common among US adolescents, with 47.1% reporting exposure to one or more specific risks and harms."
  15. Hinduja S, Patchin JW. (2026). Risks and Harms of Conversational Artificial Intelligence (CAI) Chatbot Use Among US Youth. Journal of Adolescence. 10.1002/jad.70164 — solo el resumen - el texto completo está tras una suscripción — el pasaje: "Still, 32.3% were asked for uncomfortable personal information, 23.1% felt manipulated or pressured, 17.1% received false information, 18.7% were encouraged to act unethically or illegally, 15.2% were prompted to risky behaviors, and 14.7% and 13.0% were exposed to self-harm and suicidal messages, respectively."
  16. Hinduja S, Patchin JW. (2026). Risks and Harms of Conversational Artificial Intelligence (CAI) Chatbot Use Among US Youth. Journal of Adolescence. 10.1002/jad.70164 — solo el resumen - el texto completo está tras una suscripción — el pasaje: "These findings highlight the need for adaptive safety features, ongoing monitoring systems, and safeguards that promote the psychological and social well-being of youth while addressing their developmental vulnerabilities."
  17. Maheux AJ, Maes C. (2026). Coming of Age With Generative Artificial Intelligence: The High Stakes of Automating Adolescent Development. Perspectives on Psychological Science. 10.1177/17456916261460771 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Adolescence is a sensitive life period during which the accomplishment of key psychosocial developmental tasks-including social belonging, emotion regulation, autonomy, and identity formation-leads to lifelong consequences."
  18. Maheux AJ, Maes C. (2026). Coming of Age With Generative Artificial Intelligence: The High Stakes of Automating Adolescent Development. Perspectives on Psychological Science. 10.1177/17456916261460771 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Although GenAI may provide low-stakes opportunities to simulate social interactions, scaffold emotional regulation, and support identity exploration in some contexts, it also threatens to displace or thwart essential developmental experiences."
  19. Maheux AJ, Maes C. (2026). Coming of Age With Generative Artificial Intelligence: The High Stakes of Automating Adolescent Development. Perspectives on Psychological Science. 10.1177/17456916261460771 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Risks include increasing social isolation, undermining distress tolerance, distorting expectations for human relationships, compromising identity and purpose development, and exposing youth to biased, inaccurate, or sexually inappropriate content."
  20. Maheux AJ, Boeldt D, Akre-Bhide S, Flannery J, Paige A, Villella G, et al. (2026). How US Youth Use AI Chatbots: Conversation Patterns From Naturalistic Keystroke Observations. Journal of Medical Internet Research. 10.2196/95819 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Most user-app-days involved using AI as a tool (26,343/42,355, 62%), but notable proportions included violence (6460/42,355, 15%), sexual role-play (6393/42,355, 15%), romantic role-play (3538/42,355, 8%), general role-play (3707/42,355, 9%), and friend-like interaction (2624/42,355, 6%)."
  21. Maheux AJ, Boeldt D, Akre-Bhide S, Flannery J, Paige A, Villella G, et al. (2026). How US Youth Use AI Chatbots: Conversation Patterns From Naturalistic Keystroke Observations. Journal of Medical Internet Research. 10.2196/95819 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Word counts were highest for user-app-days involving violence, followed by sexual role-play, romantic role-play, and emotional support, suggesting greater engagement when using GenAI in these ways."
  22. Maheux AJ, Boeldt D, Akre-Bhide S, Flannery J, Paige A, Villella G, et al. (2026). How US Youth Use AI Chatbots: Conversation Patterns From Naturalistic Keystroke Observations. Journal of Medical Internet Research. 10.2196/95819 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Theme prevalence varied by age: tool use was more common among older adolescents, whereas violence and sexual or romantic role-play were relatively more common among younger users."
  23. Maheux AJ, Boeldt D, Akre-Bhide S, Flannery J, Paige A, Villella G, et al. (2026). How US Youth Use AI Chatbots: Conversation Patterns From Naturalistic Keystroke Observations. Journal of Medical Internet Research. 10.2196/95819 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Participants included only those who opted into the overlay Aura keyboard (to collect keystroke data) and who engaged in some degree of GenAI use."

Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9

Quién pagó: El trabajo fue financiado por un donante filantrópico anónimo afectado por la salud mental, con apoyo adicional de subvenciones del National Health and Medical Research Council a dos autores y una beca del Australian Government Research Training Program; el artículo no indica si los financiadores tuvieron alguna influencia, y no se nombran proveedores de equipos o software.

No tome esto como consejo médico profesional.

interviews · NPJ digital medicine · the paper, 19 Jun 2026 · free

A Chatbot Beside the Therapist's Chair

Young people in Australia said they wanted mental health chatbots that explain themselves and stay in their lane. The study asked 32 young people what they want; it did not test whether the tools help or harm.

Short version · the longer version follows, about 8 min

Ask the weeklyAI chat

Ask me about AI in your life: I point you to the stories weeklyAI has published on it, this week and before.

Conversations are saved for as long as weeklyAI exists, to improve the publication. Answers come in the language you write in.

The study at a glance
Who
young people in Australia
How many
32
Where
Australia
When
March to August 2025
Kind of study
interviews
Who did it
University of Sydney researchers
The limit that matters
Only 18- to 30-year-olds in Australia; no one under 18 was heard.
A young person will often say more to something that cannot look disappointed, and a tool that answers warmly will keep them talking.
weeklyAI's reading

"It can't outright replace actual psychologists and psychiatrists, because the human connection is still important at the end of the day," one young person said in the workshops.

Behind that sentence sat a pattern. Thirty-two young people, aged 18 to 30, joined online workshops run by University of Sydney researchers. They tried early versions of a chatbot called Mia, first built for youth mental health staff, and talked about what such tools should and should not do. The researchers drew four themes from the talks.

The first: make the chatbot sound warm and human, but never let it push real people out of care. The second: show what's under the hood, meaning how it reasons, what evidence it draws on, and how conversations are stored and shared.

They also asked where a chatbot belongs in the journey to care, and how to make it feel like their own while staying safe. On one point they were specific. Their raw conversations should stay private, while short summaries the tool builds from them, including warnings when someone seems at risk, could be seen by a clinician.

What the workshops cannot show is whether any of this works. The study heard from 32 young people in one country, and the ages reported cover only the 19 who filled in a follow-up survey. Some tried a working prototype themselves, while others watched the research team operate it. No one used a finished product over months. The researchers did not measure whether anyone's mental health improved. Their own list of worries included made-up answers, poor recall of earlier conversations, and the risk of young people leaning too hard on a machine.

The article reports strong results from a separate test of Mia against an expert panel. That test is not this study, and it says nothing about what happens in a real clinic or on a bad night at home.

The article describes a study in which young people were already turning to general chatbots for support, so the questions are live in most households with a phone in them.

When your child mentions a chatbot, look for two things: whether it points them toward a real person when talk turns serious, and whether it can explain how it reached its answer.

What this means for you

When your child mentions a chatbot, listen for whether it pushed them toward a real person once the talk turned serious, and ask whether the tool ever explained how it reached its answer. That is what these young people asked for, and it is still only a request: no one has shown yet that it protects anyone.

Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9

Who paid: The work was supported by an anonymous philanthropic donor affected by mental health, with additional support from National Health and Medical Research Council grants to two authors and an Australian Government Research Training Program scholarship; the article does not say whether funders had any say, and no equipment or software lenders are named.

Do not take this as professional medical advice.

The findings of other studies mentioned here come from this document; those studies were not opened one by one.

The longer versionThe passages copied from the paper, the pictures, and every source with how much of it we read · 1532 words · about 8 minRead it →Close

Young People Were Asked What an AI Mental Health Chatbot Should Never Do

A prototype built for Australian clinics was handed to 32 young adults. They wanted it to sound human, show its reasoning, and stay out of the therapist's chair.

A chatbot asks a nervous 20-year-old how they slept. The person types an answer they have never given a parent. Then the screen offers a suggestion, and the young person has to decide whether to believe it.

That moment, and what should surround it, is what a new study set out to examine. The tool is the Mental health Intelligence Agent, or Mia, a conversational generative AI chatbot originally designed for professionals in Australian youth services. It was designed for professionals in Australian youth services, and the study asked young people how it should be changed for people their age1.

