1. Is there an urgent need for a digital therapist (DT)? Our answer
YES! The need for behavioral health services cannot be solved by therapists alone. There are not enough, and we can’t produce enough. We have known that for six decades. The public health need is urgent and massive. Individuals are desperate for
effective therapy, but
quality therapists are hard to find and prohibitively expensive.
Therapists are almost non-existent in many of the poorest countries.
There are already many excellent and useful mental health apps and chatbots in the market today, and more to come, that address this problem, each working on PART of the problem but none claiming to be the robust DIGITAL THERAPIST that could go a long way to solve this problem, not only in the U.S., but worldwide.
2. Can DT be done by the newest AI programs without a computational model base? Our answer
No, you would have all the common challenges being talked about today that have not been solved; guardrails, safety, ghosting and hallucinating. Such an AI-alone program could not be used responsibly or credibly in a medical environment.
But first a lesson from history. It’s easy to build a digital therapist that only appears to be an effective therapy.
In 1976, Professor Weizenbaum from M.I.T. launched the famous Eliza digital therapist. It was purposively and cleverly programmed not to be therapeutic but built instead to create the “strongly felt illusion of being understood.”
It worked spectacularly. People felt understood (but they weren’t, and it wasn’t therapy). It eventually became a popular parlor game.
Without strong guardrails, 7 decades later, that could happen again just because the modern AI language programs are now so good at carrying on a conversation that they naturally (not purposively) produce the “strongly felt illusion of being understood” without actually being either therapeutic or of being understanding.
AI alone can’t create a DIGITAL THERAPIST. You need an underlying robust model that captures and organizes the wisdom of therapy; that is broken down into micro steps; is thoroughly mapped; identifies context; is already programmed; and has been deployed and studied.
3. Can it be done by the newest AI with the right computational model? Our answer
Now, with access to generative AI and the base computational model described here, the essential ingredients are in place to build the science based-AI driven DIGITAL THERAPIST that is more like what movies, books and sci-fi fans have imagined since the early days of computers.
What was sci-fi yesterday is actually possible today…or at least possible enough to be built with confidence.
4. Do we have the right computational model? Our answer
We have already created, deployed and tested the state-of-the-art digital therapist component programs over decades through partnership contracts with major healthcare companies: CignaHealthPlan, Johnson and Johnson, IBM Watson and a large chain of psychiatric hospitals.
These programs have been successfully and safely used in hospital and outpatient settings by 100,000 patients; demonstrate health outcomes comparable to or surpassing in-person therapy. Patients love using them. They experience therapeutic outcomes they were empowered to make happen with DT, either with a therapist or totally on their own.
* Studied independently by UCLA/Kaiser and reported in peer reviewed journals, DT used alone without a therapist was determined to be as effective as with an in-person therapist.
* Studied by Managed Health Network, DT with a therapist outperformed traditional therapy by completing a case to patient satisfaction in less time and with better results.
The patients using our programs felt understood and they were helped.
We also believe we have the right and the best model base for the new AI driven Digital Therapist because we started from a brand agnostic position. Instead of picking one school of therapy as our model, we spent three years asking fundamental questions about the specific processes underlying the therapy practiced in this country before we did any coding.
We then coded, deployed, iterated, tested and launched.
In contrast, each of the many other mental health apps in the market today have had to choose one school of therapy (out of the dozens taught in this country) to be their therapy model.
We started building this broad sturdy base for a Digital Therapist by asking two fundamental questions:
“What do ALL therapists teach during a session?”
“How do patients learn how to take charge of their life when they are stuck?”
It took our research team 3 years to answer those two questions.
We then, in collaboration with an AI pioneer from Stanford, Dr.Ken Colby, translated those answers into very exact notational language that has guided the mapping and programming of our DT intellectual property ever since.
Instead of picking any one the dozens of brands and types of therapy to digitize as others have done, our initial 3 year research project asked: What are the explicit and implicit principles that all therapists teach during a therapy session? And what do patients learn as a result of that?
With that foundation we mapped out highly detailed step-by-step DT programs to help patients:
- Pinpoint what exactly is bothering them.
- Discover and articulate their best approach to address what’s bothering them.
