I want to start this conversation with a caveat: I am not inherently anti-AI. I don't love the environmental impact of data centers, and I worry about how that will affect the already concerning trajectory of global climate change. But that's not what this conversation is about.
My husband is heavily involved in AI and has led teams that have developed AI processes and agents to help millions of patients in hospitals and veterans receive better care. I use ChatGPT regularly to "clean up this email" or "help me not sound so mad in this email, even though this is a stupid email that I have to respond to." I think AI is a great tool that, when used appropriately, can be incredibly helpful. Like any other tool, its value depends on how we choose to use it.
My concern is with my own field of mental health therapy, and what we stand to lose when we turn to AI agents to handle the work we decide isn't worth our time.
Hey, fellow clinicians: do you know why we write notes? No, really. Why do you think we write notes? Of course, the easy answers are "so we get paid," "because the code of ethics says I have to," or "because my supervisor is mean."
I had no idea this was such a source of contention in our field. I so appreciate Dr. Burns from Western Michigan University (my graduate school advisor, as well as a professor whose classes I had the privilege of taking throughout my time there). Dr. Burns really impressed on me and my fellow students the importance of note-taking and documentation. Of course, she told us about the American Counselor’s Association’s policy that if a session is not documented, it may be treated as “not done,” so timely, accurate note-taking is part of ethical compliance. And if something is “not done” it can’t be ethically submitted to insurance companies for reimbursement.
Beyond that however, notes are how we remember the session.
Did you know that writing notes, specifically by hand (even if you’re writing by hand via tablet and stylus) impacts how well our brains retain information. According to Scientific Impact,
“One of the most well-documented effects of writing is improved memory, and the mechanism behind it comes down to how many senses you involve. Handwriting activates a larger network of visual and motor regions than typing does, and this multisensory integration plays a direct role in how deeply information gets encoded. When you physically form each letter, your brain processes the shape, the movement, and the tactile feedback from the pen and paper simultaneously. That layered input creates a richer memory trace.”
Memory is one of the biggest tools we have in therapy. How can you recognize patterns in a client's thinking and behavior if you can't remember that they've done or told you something similar in the past? I've had so many clients who have been in therapy with other clinicians tell me, "I can't believe you remembered that! When did I tell you that?"
These clients aren't complimenting me, or at least, that's not how I interpret it. They're telling me that they're feeling like the things they've shared with me are important. That those things have relational significance. Being remembered can communicate that you're paying attention, that their experiences aren't interchangeable, and that you're holding the continuity of their story.
In addition, understanding and pointing out patterns that clients may be using to reinforce maladaptive behaviors or ways of thinking is Therapy 101. If you don't remember what's happening from session to session, you can't be as useful to your clients as you might otherwise be. You can use whatever techniques and evidence-based strategies you want, but if your clients aren't experiencing that attunement, you're missing the ground floor of therapy.
And, as we throw around regularly (and as has probably been attributed to a million different writers and researchers), the relationship between the clinician and the client is what makes therapy work.
I think that it’s a trap to rely on AI to write our notes for us, and disconnects us from the opportunity to remember the session and solidify it into our memory so we can be a better clinician.
But beyond the memory benefits we may be missing out on, I have a lot of other concerns about using AI in this way.
I hear from plenty of clinicians who are using these services, and I'm constantly fed the same advertisements on SimplePractice, Psychology Today, and elsewhere. You know the ones: "Just turn on the recording device, and we'll listen to your session and write your note for you." These services are becoming increasingly popular, and I have a couple of thoughts.
Number one: Where is that data being stored? Is it being stored in a HIPAA-compliant environment? Do I have a Business Associate Agreement (BAA) on file with these outside programs? Am I potentially exposing my clients to having their private health information stolen or, even if that information is being used to train models, is it being stored securely?
We are trained to be HIPAA-compliant in our recordkeeping, including understanding who has access to our clients' protected health information and how that information is handled. But have you personally read through the agreements you're signing when you agree to use these programs?
I know that maintaining proper documentation requires time and energy, usually unpaid time and energy. I understand the appeal of having someone (or something) take that work off your plate. But we still have an ethical AND legal responsibility to protect our clients. Convenience doesn't eliminate that responsibility.
Number two: What are we giving away of ourselves?
I am a human. My humanity is the point. It's what I bring into my sessions at the most fundamental level, even before we get into the evidence-based techniques I'm using. Remember that relationship we talked about earlier? People want a relationship with a human being. That's what they're coming to therapy for.
If an AI program is listening in on my sessions, it's learning my style and how I talk to people. My personality. The way I ask questions, the timing of when I choose to let things "hang," my humor, and how and when I choose to respond to certain things. These are all part of how I practice therapy. They're part of what makes me me as a clinician.
Recently, Headway, a platform that connects therapists with clients and facilitates insurance credentialing and billing, made headlines when it introduced a proposed provider agreement that included a perpetual, irrevocable, royalty-free license to use providers' names, images, voices, and other submitted materials. Following significant pushback from clinicians, Headway paused the rollout and said it would issue a revised agreement. And rightly so, clinicians had questions.
But imagine if those provisions had gone through as written.
If a model has listened to enough of your sessions to recreate a suitable facsimile of you as a clinician, and a company also has the rights to your name, voice, and likeness, what's stopping someone from creating a Melodie-Bot to provide "good enough" therapy? What's stopping insurance companies from incentivizing their subscribers to use AI therapy trained on actual therapy sessions, using information and clinical material that clinicians may have handed over without fully understanding the potential implications?
I realize that at this point, I probably sound like I'm putting on my tin foil hat. And I can hear the response now: "It's just note-taking software, Melodie. It's not that big of a deal." Or, "If people were going to use a chatbot as 'therapy,' they weren't going to come to therapy anyway."
And that's all true. But I, and many other clinicians, need to make a living. I have student loans. I have a mortgage. I have a dog who needs toys and food and training classes.
This isn't about feeling bad for the poor clinicians who are getting left behind by technology. This is bigger than that. Our field has a history of being necessary. We do meaningful work to help people heal, grow, and achieve the happy, healthy lives they want to live.
This is about what happens when the people making decisions about mental health care start treating the relationship, judgment, and presence of a clinician as optional extras rather than essential parts of the service?"
Insurance in the United States is a for-profit enterprise. And while insurance companies can and do provide valuable services, they also have shareholders to answer to and a responsibility to make money year after year.
COVID showed us that, when it comes to health care, sometimes "good enough" becomes the standard for what gets funded. As an example, maybe for a particular unit, best practice might be a nurse caring for five patients at a time. But if we can make it work well enough with seven patients per nurse, we save the hospital system the cost of another nurse's salary. That's money that can instead contribute to the bottom line or shareholder returns.
So how long before that same thinking trickles down to mental health care? We already have all this data. Why not build a bot that we don't have to pay at all, or that costs significantly less than what we're currently reimbursing clinicians to provide care?
But what if I'm wrong? What if therapy isn't going to be replaced? What if the data is being stored safely, or isn't being stored at all? What if this is just some big boogeyman my brain is serving up?
I still think there's value in taking your own notes and doing your own documentation. I still think it's better for both the clinician and the client when clinicians actively engage in the process of remembering, reflecting on, and documenting what happens in their sessions. It's another way of strengthening the clinical relationship and improving our ability to recognize patterns, maintain continuity, and show up for our clients.
And if I’m right? Well, I’m not going to be handing over my likeness anytime soon.