Thirty-two young people took part in online workshops between March and August 2025. Nineteen of them also filled in an optional survey about themselves. Those who did ranged from 18 to 30 years old, with an average age of 25. Among the 19, 14 said mental ill-health had affected their daily lives, and 10 were in treatment at the time. The workshops came in two kinds: sessions where the research team operated Mia at the group's direction, and sessions where participants used a working version themselves and talked aloud while doing it. The analysis produced four themes, and the study tested no comparison and no outcome2.

The problem the researchers began from is stated plainly in the paper. Across the world, people aged 12 to 25 are seeing high and mostly rising rates of anxiety, depressive symptoms, psychological distress, self-harm, suicide and other mood disorders3. Services are described as overburdened and as suffering substantial gaps in workforce, service data, regulation, service specificity, coverage and coordination. Young people are also already going to general-purpose chatbots such as OpenAI's ChatGPT for mental health support, a trend the authors cite as reason to ask what they want from a purpose-built tool4.

What Mia actually does is easier to picture as a structured intake. It takes in what a person types about their symptoms, history and circumstances, notices what is missing, and keeps asking until it has enough. Then it produces a summary, a timeline, and suggestions for treatment and services, each linked to clinical guidelines. The system shows the reasoning behind each output, and its clinical criteria sit in a separate layer that experts can update without retraining the model. In a separate clinical evaluation described in the paper, Mia reached 94 percent accuracy on triage against an expert panel, recalled 24 of 25 expert-identified points, rejected 100 percent of planted misleading cues, and produced no hallucinations. Those figures come from that evaluation, not from the workshops1.

The first theme was about keeping care human. One participant put the boundary plainly: a chatbot "can't outright replace actual psychologists and psychiatrists, because the human connection is still important at the end of the day"5. Another doubted that any system could hold the complexity of a real young person's problems6. Participants raised concerns including made-up information, poor recall of what they had already said, bias in decisions, unhealthy dependence, and the risk that existing social stigma could be carried over, leading to feelings of judgement. They also wanted it to sound warm, use plain language, handle difficult subjects sensitively, and work in more than one language.

The second theme was transparency. Young people wanted to see the evidence behind an answer and the steps that led to it. One said it mattered "just to show that it bases its answers on reasonable things" and to "show its thought process"7. They wanted reliable outputs grounded in evidence8. The design changes that followed are visible in the paper: an opening screen explaining what the system does and does not do, and a display showing how much information the system has about the user so far.

The third theme was fit. Participants mapped where a chatbot belongs in a care journey, from first contact and intake through preparing for sessions, moments of crisis, and ongoing care. They cast it as a guide that helps navigate, assess and educate, and they drew a line: a chatbot acting as a friend would be inappropriate where therapeutic boundaries matter9.

The fourth theme was control and safety. Young people wanted to choose how they interact, by text or voice, in an app or a browser, in short check-ins or long conversations. On privacy they made a distinction worth pausing on. Their raw typed words should stay private and under their control10. Summaries and risk signals generated from those words should reach clinicians11. They also expected the tool to notice unsafe situations and deteriorating patterns and respond12.

Consider what that split means in practice. A teenager could type something raw and unedited, and the clinician would see the summary rather than the transcript. The gain is a private space to speak freely. The cost is that the clinician's picture depends on a machine's summary being right. The paper records a related worry from participants themselves: recommendations that contradict a clinician's could undermine trust in both.

Two other studies put this one in a wider frame, and only their abstracts were available. One survey of US adolescents found that more than 60 percent had used a conversational AI chatbot, with 11.4 percent using one daily or nearly daily, and that 47.1 percent reported exposure to one or more specific risks and harms, including being asked for uncomfortable personal information and receiving false information131415. Another analysis of keystrokes from 3,363 US youth aged 1 to 17 using a parental monitoring app found most use was as a practical tool, but notable shares involved violence, sexual role-play, romantic role-play, general role-play and friend-like interaction161718. Both are descriptions of what happens now, not tests of what a purpose-built tool would do.

Here is weeklyAI's reading. A young person will often say more to something that cannot look disappointed, and a tool that answers warmly will keep them talking. That is the useful half. The other half is that warmth is not the same as judgment, and a machine that never pushes back can quietly become the place a teenager goes instead of a person. Expect your child to have told a chatbot something you have not heard. This reading would be wrong if teenagers turned out to hold back from chatbots the way they hold back from adults, or found the friendliness obviously hollow and not worth confiding in. What you can do with it: ask what they have told one, and treat the answer as a door opening rather than a confession.

