People can use general-purpose AI (GPAI) for mental health advice, but there are better options via using purpose-built AI (PBAI).
getty
In today’s column, I examine an important tug-of-war that is rapidly emerging between people using general-purpose AI (GPAI) for mental health guidance and those same people leaning into purpose-built AI (PBAI) that avidly specializes in providing mental health guidance.
Here’s the backstory. Hundreds of millions of people are currently using general-purpose AI such as ChatGPT, GPT-5, Claude, Grok, CoPilot, Gemini, etc., as their go-to for getting mental health advice. This makes sense since they are already using that same AI for answering questions about fixing their car, cooking a roast, and resolving other daily considerations. The use of the same AI for mental health purposes is easy, seamless, and friction-free.
The downside is that general-purpose AI is not especially suited for acting as a mental health advisor. The AI is a generalist, not a specialist. Meanwhile, customized specialty AI that is purpose-built for mental health guidance is increasingly entering the marketplace. This tailored AI is intentionally designed and shaped to better aid people in their mental health quest. For my in-depth analysis of general-purpose AI versus purpose-built AI, see the link here.
A confounding problem is emerging between these two options. Suppose a person uses general-purpose AI for mental health advice and then discovers a purpose-built AI for mental health that they really like. The person might decide to exclusively use the purpose-built AI whenever a mental health aspect comes up. But that ideal way to proceed has potholes. Some of their mental health discussions might already be trapped in the general-purpose AI and not known to the purpose-built AI. This forces them to start all over again from scratch. Also, they might suddenly have a flash aspect to ask while otherwise in the midst of using general-purpose AI and find it difficult to suddenly switch over to their purpose-built AI.
I assume you can see the dilemma at hand. An obvious and unrealistic solution would be if the general-purpose AI refused to answer mental health questions; a user would then have no choice but to switch to the purpose-built AI for those facets. There is almost no reasonably conceivable way that the AI makers of general-purpose AI are going to give up on their AI handing out mental health guidance -- it is a monetizing lure for existing and prospective users. It is the tail wagging the dog.
What can a purpose-built AI do to try to entice users to predominantly use the purpose-built AI for their mental health and only sparingly use the general-purpose AI in that manner?
Let’s talk about it.
This analysis of AI breakthroughs is part of my ongoing Forbes column coverage on the latest in AI, including identifying and explaining various impactful AI complexities (see the link here).
As a quick background, I’ve been extensively covering and analyzing a myriad of facets regarding the advent of modern-era AI that produces mental health advice and performs AI-driven therapy. This rising use of AI has principally been spurred by the evolving advances and widespread adoption of generative AI. For an extensive listing of my well over one hundred analyses and postings, see the link here and the link here.
There is little doubt that this is a rapidly developing field and that there are tremendous upsides to be had, but at the same time, regrettably, hidden risks and outright gotchas come into these endeavors, too. I frequently speak up about these pressing matters, including in an appearance on an episode of CBS’s 60 Minutes, see the link here.
Millions upon millions of people are using generative AI as their ongoing advisor on mental health considerations (note that ChatGPT alone has over 900 million weekly active users, a notable proportion of which dip into mental health aspects, see my analysis at the link here). The top-ranked use of contemporary generative AI and LLMs is to consult with the AI on mental health facets; see my coverage at the link here.
This popular usage makes abundant sense. You can access most of the major generative AI systems for nearly free or at a super low cost, doing so anywhere and at any time. Thus, if you have any mental health qualms that you want to chat about, all you need to do is log in to AI and proceed forthwith on a 24/7 basis.
There are significant worries that AI can readily go off the rails or otherwise dispense unsuitable or even egregiously inappropriate mental health advice. Banner headlines last year accompanied the lawsuit filed against OpenAI for their lack of AI safeguards when it came to providing cognitive advisement.
Today’s generic LLMs, known as general-purpose AI, such as ChatGPT, GPT-5, Claude, Gemini, Grok, CoPilot, and others, are not at all akin to the robust capabilities of human therapists. Meanwhile, specialized LLMs are being built to attain those desired qualities, though such AI is still primarily in the early development and testing stages. For more about purpose-built AI apps in mental health, see my in-depth coverage at the link here and the link here.
Consider these five major patterns of how people dip into using AI for getting mental health advice:
Users are fluidly moving from category to category as they become more aware of GPAI limitations and become cognizant of BPAI availability.
The mainstay of the world right now is currently in category #1 of using general-purpose AI exclusively as their mental health guidance. These users either don’t know about the purpose-built AIs or are unsure of which purpose-built AI is safe and trustworthy to utilize. The users are not making a conscious decision to avoid purpose-built AI for mental health. Instead, they are comfortable using general-purpose AI and do not know of or perceive a suitable need to use any purpose-built AI to do so.
The next in popularity would be category #2 of mainly using general-purpose AI for their mental health advice, plus sometimes using purpose-built AI. These users are considered “innovators” and “early adopters” who are the first to opt to try using purpose-built AI for mental health and see if it is any good. They are sitting on both sides of the fence for the moment. They still like using general-purpose AI and the convenience of seeking mental health guidance while logged into the GPAI. Nonetheless, they are also attracted to the purpose-built AI and use it too.
Category #3 is inching gradually upward, whereby BPAI for mental health is a mainstay, and those users consider GPAI-based mental health advisement to reside lower on the totem pole. This is admittedly making slow headway. Meanwhile, a very small portion of people are in category #4, whereby they always and only use BPAI for their mental health guidance. They stridently avoid doing so in general-purpose AI. There aren’t many like this when compared to the vast volume of AI users all told.
