A Mental-Health Plan for 42,000 Chinese Students, With the AI Trials Still Unstarted

US · CanadaAmérica Latina

Weekly AI Watch

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

Week of 30 October 2026EN | ESPast editions15,230 papers found · 14,455 read · 4,301 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 30 de octubre de 2026EN | ESEdiciones anteriores15,230 estudios encontrados · 14,455 leídos · 4,301 correspondían · 46 notas pasaron todas las comprobaciones

weeklyAI · Week of 30 October 2026weeklyAI · Semana del 30 de octubre de 2026

← Your rights← Sus derechos

other · Frontiers in psychology · la publicación, 14 oct 2026 · gratis

Unos 42,000 estudiantes chinos, cinco años y un plan de IA que todavía no ha mostrado nada

El proyecto prevé ocho terapias y un sistema que lee caras, voces y escáneres cerebrales, pero ningún ensayo ha empezado y no hay resultados.

Versión breve · la versión detallada sigue, unos 6 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: Estudiantes universitarios de 18 a 30 años en China
Cuántos: unos 42,000 estudiantes
Dónde: Siete universidades de China
Cuándo: Inscripción iniciada en septiembre de 2025; protocolo maestro versión 2.0 fechado el 5 de septiembre de 2026
Tipo de estudio: experiment
Quién lo hizo: Universidad Médica de Henan, China
El límite que importa: Los ocho ensayos siguen en planeación y no han reclutado ni asignado a nadie.

Un sistema que dice no decidir nada sigue necesitando que alguien decida por él.
Lectura de weeklyAI

Ver · 0:31

En abril de 2025, un comité de ética de la Universidad Médica de Henan, en China, aprobó una evaluación de salud mental para estudiantes. En septiembre de ese año empezó a inscribir alumnos en esa universidad. Así arranca el registro del programa CAMPUS.

El plan creció. Ahora abarca siete universidades del centro, el noreste, el noroeste, el este y el sur de China. La meta es seguir a unos 42,000 estudiantes, de 18 a 30 años, una vez por semestre durante cinco años.

Dentro de ese grupo, hasta 14,000 personas pasarían por una batería más pesada. A todo el grupo se le grabaría la voz leyendo un texto y el rostro mientras responden preguntas abiertas, en cada ronda del estudio. Las demás mediciones serían solo para ese subgrupo más pequeño. A esas personas del subgrupo se les mediría el movimiento de los ojos, la actividad eléctrica del cerebro, la respuesta del corazón y la oxigenación de la sangre en la corteza. También se les haría resonancia magnética.

Hay una capa más, de sangre, orina, heces y saliva, para análisis genéticos y moleculares. Todavía no está activada. No se tomará ninguna muestra hasta que existan procedimientos escritos para cada tipo de muestra y las aprobaciones correspondientes.

Las ocho terapias siguen en la etapa de planeación. Son terapia cognitivo-conductual, terapia interpersonal y de ritmos sociales, ejercicio, música, biofeedback, realidad virtual, arte expresivo y una intervención de cinco sentidos. Ninguna ha comenzado a reclutar ni a asignar participantes al azar. En cada ensayo se comparará contra un grupo de control que aún no se define.

El sistema de inteligencia artificial se entrena para acertar dónde cae una persona en cuestionarios de depresión, ansiedad, insomnio e ideas suicidas. No trabaja con diagnósticos clínicos. Su precisión se probará solo dentro de estas mismas siete universidades, con datos de los mismos estudiantes. Un buen resultado ahí no diría si el sistema acierta con usted ni si sería legal en su país. Las universidades se eligieron por colaboraciones de investigación, y la muestra se describe como mayoritariamente de etnia han.

Los propios autores lo dicen: la IA no decidirá diagnósticos, ni el nivel de riesgo de suicidio, ni quién entra a un ensayo, ni qué tratamiento recibe alguien. Cualquier respuesta positiva en el cuestionario de ideas suicidas genera una alerta electrónica inmediata para el personal capacitado. Cada alerta recibe una revisión humana pronta, que decide si hace falta un contacto urgente ese mismo día.

Lo que falta ver es el protocolo de cada ensayo, su registro público y el criterio de comparación. Antes de creer que un sistema así funciona, pregunte qué tan bien detecta algo que un cuestionario no detecte ya.