A second thing to watch is where the approval lands. A tool that mirrors a teenager's words back with agreement can become the place they go to feel seen. This reading would be wrong if young people used these tools only for practical questions, with no sign of needing the tool's approval. Before your child starts with any mental health chatbot, look together at what it says it does with what is typed, and agree on what they will never share with it. Keep a human they can reach in the loop.

The limits are specific and worth holding onto. Everyone in the study was 18 to 30 and willing to take part in mental health research in Australia, so no one under 18 was heard and the most marginalised voices are missing19. The workshops were all remote, and internet problems sometimes affected the transcripts. Participants used working prototypes at a single time point, not a finished product over weeks, so nobody yet knows whether such a tool holds attention or helps in real family life20. The study reports what young people said they need. It cannot tell you that a chatbot is safe or effective for your child.

The money and interests are disclosed in the paper. The work was funded by a philanthropic donor who wishes to remain anonymous, with national research grants supporting two of the authors and a government scholarship supporting another. One senior author holds a 3.2 percent equity share in a company focused on digital transformation of mental health services, and another reports advisory and speaking roles with health insurers and pharmaceutical companies. The study was approved by the University of Sydney's human research ethics committee.

What follows for a household is not a product recommendation but a set of questions. Ask any mental health chatbot, or the service offering it, four things: what it does with what is typed, whether a clinician sees a summary and who writes that summary, what it does when someone types something alarming, and whether it is meant to sit beside human care or in place of it. Those are the terms the young people in this study set. A tool that cannot answer them plainly has not met the bar they described.

The chatbot still asks how the 20-year-old slept. What has changed is the list of things a parent can now ask before the answer matters: who reads it, who checks it, and who is there when the screen goes dark.

