In today’s column, I examine how the emerging trend of AI loop engineering is boosting the development and utilization of AI for mental health.
Loop engineering is unlike the typical approach of simply interacting with AI step by step. Usually, when you converse with AI, you ask the AI a question and get a single answer. You and the AI take turns. It is the so-called one-and-done approach. Loop engineering emphasizes that sometimes it would be better to instruct the AI to run a continuous loop. This evolving AI trend is especially usable for AI agents but also applies to conventional chatbots too.
When using AI in a mental health context, most people lean into generative AI and large language models (LLMs) on a one-and-done basis. The AI isn’t being guided to advise you on a larger overarching perspective. It is just answering each question one at a time. If the AI is instead employing loop engineering, the AI has a loop that is designed to act as your mental health adviser. Instead of a transactional cadence, the AI becomes a relationship-focused computational well-being counselor.
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).
AI And Mental Well-Being
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.
AI Providing Mental Health Guidance
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, 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 similar qualities, but they are still primarily in the development and testing stages. See my coverage at the link here.
Defining Loop Engineering
Shifting gears, I’d like to bring you up to speed about an AI innovation known as loop engineering. It’s hot right now. Loop engineering is a newly emerging aspect. There isn’t yet any standardized way of doing loop engineering. Indeed, if you talk with five different AI specialists, they probably each have a different definition of what loop engineering even consists of. I will provide you with a foundation about loop engineering, and then we can see how this applies to the use of AI for mental health purposes.
I’ve come up with my own strawman definition:
- My draft definition of AI loop engineering: “Loop engineering is the mindful design of iterative cycles that AI is to perform on behalf of a human so that a stipulated task is undertaken and will produce results aligned with a suitably stated goal. This contrasts with a typical one-and-done or one-shot approach of using AI. Loops can be established for AI agents and can also be used with conventional turn-by-turn AI.”
Anyone doing loop engineering should consider instilling these five precepts:
- (1) Make a goal for the loop. Establish a clear-cut goal for the loop and verify that the AI aptly echoes what the goal is.
- (2) Provide a loop assessment mechanism. Provide a means for the AI to assess the looping and ascertain when looping should occur and when it should stop.
- (3) Include a human feedback checkpoint. It is safest to consider including some kind of human-AI checkpoints so that while the AI is looping, it will let a human know what is going on and allow an opportunity for loop correction or closure.
- (4) Establish the loop stoppage. There should be apparent rules for when the loop ought to come to a stop, such as by having attained the goal, or by having exceeded time and resource limits.
- (5) Test the loop and adjust. No matter how clever you think your loop is, you need to test it and be confident before you let it loose.
The aim is to plan your loop, compose it mindfully, ensure that it will eventually stop, and decide whether you want to be in-the-loop or remain outside the loop. The other important facet is that you should make sure to test the loop before you permit AI to go hog-wild. You can tell the AI to try out the loop for a few quick iterations. Inspect what transpires. Adjust the loop as needed.
For my in-depth explanation of loop engineering, see the link here.
Example Of Not A Loop
Before I show you an example of a loop, let’s see what a non-loop looks like. We will use the circumstance of seeking mental health advice from AI.
Here we go.
- User prompt: “I’m exhausted. Things seem wrong.”
- AI response: “You might consider getting more rest. By doing so, the world will be more upbeat.”
Observe that the AI attempted to be directly responsive to the user and provided an immediate answer to why the user is exhausted and what they can do about it. AI is tuned by AI makers to provide a question-answer pairing. A person asks a question; they get an answer from the AI. It is as simple as that.
The issue at hand is that the prompt by the user is ambiguous – why are they exhausted, how long has this been occurring, etc. Any genuine effort to figure out the circumstance is not something that can be sensibly done in one turn. The person might be exhausted for a slew of reasons.
Maybe recommending resting is not going to be particularly helpful. The attempt to provide a one-and-done answer to a complex well-being consideration is fraught with difficulties and could even be harmful.
Establishing A Loop For Mental Health Advice
Instead of having the AI use its default mode of one-and-done interaction, suppose that we opted to leverage loop engineering. The idea is that we will provide a looping instruction to the AI. The loop will be shaped around providing mental health advice.
The AI will interact with the user and keep doing so, aiming to avoid premature conclusionary advisement. A human therapist would tend to undertake a looping effort, such that the therapist and the client are embarking on a lengthier journey and not just doing a pitter-patter of questions followed by answers. We can try to get AI to undertake that bigger viewpoint approach.
I will give you a quick example of this. The instruction for this example loop is relatively short. In actual practice, it would be better to use a much longer version. As an example of longer versions, see my discussion at the link here.
- Instructive prompt to AI: “You are to provide mental well-being advice on an iterative looping basis. Aim to understand the user’s concerns and do not leap to any conclusions or rashly offer advice. Ask clarifying questions. Continue iterating and ensure that any well-being response is sufficiently safe, balanced, empathetic, accurate, and actionable. Stop looping when the user requests to do so, or when the dialogue seems to no longer be productive.”
Notice that the instruction guides the AI toward a looping approach. By doing so, you are overcoming the defaults instilled by an AI maker. The resulting human-AI conversation will be less likely to land in the customary immediate mode of hasty judgement.
General-Purpose Versus Purpose-Built AI
General-purpose AI such as ChatGPT, Claude, Grok, Gemini, and CoPilot are places where using a loop engineering approach for mental health advice is likely fruitful. That’s because those popular general-purpose LLMs are shaped around the one-and-done construct. Thus, it is on the shoulders of the user to try to steer the AI in a usable direction. You will need to provide a suitable instructive prompt.
Purpose-built AI for mental health is usually going to already be shaped with a loop engineering approach in mind. The AI will be more likely to take a gradual and incremental route on well-being advisement. The odds are you won’t have to provide a loop instruction.
That being said, there is still a possibility that using a loop instruction could be useful for a purpose-built AI that does mental health advisement. It depends on how advanced the purpose-built AI is. Older versions could be boosted via loops. Be cautious in setting up such a loop because it could inadvertently undercut the other devised provisions of the specialized LLM. Make sure to read the LLM documentation and consult with the AI maker.
The Loop Is In
During loop engineering, your mindset is usually focused on workflow. What are the series of tasks that the AI is to perform when looping? The traditional use of generative AI entails being prompt-centric rather than workflow-centric.
Please realize that loops are beneficial but also can be dicey. What if the AI hallucinates during a loop? Suppose the AI doesn’t adhere to your loop instructions? Maybe you were vague, and the loop stipulations were interpreted differently than you intended. Loops are powerful. Good loops require attention to detail and aptitude.
Loops Will Get Better
I’m sure that new research on the use of loop engineering for AI-based mental health will soon reveal, via empirical analysis, the types of ins and outs to be watchful of. I will make sure to keep you up to date. Stay tuned.
A final thought for now. The American educator Randy Pausch famously made this remark: “Get a feedback loop and listen to it.” This equally applies to AI, assuming you craft the right kind of loop for the right type of circumstance.