Qué significa para usted

Ese plan no le dice nada todavía sobre su propia cara ni su propia voz. Los ocho ensayos siguen sin empezar y no hay ningún resultado, así que no puede concluir que la inteligencia artificial detecte mejor la depresión que un cuestionario. Cuando se publique el registro de cada ensayo, fíjese con qué grupo se compara y qué tanto detecta algo que el cuestionario no vea ya.

Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887

Quién pagó: El trabajo fue financiado por organismos públicos chinos, principalmente la Fundación Nacional de Ciencias Naturales de China y programas provinciales de Henan; el artículo afirma que los financiadores no tuvieron papel alguno en el diseño, la recogida de datos, el análisis, la interpretación ni la preparación del manuscrito, y no menciona 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 · 1153 palabras · unos 6 minLeerla →Cerrar

Un programa chino planea encuestar a 42,000 estudiantes sobre su salud mental.

El protocolo describe cómo se recogerán los datos y qué no decidirá la máquina. No informa de ningún resultado, porque todavía no hay ninguno.

Responde un cuestionario sobre su ánimo, su sueño y sus preocupaciones. Todo eso alimentaría un sistema de inteligencia artificial que busca clasificar el riesgo según los puntajes de los cuestionarios, no diagnosticar. Ese es el plan que describe un protocolo publicado en *Frontiers in Psychology* por un equipo de la Universidad Médica de Henan, en China.

El programa se llama CAMPUS y tiene tres partes: un seguimiento de cinco años a unos 42,000 estudiantes de siete universidades chinas, ocho ensayos clínicos aleatorizados que todavía no arrancan, y análisis exploratorios de IA y de moléculas. Los cuestionarios miden depresión, ansiedad, insomnio, estrés y pensamientos suicidas¹. A eso se suman electroencefalogramas, resonancias magnéticas, ritmo cardiaco, seguimiento ocular, análisis de expresiones faciales y acústica de la voz¹. La IA que planean es un modelo que combina todos esos datos y luego explica cuáles pesaron más en su cálculo².

Piense en lo que abarca una sola visita. Los estudiantes empiezan respondiendo cuestionarios en una aplicación llamada AI Muxin; luego pasan a las mediciones fisiológicas en sitios designados. Los cuestionarios se repiten cada semestre durante cinco años, y las mediciones corporales, una vez al año¹. La parte de las muestras biológicas todavía no se activa: sangre, orina, heces y saliva son candidatas, pero el calendario y los procedimientos se cerrarán antes³.

Lo curioso es que el propio protocolo pone límites muy claros a su máquina. El sistema no va a diagnosticar nada: las tareas son contra puntajes de cuestionario, y no se planea ninguna entrevista diagnóstica estructurada ni una revisión clínica independiente⁴. Los modelos se desarrollan y evalúan fuera de línea, y sus resultados no van a determinar diagnóstico, disposición de riesgo suicida, elegibilidad para un ensayo, asignación al azar ni elección de tratamiento⁵. Tampoco van a sustituir ni retrasar la evaluación profesional ni la acción de emergencia⁶. Y ningún puntaje del modelo va a inscribir a alguien en una intervención por su cuenta⁷.

Esa última parte importa porque hay una alerta de seguridad de por medio. Cada vez que un estudiante marque un reactivo de ideación suicida, salta un aviso electrónico a personal capacitado, que revisa el caso y decide si hace falta contacto el mismo día. El nivel de riesgo lo determina un profesional de salud mental, no el algoritmo. El protocolo describe tres niveles de respuesta ante una alerta de este tipo. El protocolo describe tres niveles de respuesta, desde un contacto en 24 horas hasta el traslado a urgencias cuando hay peligro inminente.

Sobre los datos personales, el protocolo separa consentimientos. Grabar voz, grabar la cara, hacer resonancias y guardar muestras para genética requiere un permiso específico para cada componente⁸. Un estudiante puede negarse a cualquiera de esas partes sin perder su lugar en el resto del estudio⁹. Las grabaciones crudas, los videos faciales, las imágenes y los datos genómicos no se depositarán en repositorios públicos¹⁰. Incluso los rasgos derivados y los pesos del modelo pasarán por una revisión de riesgo antes de compartirse, porque pueden conservar información de quienes participaron¹¹.