Where each piece of context comes from, and how much of it we read

  1. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "We examined the Mental health Intelligence Agent (Mia), a genAI chatbot originally designed for professionals in Australian youth services."
  2. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "Following co-design, 32 young people participated in online workshops exploring their perceptions of genAI chatbots in youth mental health and to develop recommendations for reconceptualising Mia for consumers and integrating it into services."
  3. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "Worldwide, young people aged 12-25 are experiencing high and, in most cases, rising rates of anxiety, depressive symptoms, psychological distress, self-harm, suicide, and other mood disorders"
  4. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "This aligns with the existing trend of young people turning to publicly accessible, general-purpose systems, such as OpenAI’s ChatGPT, for mental health support"
  5. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "It can’t outright replace actual psychologists and psychiatrists, because the human connection is still important at the end of the day, it’s not something that can be replicated by AI or chatbots"
  6. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "I feel a lot of the problems that I experience as a young person with lived experience are a lot more complex than maybe the advice that a system like this might be able to offer"
  7. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "A lot of people say, “Oh, you can’t trust AI 100%”. And I’m sure you can’t really trust anything 100% but just to show that it bases its answers on reasonable things, I think it’s important to show its thought process"
  8. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "Young people wanted genAI chatbots to interpret user inputs aptly, draw appropriate inferences from them, and provide reliable outputs grounded in evidence."
  9. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "Importantly, young people distinguished the guide role from companionship or friendship, emphasising that a genAI chatbot functioning as a friend would be inappropriate in mental health contexts where therapeutic boundaries and clinical relationships are paramount."
  10. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "Young people emphasised that for the sake of promoting unfiltered interactions and protecting privacy and autonomy, their raw data should remain private and under their control."
  11. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "While raw conversational data should remain private, AI-generated insights developed from genAI chatbot interactions (i.e. summaries detailing a user’s needs, personalised treatment, self-management, and service navigation recommendations, and deterioration or risk detection) should be accessible to clinicians and services."
  12. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "Young people expected these tools to detect when users discuss unsafe situations (e.g. sharing imminent plans of self-harm or violence) and when patterns suggest deteriorating mental health or elevated risk."
  13. Hinduja S, Patchin JW. (2026). Risks and Harms of Conversational Artificial Intelligence (CAI) Chatbot Use Among US Youth. Journal of Adolescence. 10.1002/jad.70164 — only the abstract - the full paper is behind a subscription — the passage: "Over 60% of the sample reported using a CAI chatbot, with 11.4% doing so every day or nearly every day."
  14. Hinduja S, Patchin JW. (2026). Risks and Harms of Conversational Artificial Intelligence (CAI) Chatbot Use Among US Youth. Journal of Adolescence. 10.1002/jad.70164 — only the abstract - the full paper is behind a subscription — the passage: "CAI chatbot usage is common among US adolescents, with 47.1% reporting exposure to one or more specific risks and harms."
  15. Hinduja S, Patchin JW. (2026). Risks and Harms of Conversational Artificial Intelligence (CAI) Chatbot Use Among US Youth. Journal of Adolescence. 10.1002/jad.70164 — only the abstract - the full paper is behind a subscription — the passage: "Still, 32.3% were asked for uncomfortable personal information, 23.1% felt manipulated or pressured, 17.1% received false information, 18.7% were encouraged to act unethically or illegally, 15.2% were prompted to risky behaviors, and 14.7% and 13.0% were exposed to self-harm and suicidal messages, respectively."
  16. Maheux AJ, Boeldt D, Akre-Bhide S, Flannery J, Paige A, Villella G, et al. (2026). How US Youth Use AI Chatbots: Conversation Patterns From Naturalistic Keystroke Observations. Journal of Medical Internet Research. 10.2196/95819 — only the abstract - the full text could not be fetched — the passage: "Participants were 3363 US youth aged 1-17 years using the Aura parental monitoring app, a commercially available service that guardians install on children's mobile devices for digital safety."
  17. Maheux AJ, Boeldt D, Akre-Bhide S, Flannery J, Paige A, Villella G, et al. (2026). How US Youth Use AI Chatbots: Conversation Patterns From Naturalistic Keystroke Observations. Journal of Medical Internet Research. 10.2196/95819 — only the abstract - the full text could not be fetched — the passage: "Results identified 11 mutually inclusive themes, including functional tool use; social practice; role-play, including general, romantic, and sexual; emotional support; friend-like interactions; and violence."
  18. Maheux AJ, Boeldt D, Akre-Bhide S, Flannery J, Paige A, Villella G, et al. (2026). How US Youth Use AI Chatbots: Conversation Patterns From Naturalistic Keystroke Observations. Journal of Medical Internet Research. 10.2196/95819 — only the abstract - the full text could not be fetched — the passage: "Most user-app-days involved using AI as a tool (26,343/42,355, 62%), but notable proportions included violence (6460/42,355, 15%), sexual role-play (6393/42,355, 15%), romantic role-play (3538/42,355, 8%), general role-play (3707/42,355, 9%), and friend-like interaction (2624/42,355, 6%)."
  19. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "the study participants consisted of young people aged 18-30 who were willing to participate in mental health services research in the Australian context"
  20. Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9 - the article this story is about — the whole article — the passage: "participants engaged with functioning prototypes at one time point rather than final products over an extended period of user testing, and thus, user experiences may differ from anticipated ones."

Poulsen, A., Hickie, I. B., Gorban, C. et al. (2026). Young people’s perceptions and recommendations for conversational generative artificial intelligence in youth mental health. npj Digital Medicine. https://doi.org/10.1038/s41746-026-02888-9

Who paid: The work was supported by an anonymous philanthropic donor affected by mental health, with additional support from National Health and Medical Research Council grants to two authors and an Australian Government Research Training Program scholarship; the article does not say whether funders had any say, and no equipment or software lenders are named.

Do not take this as professional medical advice.

The free weekly drip by email

One email a week: the free story of every section, with its picture, and the headlines of the rest.

No payment, nothing to cancel. Unsubscribe from any issue.

El goteo semanal gratis por correo

Un correo por semana: la nota gratuita de cada sección, con su imagen, y los titulares de las demás.

Sin pago y sin nada que cancelar. Puede darse de baja desde cualquier envío.

From the house

About

Every week the new research on AI is read here, and this page tells you what was found, how far it can be trusted, and what you can do with it. This page is free, and every line on it says what was found.

The studies are read by a machine built and ruled by a professor of communication at Sacramento State and Universidad Nacional Autónoma de México (UNAM). Who makes this →

Diego Bonilla, Ph.D.

