Therapy chatbots and emotional complexity: do therapy chatbots really empathise?

Area of Science:

• Mental Health Technology
• Human-Computer Interaction
• Digital Therapeutics

Background:

• Internet-based cognitive behavioural therapy (iCBT) and AI-powered therapy chatbots are increasingly utilized for mental health support.
• Existing research highlights both the potential benefits and significant risks associated with these digital tools.

Purpose of the Study:

• To review existing studies on therapy chatbots, focusing on their risks, benefits, and 'empathetic design'.
• To critically evaluate the role of AI, including large language models (LLMs) like ChatGPT, as 'AI therapists'.

Main Methods:

• Literature review of studies examining therapy chatbots.
• Analysis of research concerning chatbot efficacy, user experience, and ethical considerations.
• Evaluation of 'empathetic design' claims and the nature of AI empathy.

Main Results:

• Therapy chatbots can alleviate symptoms of stress, depression, and anxiety through emotion tracking and self-reflection.
• Concerns include short-term study durations, limited sample sizes, self-report biases, privacy, data security, and lack of human oversight.
• Users may anthropomorphize AI, leading to emotional dependency, while AI exhibits cognitive but not affective empathy.

Conclusions:

• While AI chatbots can offer certain therapeutic benefits, attributing human characteristics like 'empathy' is problematic.
• The potential for manipulation and user dependency necessitates careful consideration and avoidance of anthropomorphism.
• Robust ethical guidelines and human oversight are crucial for the responsible development and deployment of AI in mental healthcare.