¿Y qué se ha probado antes? Los propios autores citan programas escolares como uno canadiense que reportó reducciones en suicidabilidad, depresión y ansiedad, aunque se apoyaba sobre todo en cuestionarios y no incluía neuroimagen ni análisis moleculares. También citan StudentLife, que mostró que se puede usar el teléfono para observar patrones de sueño y conducta en universitarios, pero sin muestras biológicas ni resonancias. La diferencia que CAMPUS promete es juntar todo eso en un solo sistema. Prometer no es lo mismo que demostrar.

Así lo leemos nosotros. Cuando alguien sabe que lo graban y lo observan, deja de comportarse con naturalidad y empieza a cuidarse. Si además su cara, su voz y su sangre se convierten en números que nadie le explica, es fácil que acepte sin preguntar. Esperamos ver a muchos estudiantes firmando cada permiso por costumbre, y a pocos pidiendo el detalle por escrito. Sabríamos que nos equivocamos si los estudiantes rechazan componentes concretos con confianza, piden por escrito para qué se usará cada dato y el estudio responde con claridad, sin presión, y sin cambiarles el trato.

También esperamos que el comité de ética y los procedimientos de seguridad existan en el papel, pero que el estudiante promedio no sepa a quién dirigirse ni a qué hora hay alguien disponible. Lo contrario se vería si el protocolo público nombra a un responsable en cada universidad, con horario y vía de contacto, y si quienes marcaron una alerta cuentan que los contactaron rápido. Y aunque el estudio diga que su IA no decide nada, conviene vigilar si algún puntaje sale del equipo de investigación y llega a una autoridad universitaria, a un profesor o a una aseguradora. Pregunte por escrito si eso puede pasar y bajo qué reglas.

Hay límites que el propio documento reconoce. Las siete universidades se eligieron por colaboraciones de investigación, no por sorteo nacional, así que el grupo no representa al estudiantado chino¹². La muestra es mayoritariamente de etnia han, lo que reduce la generalización. El seguimiento de cinco años sufre desgaste. Quienes acepten resonancias y muestras pueden ser distintos de quienes no, y eso sesga la parte biológica. Los aparatos varían por centro. Se excluye a estudiantes con diagnóstico previo de depresión mayor o ansiedad generalizada, lo que deja fuera parte del espectro. Y el sesgo algorítmico es un riesgo reconocido, con desempeño distinto según sexo, avance académico o centro¹³.

Nada de esto es un resultado. Los ocho ensayos siguen en planeación y no han reclutado, aleatorizado ni entregado una sola sesión¹⁴. El reclutamiento del seguimiento comenzó en la Universidad Médica de Henan en septiembre de 2025 y se planea ampliar a las siete universidades¹⁵. El documento es la versión 2.0 del protocolo maestro, fechada el 5 de septiembre de 2026, y no reporta hallazgos de ningún participante. Es decir: no hay evidencia de que esta IA funcione, ni de que falle.

Aquí va lo que usted puede hacer con esto. Si alguna institución, escuela o clínica le ofrece una herramienta de IA para evaluar salud mental, pregunte tres cosas antes de firmar: qué se graba exactamente, quién lo guarda y por cuánto tiempo, y si puede retirar solo una parte sin perder el resto. Guarde copia de lo que firmó y de lo que le contestaron. Pida el nombre y el contacto del comité de ética que aprobó el estudio y del responsable de seguridad. Si le muestran un puntaje sobre usted, pida que le expliquen qué datos lo produjeron y quién lo interpretó. Un sistema que dice no decidir nada sigue necesitando que alguien decida por él.

Lo que sí puede hacer es preguntar, antes de entrar, quién verá esa grabación y qué pasará si dice que no.

