Can AI Replace Therapy? 3 Limits of AI Mental Health Tools
A growing number of people are using AI tools to support their mental health. Some individuals are going as far as to use AI as a therapist. In the age of both rapid developments in AI and a deepening mental health crisis, it is easy to see why we would turn to AI tools to cope with mental health distress– they are accessible, easy, and provide quick answers. All of these factors make AI seem like the perfect choice for mental health care but I want to consider benefits to therapy with a human provider that are foundational for effective care.
The Benefit Of Human Empathy
AI struggles to replicate human empathy. Though many users of AI can be validated, reassured, or understood through conversations with various AI tools, human empathy in the therapy space is unique. While AI can recognize patterns to notice feelings in the user and can mimic supportive language, this is simply cognitive empathy. Authentic empathy requires the ability to feel through biological reactions, and shared vulnerability. Authentic empathy also is based on a choice to care– as AI is probability based generation, therapy can provide genuine empathy through therapist’s choice and morality to care. Sitting in a room with a therapist, speaking and experiencing feelings with another human, is a connection that cannot be replicated by AI. AI is ultimately a machine– and while it can provide responses that mimic feeling, it cannot feel itself. Therapeutic healing relies heavily on the therapeutic relationship which also includes the therapist's capacity to be attuned and feel with the client. This change via relationship is something AI struggles to replicate.
The Importance of Feedback And Bias
One of the unique and integral parts of the therapeutic relationship is the therapist’s focus on bias and honest feedback. In an interaction with AI, AI learns from the information you put into the system and seeks to give you information that you want. For instance, if you are asking AI to help you navigate a dispute with your partner that is causing you anxiety, AI will use the information you give it to understand how you feel. If it senses that you are seeking a specific outcome, AI will tailor its responses to the emotions and outcomes it thinks you are seeking based on its probability algorithms.
In therapy, a therapist can balance between emotional attunement and evaluating bias. A therapist’s job is not to always give you what you are seeking– it is to help you notice when what you are seeking may be enacting a pattern in your life that is unhealthy or unhelpful to you. Sometimes, this means noticing what you are seeking and working to understand why you are seeking it. It also can mean offering a different perspective or observation that goes against what you are immediately seeking. While opposite observations like this can be challenging to tolerate, they are where the real therapeutic change takes place. Healing and change do not happen in relationships that further enact patterns, like with AI. Healing and change occur in relationships where patterns are noticed and therapists can offer nuanced observations and approaches to examine these patterns and engage with the client in a way that offers a new perspective and emotional experience. Having an unbiased, honest, perspective is essential to therapeutic care and is difficult to achieve with information fed, probability algorithms, present in AI.
The Speed of Answers
One of the major benefits of AI is the speed at which it can provide responses. This can be extremely helpful in breaking down stigmas around mental health as AI can quickly spread information for important resources that help many individuals gain access to care. When AI is used as a therapist, speed can become dangerous. What feel like quick, sharp, and validating responses often lack the full context and exploration that a therapist would seek before drawing a conclusion. This matters because even in therapy, clients are particular in the information that they share. Therapists are trained in this dissemination of information and know how to investigate further to understand a symptoms or situation holistically before making an observation or recommendation.
AI is answer driven– meaning that its goal is not mental exploration but rather solving challenges presented to its models. While its speed of solutions provides instant gratification, it does not always take into account missing information, broader context, or historical information about the client. This is where speed gets dangerous. When answers are given quickly, they often disregard important parts of a client’s past that are also intertwined in the presenting problem but not explicitly stated when describing the problem. A therapist with a lasting relationship with a client will make these historical connections and help draw parallels for the client that extend beyond quick, specific, answers but give a more well-rounded view of the impact of a particular experience on a client.
While speedy answers make us feel good, they are not often where deeper therapeutic work is. That work is in learning to tolerate discomfort, ride out our emotions, and tolerate uncertainty. AI works oppositely to that challenge and so is not as effective in helping clients create sustained change.
As AI evolves I am sure its capabilities to relate and support the mental health concerns of humans will also evolve. For now, I would encourage you that if you find yourself leaning on AI for mental health support, you may find greater solace, change, and connection in a human therapeutic relationship. If you would like to talk about what that looks like, I invite you to reach out through my website for your free consultation.