Tiếp theoHướng dẫn tiếp theo
Đạo đức AI dành cho cố vấn và nhà trị liệu
xã hội
HƯỚNG DẪN xã hội
AI mental health chatbots can deliver structured techniques and support at any hour, and some trials show short-term symptom improvement.
However, they cannot match a licensed human therapist's accountability, clinical judgment or reliable handling of crises. The research supports chatbots as a supplement or bridge, not a replacement. The biggest documented risks are mishandled suicidal disclosures, harmful advice and over-agreeable responses.
Early mental health chatbots such as Woebot and Wysa were largely scripted. They guided users through cognitive behavioral therapy (CBT) exercises using pre-written responses, which made them predictable but limited. Newer tools use large language models, which respond more flexibly but less predictably. Many people also use general-purpose chatbots for emotional support, even though these were never designed or tested as therapy. The research is mixed. Randomized trials and meta-analyses of chatbot interventions generally report small-to-moderate short-term reductions in depression and anxiety symptoms. Follow-up periods are often short, many users drop out and comparison groups vary. In 2025, Dartmouth researchers published a randomized trial of Therabot, a generative AI chatbot built for therapy, in NEJM AI. It reported symptom reductions compared with a waitlist, which is a weaker comparison than active treatment by a human. The therapeutic alliance, meaning the bond, agreement on goals and agreement on tasks between client and therapist, is one of the most consistent predictors of outcome in human therapy. Some chatbot studies report that users rate their sense of bond with the bot fairly high. Whether that feeling works the way a human alliance does, including repairing ruptures and challenging the client, is not established. The risks are well documented. In 2023 the National Eating Disorders Association took down its Tessa chatbot after it gave weight-loss advice. Research testing chatbots with crisis scenarios has found missed or inappropriate responses to suicidal statements. Lawsuits have alleged that companion and general chatbots contributed to teenagers' suicides. Chatbots also tend to agree with users, which can reinforce distorted beliefs. Human therapists are licensed and accountable, and they can assess risk, arrange emergency care and coordinate treatment. A common misconception is that it has to be one or the other. The most defensible use is chatbots alongside human care, not instead of it. Some states, including Illinois in 2025, have passed laws restricting AI from providing therapy.
Những tác hại thảm khốc và thường ngày của AI đều phụ thuộc vào việc ai hiểu được rủi ro và ai có thể hành động.
Kiến thức công cộng và chuyên môn định hình liệu chính sách an toàn mạnh mẽ có khả thi về mặt chính trị hay không.
Những lời giải thích rõ ràng làm giảm sự thu hút bởi sự cường điệu, PR trong phòng thí nghiệm và sân khấu đạo đức mơ hồ.
Expect more controlled trials comparing chatbots with active treatments rather than waitlists, and more states setting rules on AI claims of providing therapy. Designs that blend human and AI care, such as chatbot homework reviewed by clinicians or AI triage that escalates to people, have the strongest case. Better crisis detection and less sycophancy are active engineering goals, but they are not solved. Whether chatbots can safely widen access to care for people who cannot reach therapists depends on independent evidence that has not yet caught up with how widely these tools are used.
A college student on a waiting list uses a CBT-based chatbot to practice spotting and reframing negative thoughts between intake and her first therapy appointment.
A therapist assigns a chatbot exercise for tracking sleep and mood as homework, and reviews the logs with the client in the next session.
A person tells a general-purpose chatbot he has been thinking about ending his life. A well-designed system stops the ordinary conversation and gives crisis resources such as the 988 Suicide and Crisis Lifeline in the US, while a poorly designed one keeps chatting as usual.
A parent notices a teenager spending hours each night confiding in a companion chatbot app. The family uses it as a reason to find a human counselor rather than treating the app as therapy.
Xử lý rủi ro hiện hữu như khoa học viễn tưởng trong khi khả năng lại phức tạp.
Nhầm lẫn giữa an toàn sản phẩm bề mặt với sự liên kết dưới quyền tự chủ cao.
Chỉ để lại những khán giả không phải người Anh và không có chuyên môn với những nguồn chất lượng thấp.
Tách biệt các tác hại của sản phẩm, sử dụng sai và rủi ro mất kiểm soát/sai lệch.
Hỏi bằng chứng nào sẽ thay đổi quan điểm của bạn về thời gian và mức độ nghiêm trọng.
Ưu tiên các nguồn chính và đánh giá cụ thể hơn các tuyên bố tiếp thị.
Xác định một lộ trình hành động: sự nghiệp, chính sách, nguồn tài trợ hoặc kỹ năng - không chỉ là nhận thức.
Free newsletter
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
AI mental health chatbots can deliver structured techniques and support at any hour, and some trials show short-term symptom improvement. However, they cannot match a licensed human therapist's accountability, clinical judgment or reliable handling of crises. The research supports chatbots as a supplement or bridge, not a replacement. The biggest documented risks are mishandled suicidal disclosures, harmful advice and over-agreeable responses.
Beating a waitlist shows the chatbot does better than nothing. It does not show the chatbot matches active human treatment.
Tessa gave weight-loss advice, which is harmful for people with eating disorders, so it was removed.
The alliance is the bond plus agreement on goals and tasks, and it is one of the most consistent predictors of outcome in therapy.
Pre-written responses avoid made-up content but are rigid. Language models are flexible but less predictable, which makes safety testing harder.
Models trained to please users tend to agree with them, while good therapy sometimes requires respectfully challenging the client.
Tiếp tục học hỏi
Đã chọn thêm hướng dẫn cho chủ đề này
Tiếp theoHướng dẫn tiếp theo
Đạo đức AI dành cho cố vấn và nhà trị liệu
xã hội