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

Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Multidimensional screening integrates psychological self-report (PHQ-9, GAD-7, ISI, BSI-CV, and others), physiological measurements (EEG, MRI, fNIRS, HRV, eye tracking, facial-expression analysis, ERP, voice acoustics), and an exploratory biospecimen component for multi-omics profiling"
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A Transformer-based multimodal AI framework with SHAP-based feature attribution will be developed and internally evaluated for questionnaire-defined risk stratification and exploratory treatment-response prediction."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The biospecimen component has not yet been activated. Blood, urine, stool, and saliva are candidate biological matrices, but the collection schedule and matrix-specific procedures will be finalized before activation."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The AI analyses will address questionnaire-defined screening and prediction tasks rather than clinical diagnosis. No structured diagnostic interview or independent clinical adjudication is planned."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "All models will be developed and evaluated offline. Their outputs will not determine clinical diagnosis, suicide-risk disposition, trial eligibility, randomized allocation, treatment selection, intervention intensity, or protocol modification."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "AI models will be developed and evaluated offline; their outputs will not be used for real-time safety management, clinical risk assignment, or disposition and will not replace or delay professional assessment or emergency action."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Cohort screening or AI-generated model outputs will not automatically enroll a participant in an intervention or determine trial selection or randomized allocation."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Voice recording, facial-expression video recording, physiological and neuro-functional assessments, MRI, and future biospecimen collection for genetic and other multi-omics analyses will be subject to separate component-specific consent or consent options, as approved by the responsible ethics committee."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Participants may decline or discontinue any optional component without affecting their continued participation in the core cohort."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Raw voice recordings, facial-expression videos, imaging data, genomic data, and other data carrying a material risk of re-identification will not be deposited in public repositories."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Derived features, trained model weights, embeddings, and other model artifacts will undergo disclosure-risk review before sharing because they may retain information about the training data."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The participating universities were selected through research collaborations rather than probability sampling from a national sampling frame. Despite their multiregional distribution, the cohort is not nationally representative"
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "algorithmic bias is a recognized risk, including differential model performance across subgroups defined by sex, academic stage, or center."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "All eight embedded intervention trials remain in the planning stage and have not commenced participant recruitment, enrolment, randomization, or intervention delivery."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Longitudinal cohort enrolment commenced at Henan Medical University in September 2025 and is planned to expand across seven participating universities."

No tome esto como consejo médico profesional.

Los hallazgos de otros estudios que aquí se mencionan los conocemos por este documento, que fue el que leímos; no abrimos cada uno de esos estudios.

other · Frontiers in psychology · the paper, 14 Oct 2026 · free

A Mental-Health Plan for 42,000 Chinese Students, With the AI Trials Still Unstarted

Seven universities plan to record faces, voices, brain scans and blood for an AI screening system. No results exist yet, and the system is matched against questionnaires, not diagnoses.

Short version · the longer version follows, about 7 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: University students in China
How many: about 42,000 students
Where: Seven Chinese universities
When: Enrolment began September 2025
Kind of study: analysis of what people did
Who did it: Mental Health and AI Research Center, Henan Medical University, with US co-authors
The limit that matters: Universities were picked through research collaborations, so the group is not representative of Chinese students.

The AI is trained to match scores on standard questionnaires, such as a depression scale with a cut-off of 10.
weeklyAI's reading

Watch · 0:34

In April 2025, an ethics committee at Henan Medical University approved an assessment of student mental health. Enrolment there began in September 2025. That approval covered one campus. The plan has since grown.

It now spans seven Chinese universities, in the north-east, north-west, east, south and central regions of the country. The researchers aim to enrol about 42,000 students aged 18 to 30, and to follow them once a semester for five years. Up to 14,000 of them would join a smaller group whose faces, voices, eye movements, brain activity and heart rhythms are recorded.

A further layer would take blood, urine, stool and saliva for genetic and other molecular analysis. That part is not switched on. No sample is taken until written procedures and separate approvals exist, the authors say.

Eight therapy trials sit inside the same programme: cognitive behavioural therapy, a sleep-and-routine therapy, physical exercise, music, biofeedback, virtual reality, expressive art, and a five-sensory stress programme. All eight are still in planning. Each will compare its therapy with a control condition that has not yet been fixed. Each is meant to recruit 420 students.

The AI is trained to match scores on standard questionnaires, such as a depression scale with a cut-off of 10. It is not checked against a doctor's diagnosis, because no diagnostic interviews are planned. Its accuracy is tested only inside these seven universities, by holding back some of their own students and some of their own campuses. A result from that test would not tell anyone whether such a system works, or is fair, outside them.

Two things about the students limit what any finding could mean. The universities were chosen through existing research collaborations, not drawn at random from across China. The authors describe the sample as predominantly Han Chinese. The authors say the group is not representative of Chinese students as a whole. The authors describe the sample as mostly Han Chinese, and they say this may limit how well the findings apply to other universities and other groups of students.