Full Professor of Communication Studies, Sacramento State, California, United States
Graduate instructor, Universidad Nacional Autónoma de México (UNAM), Mexico

hypergraphia.com

I am a professor of communication at Sacramento State, where I created the course Information Management and Privacy and have taught it many times, and at UNAM, in Mexico City, I teach the graduate course Electronic and Generative Literature with AI. In both I teach AI. I have taught since 1989, at first mathematics and physics to children who found them hard, and since 1997 one subject that has changed its name many times: media literacy, computer literacy, digital literacy, information literacy, ICT literacy, and now AI literacy.

This page exists because most people cannot go back to a university, and these systems are changing their work, their children's schooling, their parents' care and their vote all the same. What a university teaches about them should reach the people who need it, every week, in plain language, without enrolling anywhere. That is what I do here, and it is what I do on Thursday mornings in a seminar for the Renaissance Society at Sacramento State, built on this publication, where anyone can send a question and hear it answered.

In 2003 my doctorate at Syracuse measured what a machine can infer about a person from the traces they leave, and what I found became the privacy course I still teach. In 2007 and 2009 I sat on the committee at ETS in Princeton that set the national standard for digital literacy, and from 2014 to 2016 I was the California State University's ICT Literacy Ambassador. Six names, one job: helping people not be at the mercy of the newest machine.

I have also worked on the other side of the counter. In 2000 I began making films that a machine assembles differently for every viewer; the first of them took first prize at the Moscow International Film Festival in 2004, two decades before the industry shipped a generative model. A generative video poem I made with a poet at UNAM is in the field's definitive anthology, and in 2025 my AI short film The Human Futures Market took Best AI Short Film at the Red Rocks AI Film Festival. My recent published research is on AI and elections, and on how these models represent the cultures they were not trained on. I do not read the papers myself, and this page will never tell you I do. I built the machine that reads them and wrote the rules it reads by, and I analyze what it gives me the way I have taught my students to analyze: asking how it knows, and who benefits from what it says.

The vendors began teaching AI literacy two years ago, because before that there was nothing to sell. I am not against them. I am simply not one of them, and no one who sells these systems pays for this page. Your subscription does, and that is what keeps it honest.

The record, dated
  • 1989 Teaching begins, Mexico City
  • 1995 Hypergraphia, an early Spanish-language web and hypermedia publisher
  • 1996 Mercadotecnia e imagen en Internet, with Jesús del Toro, a book on marketing on the Internet, among the first in Spanish, Editorial Iberoamérica
  • 1999 Autorretrato, first generative poem, published by UNAM's Instituto de Investigaciones Filológicas
  • 2003 Ph.D., Syracuse University, Doctoral Prize, for the dissertation The Medium Is the Measure of Itself
  • 2004 A Space of Time: First Prize, Internet/Multimedia, XXVI Moscow International Film Festival
  • 2007, 2009 ETS iSkills committee, Princeton
  • 2008 Making Sense of Tracking Data, a book on how the traces a person leaves online allow inferences about them, including how they think, VDM Verlag
  • 2011 Keynote on generative work, UNAM
  • 2014–2016 ICT Literacy Ambassador, California State University · California Open Educational Resources Council
  • 2019 Big Data, with Rodolfo Mata: Electronic Literature Collection, Volume 4
  • 2020 Tablada Hipertextual, with Rodolfo Mata, UNAM: the computer reads a poet's words and builds a different reading from them, the ground the language models now work on
  • 2021 Uku Pacha, a generative VR film: world premiere, Vancouver International Film Festival
  • 2023 Keynote on the representation of Latin American cultures in AI, Ibero-American Meeting of Audiovisual Narratives, Trujillo, Peru · Graduate teaching at UNAM begins
  • 2023–2026 Directed the faculty learning community Advancing AI Literacy, Sacramento State · Principles for Cognitive and Pedagogical Sovereignty
  • 2024 A weekly AI publication run on real pipelines, every week since
  • 2025 Best AI Short Film, Red Rocks AI Film Festival
  • 2026 Keynote, with Nancy Huante-Tzintzun, Latin American and Caribbean Digital Humanities symposium, UT San Antonio · AI Watch, a weekly seminar for the Renaissance Society at Sacramento State

The research record →

TertulIA: weeklyAI.watch/tertulia/en/Pictures are illustrations generated by weeklyAI.watch, never photographs