Category #5 consists of those who might be using GP AI for a variety of uses but refuse to use it for any mental health advice. Likewise, they aren’t using purpose-built AI for mental health guidance. Why the overarching reluctance? Several plausible reasons. Perhaps they don’t believe AI can be helpful or prefer not to divulge their innermost secrets to AI.
All else being equal, people are going to gravitate toward using general-purpose AI over the use of purpose-built AI for their mental health guidance. An exception would be if a human therapist assigned a person to use a purpose-built AI app as part of the therapy under the supervision of the therapist. Even in that case, the chances are that the person would still occasionally dip into asking GPAI for their mental health questions.
The basis for this gravitational pull toward GPAI is shaped by these nitty-gritties:
In contrast, PBAI for mental health purposes tends to be perceived this way:
General-purpose AI has an aura of holistic coherence, albeit it might be somewhat weaker on mental health facets than PBAI, but that’s okay, and people are willing to sacrifice or undercut the technical psychological precision accordingly.
What can a maker of purpose-built AI for mental health do about this conundrum?
First, face reality. There are some PBAI mental health firms that are utterly convinced that they have the best thing since sliced bread, and it has consequently blinded them. They fervently believe they have made a better mousetrap. The world at large should march to their doors. The gist is that they assume that once people get a taste of their PBAI, they will find it irresistible and forgo the use of GPAI for any mental health advisement. Sorry, wake up and smell the roses.
Second, acknowledge that people will undoubtedly gravitate toward GPAI. Accept this as a key precept. Once that’s firmly implanted as a proposition, rethink how your purpose-built AI can live in coexistence with the use of general-purpose AI. Set aside the us-versus-them mindset. Aim to shape your PBAI as a considered complementary or augmentation of GPAI usage. Say to yourself, how can the two worlds work hand-in-hand?
Third, reduce the switching cost as much as possible, driving toward near zero. Make the PBAI so that it is easily left open and ready for use, such as integrating with browser extensions, OS-level hooks, keyboard shortcuts, and the like. There should be an always-on presence that reduces hesitation about conferring with the PBAI.
Fourth, ensure that the PBAI lives up to its perception as a technical psychological “expert” that outdoes and overshadows the capabilities of GPAI in the mental health realm. Deliver aspects that GPAI cannot do or does poorly. For example, have available structured therapeutic journeys, maintain tiptop longitudinal memory about the person, produce actionable follow-through in addition to giving advice.
Fifth, play up the fact that your PBAI is authentically rooted in mental health foundations and, therefore, unlike the haphazardness of GPAI when it comes to psychological underpinnings. Think of handy ways to stake your claim. Prominently indicate clinical frameworks that are employed (e.g., CBT, DBT, ACT). One cautionary note is do not overplay that technical precision hand. If the BPAI seems sterile and institutional, this will drive users back over to GPAI to get that natural, conversational tone and semblance. Be clinically grounded, but conversationally fluid.
Sixth, build habits. PBAI makers can encourage people to swing into the BPAI for daily check-ins, dear diary updates, streaks, or progress tracking, and for other useful mental health purposes. Over time, the user starts to naturally consider the PBAI as the right place to be. The mantra should be “this is the place I go to aid my mental well-being and enhance my mental health.”
The strategic goal is to substantively prove to people that your PBAI is where they actively work on bettering themselves and improving their mental status, and that this coexists with the use of GPAI. GPAI is where they ask questions, while BPAI is where they get answers (well, that’s kind of a cutesy phrasing, but you get my drift).
I’ve previously covered the potential for business relationships and AI-to-AI aspects of PBAI makers and GPAI makers, namely, connecting to each other behind-the-scenes, see my discussion at the link here. In those analyses, I’ve predicted that some GPAI makers will gradually end up opting to “embed” a BPAI mental health app within their general-purpose AI. This has many upsides, including contending with newly emerging AI laws about the use of AI and mental health, see my coverage at the link here.
Here’s how this might work.
Generic AI is acting somewhat like a generalist when it comes to mental health. At some juncture, the generic AI opts to transfer the user to the customized LLM (GPAI links to BPAI). There are many ways that this might be triggered. In any case, assume that the generic AI and the customized LLM are sharing conversations and contexts.
The customized LLM doesn’t skip a beat and immediately computationally discerns what the user has been discussing with the generic AI. If the customized LLM needs to deal with something that the generic AI stated, so be it. The customized LLM at least is already up-to-speed and can adroitly address whatever else was already told to the user.
In fact, here’s an even more elevated or deeply embedded approach. While a user is using generic AI, the customized LLM is always silently paying attention to the conversations taking place. The advantage is that the customized LLM can perhaps detect that a mental health matter is arising. The customized LLM alerts the generic AI accordingly. Thus, before the generic AI gets itself into a bind, it hands over the conversation to the customized LLM.
There are going to be a lot of PBAI for mental health that arise and then come and go. Sometimes they will make a big splash and seem on the verge of taking over the marketplace. Sadly, we’ve seen ones already that have crashed and burned. It is a rough road for PBAI in mental health.
That being said, the possible upsides for high-quality PBAI in mental health are tremendous and provide a sensible, worthwhile ambition. Just look at the stats of societal facets. The growing woes of mental health difficulties on a global, widespread basis are a phenomenon that rationally calls for AI to come to the rescue. There are not enough human therapists that can meet burgeoning needs. AI can. The AI must, though, be done right, provide mental health guidance that is safe and sound, and be discoverable and known well enough to survive past the early days of entering this rapidly evolving marketplace.
The famous words of Leo Burnett come to mind: “Don’t tell me how good you make it; tell me how good it makes me when I use it.”
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。