The authors are direct about the machine's role. Its outputs will not set a diagnosis, a suicide-risk level, entry into a trial or a choice of treatment. Any positive answer on the suicide questionnaire sends an alert to trained staff and the site safety officer, who review it quickly. A qualified professional then decides what happens next. When the answers show active thoughts of suicide with intent, a workable plan, steps taken toward it, or a risk that the student cannot stay safe, a qualified professional sees the student the same day, and staff stay in contact until a plan is set. When the answers are judged not urgent, staff contact the student within 24 hours and a professional completes a full safety assessment within 48 hours.

Watch for the separate protocol and public registration of each trial before it starts. The open question is whether this system detects distress any better than the questionnaires students already fill in.

What this means for you

For now there is nothing to act on, because no student has been screened by this system and none of the eight trials has begun. If a university near you announces such a programme, the question to ask is what its own trial protocol and public registration say, and whether the machine was ever checked against a doctor's assessment. Until then, treat any claim that this detects distress better than a questionnaire as unproven.

Who paid: The work was funded by Chinese public bodies, chiefly the National Natural Science Foundation of China and Henan provincial programs; the article states the funders had no role in study design, data collection, analysis, interpretation or manuscript preparation, and names no equipment or software lenders.

Do not take this as professional medical advice.

The longer versionThe passages copied from the paper, the pictures, and every source with how much of it we read · 1437 words · about 7 minRead it →Close

China Plans to Read Students’ Faces, Voices and Brain Scans for an AI Mental-Health Program. No Results Yet.

Seven universities, about 42,000 students, five years of questionnaires, and eight therapy trials that have not started. The researchers say the AI will never decide anyone’s care.

Picture the first visit. A student sits down, answers a long set of questionnaires on a phone app, and then moves to a lab room where a cap of electrodes records the brain’s electrical activity, a scanner maps it, a camera watches the face, a microphone takes the voice, and an eye tracker follows the gaze. Later, blood, urine, stool or saliva may be collected. All of it is meant to feed one system.

That system is CAMPUS, the Comprehensive Artificial Intelligence-Assisted Mental Health Program for University Students in China. A protocol published in Frontiers in Psychology describes a five-year study across seven Chinese universities in seven provinces, from Henan in the center to Hainan in the south. The authors are a large team from institutions across China and the United States; the Mental Health and Artificial Intelligence Research Center at the Second Affiliated Hospital of Henan Medical University is named as the body responsible for project and data management, working with site investigators at each university. Funding came from the National Natural Science Foundation of China and Henan provincial programs; the authors declared that the work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest, and no independent outside evaluator is named.

Start with the size, because the size is the point. The cohort aims to enroll about 42,000 students aged 18 to 30, roughly 6,000 from each university, and to check in on them once a semester for five years, nine follow-up waves in all. The authors say the reason is capacity: depression, anxiety, insomnia and suicidal thoughts among university students are rising faster than campus counseling services can absorb¹. Whether that is true where you live is a separate question, and this study does not answer it.

The technology is not one device but a pile of them. The screening combines self-report questionnaires with physiological measurements — EEG, MRI, fNIRS, heart-rate variability, eye tracking, facial-expression analysis, event-related potentials and voice acoustics — plus an exploratory biospecimen component for multi-omics profiling². On top sits a Transformer-based multimodal AI framework, with SHAP-based feature attribution, to be developed and internally evaluated for questionnaire-defined risk stratification and exploratory treatment-response prediction³. In plain words: software that learns patterns across many kinds of data at once, plus a method for saying which input pushed a given prediction.

Here is the sentence that matters most for anyone worried about a machine judging them. The AI analyses, the authors write, will address questionnaire-defined screening and prediction tasks rather than clinical diagnosis, and no structured diagnostic interview or independent clinical adjudication is planned⁴. All models will be developed and evaluated offline, and their outputs will not determine clinical diagnosis, suicide-risk disposition, trial eligibility, randomized allocation, treatment selection, intervention intensity, or protocol modification⁵. The models will not be used for real-time safety management, clinical risk assignment or disposition, and will not replace or delay professional assessment or emergency action⁶. Screening or AI output will not automatically enroll anyone in an intervention or decide which trial they enter⁷.

Consider what the clinic visit covers, because the protocol is unusually explicit about consent. Voice recording, facial-expression video, physiological and neuro-functional assessments, MRI, and future biospecimen collection for genetic and other multi-omics analyses will each be subject to separate component-specific consent or consent options, as approved by the responsible ethics committee⁸. A participant may decline or discontinue any optional component without affecting continued participation in the core cohort⁹. Raw voice recordings, facial videos, imaging data, genomic data and other data carrying a material risk of re-identification will not be deposited in public repositories¹⁰. Derived features, trained model weights and embeddings will undergo disclosure-risk review before sharing, because they may retain information about the training data¹¹.