- Identify and understand their resistance to actually taking the action they believe is the correct action to try first.
- Overcome their resistance to actually doing what they determined they needed to do.
The patient has learned and practiced a new way to master their fears and worries through the right combination of insights and action experiments.
The result is not just relief from symptoms. The patient is also equipped with an enhanced agency skill to build their life and address the challenges of the life cycle.
That critical piece of personal development is the often-overlooked implicit and ultimate goal of ALL psychotherapies.
That’s what therapists teach, and that’s what our foundational programs help patients learn, step-by-step.
OUR SOLUTION TO THE GUARDRAIL/EFFICACY/SAFETY CHALLENGE
Use generative-AI to enhance and expand our existing well-tested- computational-digital therapy program that works (understands, safe, and truly therapeutic).
5. Can DT be both safe and effective? Our answer
Yes, if the new AI programs are built ON TOP OF a carefully thought through, well-tested and detailed already programmed model like ours that:
- Establishes safe guardrails and knows when a mental health professional should be involved.
- Represents the core wisdom of successful psychotherapy.
- Truly “knows” the patient every step of the way.
- Establishes finely grained contexts at each step in the process.
- Helps the user discover, practice and learn a new way of thinking to master the conflicts blocking them from becoming the more effective and self-confident agent of their own life going forward.
Our therapeutic model is built to satisfy all five of these requirements for a safe and effective AI driven next generation digital therapist.
6. Will it be as effective and able to understand the patient as well as a human therapist? Our answer
Probably not as good as the very best therapists, but we could come close when we add the power of AI to deal with complex sets of data generated in our model. Every user goes through their own unique pathway when using our program. As they go through, we add to our context information about what and how they are thinking, and what feelings are attached to those thought patterns.
Since those pathways are already built, we have a known context that gets refined and elaborated each step of the way. Having a small context within a nest of other context modifiers is one of the conditions necessary to “understand” natural language.
This is what AI can do and a computational model cannot do. But combined, there is a way for DT to understand even more like the way that humans understand.
The AI partner can do more. Psychotherapists are taught to recognize psycho-dynamic patterns that are well documented in the literature. Each of those patterns can be captured and built into the program. We have already built two of the key psycho-dynamic patterns into our core program (self-doubts that erode self-confidence and the illusion of safety that interferes with making changes) both of which have been the subjects of books I have written.
If we feed the AI program ALL of the psycho-dynamic patterns therapists are taught, the Digital Therapist (with it’s almost perfect memory) may at times be more reliably useful than a therapist.
7. Can DT become more than a therapist? Our answer
Yes, with relatively little modification the DIGITAL THERAPIST can also become a
self-development educational program that takes into account the rules and values of other cultures, and how that influences the decisions each person has to make as they carve out a life and discover who they are becoming.
“A mind is a difficult thing to have…”
That’s how a man born deaf described his experience of being alone with his thoughts before a cochlear implant enabled him to hear and talk with others.
It’s a vivid reminder of the inherent self-talk activity of the brain/mind that has to be understood and processed by our conscious rational mind in order to make good decisions about your life, and your relationships.
Imagine having your own digital therapist helping you do that!
Can we really hope to understand, and help a person cope with and master that nuance and complexity of thought/feeling with artificial intelligence?
I am optimistic that can happen because the centerpiece of our core program is already about every-day decisions and actions. We can systematically explore how well known influencers distort rational decision processing.
8. Are we trying to create a digital version of the human mind? Our answer
No. The beauty and complexity of the sentient-self aware human mind is awesome. Not something AI or any computational program, or both combined, can replicate.
But an AI DigitalTherapist (DTAI) built on top of our computational program (DT) can interact with a human mind to carry on a meaningful dialogue that helps patients on a mass scale engage in a powerful therapeutic experience.
9. What can AI add to what we already have? Our answer
We already have the model and programs that work well, are stable and well tested, but which would work even better with an integrated AI approach (more languages, nuance, personalization, expanded scope and growing knowledge base about the human mind in pain and conflict).
In addition, AI can better track the emotional non-linear thinking processes over time, and with that knowledge we can enhance the strength of the critically important therapeutic alliance.