Could AI chatbots replace human therapists? - What in the World podcast, BBC World Service

Key Concepts:

• AI Therapy Chatbots: Generic (e.g., ChatGPT, DeepSeek) vs. Dedicated Mental Health Apps.
• Cognitive Behavioral Therapy (CBT): A structured therapy approach focused on reframing thoughts.
• Large Language Models (LLMs): AI models trained on vast datasets to generate responses.
• Data Privacy and Security: Concerns about the use and potential misuse of personal mental health data.
• Accessibility and Affordability: AI therapy as a potential solution to the shortage and cost of traditional therapy.
• Cultural Considerations: The influence of cultural norms and stigma on mental health care access.
• Limitations of AI Therapy: Inability to address complex issues like self-harm, trauma, or psychosis.

1. Introduction to AI Therapy

• Many therapists are divided on AI's role, with some seeing it as a helpful assistant and others expressing fear.
• Millions use generic AI chatbots like ChatGPT and DeepSeek for mental health support, alongside dedicated therapy apps.
• The central question is whether AI therapy is safe, effective, and a potential replacement for traditional therapists.

2. The Mental Health Care Access Problem

• There's a significant shortage of psychiatrists and therapists globally due to training costs and time.
• Increased awareness and reduced stigma around mental health have led to higher demand, exacerbating the shortage.
• Private therapy can be expensive, leading people to explore online alternatives, including AI.

3. Types of AI Therapy Apps

• Generic AI Chatbots: Used for various tasks, including mental health support (e.g., ChatGPT).
• Dedicated Therapy Apps: Specifically designed for mental health, often based on CBT principles.
• Early AI therapy apps were basic, resembling online journals for tracking thoughts and moods.

4. User Experiences with AI Therapy

• Michael from Austria used ChatGPT for money anxiety, receiving a seven-day plan with exercises that proved effective.
• He highlighted the accessibility and cost-effectiveness of AI compared to the long wait times and high costs of traditional therapy (100 euros per hour in Austria).
• Michael found the AI empathetic, even more so than some therapists.
• The host, Hannah Gelbart, found generic chatbots overly positive and desired more realism.

5. How Dedicated Therapy Apps Work

• Many dedicated apps are built with mental health professionals and utilize CBT.
• CBT helps users reframe thoughts, particularly for anxiety and low mood, by prompting them to think differently.
• Some apps use LLMs, but the source of their training data is unclear, raising concerns about privacy and accuracy.
• Therapy data is typically private and not available online for LLMs to learn from.

6. Research on the Effectiveness of AI Therapy

• Research is in early stages, with varying results due to the diversity of apps.
• A New England Journal of Medicine study found a generative AI therapy bot as effective as human therapy for depression, anxiety, and eating disorder risk.
• The AI model was trained on custom datasets based on evidence-based practice and monitored by trained researchers.
• Long-term studies with diverse populations are needed to understand the full potential and limitations.

7. Cultural Considerations in AI Therapy

• Cultural experiences significantly impact therapy access and attitudes.
• The Chinese AI app DeepSeek has gained popularity among young people seeking emotional support due to privacy and anonymity.
• DeepSeek displays its thought processes, including reminders to be empathetic, which resonates with users.
• In China, only 20 out of a million people have access to mental health services, highlighting the need for alternative solutions.
• Mental health remains a taboo in many Asian families, making AI apps a preferred option.

8. Professional Perspectives on AI Therapy

• Therapists acknowledge the supply and cost issues in human therapy, making AI a potentially beneficial tool for accessibility.
• AI may help break down stigma, especially for digital natives comfortable sharing feelings online.
• Concerns arise when AI is seen as a direct replacement for human therapy.
• Psychologist Maytal Eyal warns that reliance on AI for emotional support could create unrealistic expectations for human relationships, leading to a decline in tolerance for discomfort and natural friction in intimacy.

9. Risks and Potential Harms of AI Therapy

• AI may provide harmful advice due to the complexity of human psychology and the nuances of mental health issues.
• In 2023, the National Eating Disorders Association replaced its helpline with a chatbot that recommended calorie restriction, leading to its suspension.
• Lack of regulation and accountability is a concern; it's unclear where to complain if an AI app provides harmful advice.
• Users should check for information on how to get more help and what happens if things go wrong.

10. Data Privacy and Security Concerns

• Sharing innermost thoughts with AI raises concerns about data leaks and misuse by marketers.
• Generic AI chatbots use data to train models, making privacy difficult to ensure.
• Mental health apps should have clear terms and conditions regarding data privacy, third-party access, and involvement of mental health professionals.
• The Mozilla Foundation found that 19 out of 32 leading mental health apps failed to protect user privacy and security.
• Reading the small print is crucial to understand how data is used and what privacy measures are in place.

11. Conclusion

• Experts advise against using AI as a replacement for therapy.
• AI companies often recommend users seek professional help, as these bots are not licensed therapists.
• AI therapy holds promise for increasing access and reducing stigma, but caution is needed regarding data privacy, potential harms, and the limitations of AI in addressing complex mental health issues.