Key Considerations for Incorporating Conversational AI in Psychotherapy

Adam S. Miner, Nigam H. Shah, Kim Bullock, Bruce A. Arnow, Jeremy N. Bailenson, Jeffrey T. HancockFrontiers in Psychiatry 2019 · DOI

Abstract

Conversational artificial intelligence (AI) is changing the way mental health care is delivered. By gathering diagnostic information, facilitating treatment, and reviewing clinician behavior, conversational AI is poised to impact traditional approaches to delivering psychotherapy. While this transition is not disconnected from existing professional services, specific formulations of clinician-AI collaboration and migration paths between forms remain vague. In this viewpoint, we introduce four approaches to AI-human integration in mental health service delivery. To inform future research and policy, these four approaches are addressed through four dimensions of impact: access to care, quality, clinician-patient relationship, and patient self-disclosure and sharing. Although many research questions are yet to be investigated, we view safety, trust, and oversight as crucial first steps. If conversational AI isn't safe it should not be used, and if it isn't trusted, it won't be. In order to assess safety, trust, interfaces, procedures, and system level workflows, oversight and collaboration is needed between AI systems, patients, clinicians, and administrators.

Introduction

The demand for mental health services is currently outstripping the supply of qualified clinicians. In the United States, the ratio of psychologists and psychiatrists to the general population has declined over the last decade. Even when patients secure an appointment, the time dedicated to "talk therapy" is often curtailed by financial and administrative burdens. Conversational AI—including chatbots and digital assistants—offers a potentially scalable solution to bridge this gap. However, its integration into clinical practice has often been haphazard. This paper proposes a conceptual framework to evaluate AI-human collaboration in psychotherapy, moving beyond a simple "human versus machine" binary.

A Framework for AI-Human Integration

The authors categorize mental health care delivery into four distinct approaches based on the level of AI involvement:

  1. Humans Only: The traditional model where sessions are private and rarely recorded.
  2. Human Delivered, AI Informed: An AI system "listens" to or analyzes sessions to provide clinicians with feedback on symptom detection or adherence to therapeutic protocols.
  3. AI Delivered, Human Supervised: Patients interact directly with an AI, while a human clinician oversees the process, managing escalations or screening patients.
  4. AI Only: Fully automated systems designed for maximum scalability without direct human oversight.

To evaluate these approaches, the authors identify four "Dimensions of Impact": Access to Care, Quality, the Clinician-Patient Relationship, and Patient Self-Disclosure.

Table 1: Comparison of care delivery approaches across dimensions of impact.

Key Insights and Clinical Implications

Several critical themes emerge from this framework that challenge traditional assumptions about therapy:

  • The Irrelevance of the Turing Test: Clinical effectiveness does not require an AI to be indistinguishable from a human. Even simple systems can perform collaborative tasks that improve care outcomes.
  • Non-Consumable Resources: Unlike human time, AI is a "non-consumable" resource. It does not suffer from fatigue, its attention does not scale linearly with cost, and it can provide long-term continuity that human trainees cannot.
  • The Paradox of Disclosure: AI can be a double-edged sword for patient honesty. While some patients feel more comfortable disclosing sensitive information to a non-judgmental machine, others may be more reluctant to share with a system that has near-limitless memory and data persistence.
  • Therapeutic Alliance with Machines: Emerging research suggests that users can develop a "therapeutic alliance" with AI. This indicates that the "human ingredient" in therapy may be more about the interactive process than the biological nature of the therapist.

Safety, Trust, and Future Directions

The transition to AI-augmented therapy faces significant hurdles. There is currently a regulatory and legal vacuum regarding the "Tarasoff duty" (the obligation to warn if a patient poses a threat) and how liability is shared between clinicians and AI developers.

The authors conclude that safety and trust are the primary prerequisites for adoption: "If it isn't safe it should not be used, and if it isn't trusted, it won't be." Establishing robust oversight workflows and fostering collaboration between AI developers, clinicians, and patients is essential for the responsible evolution of mental health care.

Summary updated

Abstract

Conversational artificial intelligence (AI) is changing the way mental health care is delivered. By gathering diagnostic information, facilitating treatment, and reviewing clinician behavior, conversational AI is poised to impact traditional approaches to delivering psychotherapy. While this transition is not disconnected from existing professional services, specific formulations of clinician-AI collaboration and migration paths between forms remain vague. In this viewpoint, we introduce four approaches to AI-human integration in mental health service delivery. To inform future research and policy, these four approaches are addressed through four dimensions of impact: access to care, quality, clinician-patient relationship, and patient self-disclosure and sharing. Although many research questions are yet to be investigated, we view safety, trust, and oversight as crucial first steps. If conversational AI isn't safe it should not be used, and if it isn't trusted, it won't be. In order to assess safety, trust, interfaces, procedures, and system level workflows, oversight and collaboration is needed between AI systems, patients, clinicians, and administrators.
BibTeX
@inproceedings{miner2019key,
  author = {Adam S. Miner and Nigam H. Shah and Kim Bullock and Bruce A. Arnow and Jeremy N. Bailenson and Jeffrey T. Hancock},
  title = {{Key Considerations for Incorporating Conversational AI in Psychotherapy}},
  booktitle = {Frontiers in Psychiatry},
  year = {2019},
  doi = {10.3389/fpsyt.2019.00746},
}
Figure