The biological half is still on paper. The biospecimen component has not been activated; blood, urine, stool and saliva are candidate matrices, and the collection schedule and matrix-specific procedures will be finalized before activation¹². The eight embedded trials are equally unstarted: all remain in the planning stage and have not begun recruitment, enrolment, randomization or intervention delivery¹³. The planned interventions include cognitive behavioral therapy, interpersonal and social rhythm therapy, physical exercise, personalized music, biofeedback, virtual reality, expressive art and a five-sensory stress-regulation program, each in its own two-arm randomized trial with its own comparator, to be finalized before activation.

Enrolment itself has begun at one site. Longitudinal cohort enrolment commenced at Henan Medical University in September 2025 and is planned to expand across the seven participating universities¹⁴. The master protocol is Version 2.0, dated 5 September 2026; the authors state that before Version 1.0 they had not viewed any post-enrolment outcome data. No participant-level or aggregate findings are reported.

The authors list the limits themselves, and they are worth reading closely. The universities were selected through research collaborations rather than probability sampling from a national frame, so the cohort is not nationally representative¹⁵. Algorithmic bias is a recognized risk, including differential model performance across subgroups defined by sex, academic stage or center¹⁶. The authors also note that the predominantly Han Chinese sample may further limit generalizability; that five years of follow-up will lose participants; that participation in MRI and biospecimen collection may differ systematically by participant characteristics, so the biological subsamples may be subject to selection bias; that devices differ by center; that lab assessments may not capture real-world stress; and that students with prior diagnosed major depressive disorder or generalized anxiety disorder are excluded, limiting study of the full severity range.

This belongs to a family of stories about institutions writing rules for AI before they can enforce them. A study of Chicago Public Schools, which we could read only in its abstract, traced how the district articulated flexible guidance, human oversight, vendor accountability and prohibitions on high-risk uses between 2023 and 2025, alongside a gap between commitments and enforcement: monitoring and differentiated supports were less documented¹⁷. Another abstract-only study of four UNESCO and OECD publications found that words like responsibility, transparency and trust provide normative coherence but remain operationally underdefined across institutional contexts¹⁸. A third, also abstract-only, looked at documents from all 48 accredited health-informatics master’s programs in the United States and found that 40 had at least one qualifying document, most offering guidance rather than formal policy, with privacy and regulatory concepts present but less common than academic-integrity themes¹⁹²⁰. The pattern across all three: the paper exists, the enforcement is thinner.

None of this tells you whether such a system is accurate, useful or lawful where you live. The protocol plans internal validation only — holding out participants, later waves, or one university at a time — and the authors say plainly that these procedures do not establish transportability, clinical utility or implementation feasibility in independent settings. The reference outcomes are questionnaire scores, not diagnoses⁴. The universities were chosen through collaborations, not sampled from a national frame¹⁵. So even a strong result later would say nothing about students in Mexico City, Lagos or Ohio, and nothing about what your own data protection authority would permit. If a school, employer or insurer near you ever offers an AI mental-health screen, the questions to ask are the ones this protocol answers about itself: who sees the raw face and voice recordings, can you decline one part and keep the rest⁹, who reviews an alarm before anything happens, and is a human clinician — not a model — making the call⁶.

What the reader now knows about that first visit: the questionnaires are real, the electrodes and cameras are planned, the blood draw is not yet scheduled¹², the eight therapy trials have not enrolled a single participant¹³, and the AI, by the authors’ own design, sits offline and answers no clinical question⁵. The study is a blueprint for evidence, not evidence. Whether a system that watches faces and voices can be trusted with a student’s worst semester is exactly what the next five years are supposed to test — and the protocol’s most useful contribution may be that it says, in advance and in writing, what the machine will not be allowed to decide.

If a university near you announces an AI mental-health program, ask for its protocol and read the consent section first: which parts are optional, who holds the recordings, and who — a person or a program — decides what happens next.

