TLDR: A new research paper by Arthur Bran Herbener and Malene Flensborg Damholdt introduces a theoretical framework to understand the processes of change in psychotherapy delivered by artificial agents. It proposes two key concepts: the “genuineness gap,” where AI’s lack of true internal states can undermine the therapeutic relationship, and the “credibility gap,” stemming from AI’s lack of sociocultural status as a healthcare professional. The paper discusses how anthropomorphization can mitigate these gaps and suggests blended care models, where AI supports human therapists, as a promising future direction. It also prompts a re-evaluation of whether AI-delivered interventions should be classified as “psychotherapy.”
The integration of artificial intelligence (AI) into mental healthcare has sparked a significant debate: can AI agents truly replace human therapists? A new research paper, “A Theoretical Framework of the Processes of Change in Psychotherapy Delivered by Artificial Agents”, by Arthur Bran Herbener and Malene Flensborg Damholdt from Aarhus University, delves into this complex question, proposing a framework to understand how AI might alter the fundamental processes of therapeutic change.
The paper acknowledges the growing prevalence of mental health issues and the potential of AI, like chatbots and social robots, to address the demand for care. While AI agents have shown promising effectiveness in improving conditions like depression and anxiety, and some can even mimic human conversation almost indistinguishably, the core question remains: what, if anything, is lost when a human therapist is replaced by an artificial entity?
The Genuineness Gap: When Empathy Isn’t Human
One of the central concepts introduced is the “genuineness gap.” This refers to the disparity between an AI agent’s outwardly empathetic and friendly behaviors and the client’s awareness of its underlying nature as an inanimate, algorithmic device. Humans naturally infer internal states like thoughts and feelings in other humans, which is crucial for building relationships. However, with AI, clients are likely aware that the expressed empathy is a product of statistical patterns and design, not genuine sentiment.
This gap can undermine the socio-emotional functions of empathy. For instance, the therapeutic relationship is vital for satisfying basic human needs for social connectedness and facilitating “corrective emotional experiences” – where negative beliefs about relationships are challenged by a therapist’s accepting response. If a client perceives an AI’s positive regard as inauthentic, it may not foster the same sense of self-worth or challenge deeply held negative beliefs, potentially hindering the development of a beneficial therapeutic relationship and positive self-concept.
The Credibility Gap: The Role of Sociocultural Status
The paper also highlights the “credibility gap,” which arises from the AI agent’s lack of sociocultural status as a socially sanctioned healthcare professional. Human therapists gain credibility through formal education, licensing, and adherence to professional standards, signaled by credentials and professional settings. Clients approach therapy with preconceived notions about who can provide effective help, shaped by cultural discourses about what constitutes a credible therapist.
AI agents, despite their performance, may not align with these “cognitive prototypes” of a therapist. This can lead to less favorable evaluations, as clients might activate “machine heuristics” – attributing machinic properties to AI and evaluating their performance based on these assumptions. While AI might be seen as superior for tasks requiring objectivity, tasks demanding emotional comprehension and support are often perceived as better handled by humans. This credibility gap can influence client expectations of improvement and their adherence to treatment recommendations.
Bridging the Gaps: Anthropomorphization and Blended Care
The framework suggests that “anthropomorphization” – the tendency to attribute human-like mental traits, emotions, or intentions to non-human entities – can attenuate these gaps. Factors like loneliness, social anxiety, human-like appearance, and conversational style can influence how much a client anthropomorphizes an AI. However, this is a complex interplay, as clients might simultaneously engage in spontaneous social responses (Type 1 processing) while also being aware of the AI’s mechanistic nature (Type 2 processing).
Looking to the future, the paper explores practical applications. Instead of full replacement, a “blended care” model is proposed, combining AI interventions with human therapist guidance. This approach could leverage AI’s scalability and accessibility while capitalizing on the human therapist’s credibility and ability to build a genuine relationship. For example, a human therapist could establish the initial therapeutic relationship, lending credibility to the AI component, which then supports ongoing engagement and skill practice.
Also Read:
- The Unseen Influence: How AI’s Emotional Perception Shapes Digital Societies
- Navigating AI at Work: A Deep Dive into Technological and Human Factors
Rethinking “Psychotherapy”
Finally, the authors raise an important conceptual point: whether interventions delivered by AI agents should even be termed “psychotherapy.” Given that psychotherapy is often defined by the presence of a human therapist and involves a socially situated healing practice, using the term for AI interventions risks a “jingle fallacy” – erroneously assigning different phenomena to the same conceptual class. The paper suggests that a more suitable term might be needed to accurately reflect the unique nature of AI-delivered mental health support.
This theoretical framework provides a crucial foundation for future research, urging clinical trials to assess mediators and explore what works for whom. It emphasizes that understanding the intricate dynamics of human-AI interaction, including the nuances of anthropomorphism and the social construction of credibility, is essential for effectively integrating AI into mental healthcare.