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

Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "University students face a rising prevalence of depression, anxiety, insomnia, and suicidal ideation that exceeds the capacity of conventional, resource-constrained campus support services."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "Multidimensional screening integrates psychological self-report (PHQ-9, GAD-7, ISI, BSI-CV, and others), physiological measurements (EEG, MRI, fNIRS, HRV, eye tracking, facial-expression analysis, ERP, voice acoustics), and an exploratory biospecimen component for multi-omics profiling"
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "A Transformer-based multimodal AI framework with SHAP-based feature attribution will be developed and internally evaluated for questionnaire-defined risk stratification and exploratory treatment-response prediction."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "The AI analyses will address questionnaire-defined screening and prediction tasks rather than clinical diagnosis. No structured diagnostic interview or independent clinical adjudication is planned."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "All models will be developed and evaluated offline. Their outputs will not determine clinical diagnosis, suicide-risk disposition, trial eligibility, randomized allocation, treatment selection, intervention intensity, or protocol modification."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "AI models will be developed and evaluated offline; their outputs will not be used for real-time safety management, clinical risk assignment, or disposition and will not replace or delay professional assessment or emergency action."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "Cohort screening or AI-generated model outputs will not automatically enroll a participant in an intervention or determine trial selection or randomized allocation."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "Voice recording, facial-expression video recording, physiological and neuro-functional assessments, MRI, and future biospecimen collection for genetic and other multi-omics analyses will be subject to separate component-specific consent or consent options, as approved by the responsible ethics committee."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "Participants may decline or discontinue any optional component without affecting their continued participation in the core cohort."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "Raw voice recordings, facial-expression videos, imaging data, genomic data, and other data carrying a material risk of re-identification will not be deposited in public repositories."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "Derived features, trained model weights, embeddings, and other model artifacts will undergo disclosure-risk review before sharing because they may retain information about the training data."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "The biospecimen component has not yet been activated. Blood, urine, stool, and saliva are candidate biological matrices, but the collection schedule and matrix-specific procedures will be finalized before activation."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "All eight embedded intervention trials remain in the planning stage and have not commenced participant recruitment, enrolment, randomization, or intervention delivery."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "Longitudinal cohort enrolment commenced at Henan Medical University in September 2025 and is planned to expand across seven participating universities."
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "The participating universities were selected through research collaborations rather than probability sampling from a national sampling frame. Despite their multiregional distribution, the cohort is not nationally representative"
Yang, G., Liu, R., Wang, C. et al. (2026). Comprehensive artificial intelligence-assisted mental health program for university students in China (CAMPUS): study protocol for a multi-center prospective cohort with planned embedded randomized intervention trials. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2026.1877887 - the article this story is about — the whole article — the passage: "algorithmic bias is a recognized risk, including differential model performance across subgroups defined by sex, academic stage, or center."
Wu E, Rodriguez L. (2026). From Articulation to Enforcement: District Governance of Generative AI Under Policy Ambiguity in Chicago Public Schools. Educational Policy. https://doi.org/10.1177/08959048261500084 — only the abstract - the full text could not be fetched — the passage: "CPS articulated flexible guidance, human oversight, vendor accountability, and prohibitions on high-risk uses, alongside an articulation-enforcement gap: commitments are detailed, but monitoring, enforcement, and differentiated supports are less documented."
Hashem R. (2026). Cheating or being cheated? Academic integrity and a changing cheating habitus in international AI governance discourse. Frontiers in Education. https://doi.org/10.3389/feduc.2026.1953305 — only the abstract - the full text could not be fetched — the passage: "They provide normative coherence but remain operationally underdefined across institutional contexts."
Eldredge C, Dolezel D, Jones J, Vooppala N, Bronsburg SE. (2026). Institutional approaches to artificial intelligence policy and guidance in health informatics and information management education: emerging trends and inconsistencies. Journal of the American Medical Informatics Association. 10.1093/jamia/ocag150 — only the abstract - the full text could not be fetched — the passage: "Forty programs (83%) had at least 1 qualifying document, 8 programs had no qualifying documents."
Eldredge C, Dolezel D, Jones J, Vooppala N, Bronsburg SE. (2026). Institutional approaches to artificial intelligence policy and guidance in health informatics and information management education: emerging trends and inconsistencies. Journal of the American Medical Informatics Association. 10.1093/jamia/ocag150 — only the abstract - the full text could not be fetched — the passage: "References to privacy, intellectual property, and regulatory concepts were present but less common than academic integrity-related themes."

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.