> "more than 6 in 10 people declare they already use AI for mental health questions. 42% of them almost always follow the advice it gives them."
If these numbers were "of people who regularly use GenAI chat tools" then I'd be surprised it was quite so high already, but not shocked and would find it completely believable.
But this seems to be "of all people surveyed", which I'm rather skeptical of - unless their sample was very biased (as an extreme example, if they recruited people to the survey only by linking to it in ChatGPT ads, but there are plenty of less extreme ways to get a sample group that's way more likely to use AI than a genuinely random sample of the whole population).
It's also worth noting (and perhaps somewhat explains numbers seeming so unrealistically high to me) that, unless I've misunderstood, "turn to AI for psychological support" isn't necessarily "using AI as a therapist", it could be uses as minor as asking "Can exercise help with my depression?" or "If I think I am having a nervous breakdown, should I talk to a doctor?"
This is almost certainly from an online panel survey of people who are paid to take surveys, which is a mini-industry in itself, people try to make a living taking as many as they can for a few cents. The people who fill these out are completely unrepresentative of the general population and quite a few are probably having their openclaw fill it out for them.
That definitely fits with how I felt about the numbers they reported
Yes, this seems unusually and suspiciously high. Given that Axa is an insurance and investment company, may have some conflict of interest.
If you have baseline epistemic hygiene there's nothing wrong with using an LLM for advice.
If you have baseline epistemic hygeine you'd also recognized this as a B2B sales pitch: Axa sell group health, employee-assistance, and corporate wellbeing products.
I guess one can argue it's a novel form of (secular) divination, like tarot, except you have no idea what sort of intentions have shaped the AI during fine tuning, and by trusting it too much, you basically get puppeteered subtly by these companies.
Despite progress in natural sciences and tech, the answer to what's the good life is still up for judgment and values, it's not some neutral factual thing that simply falls out of AI training. Without fine tuning, supervised and RLHF, it could just as much role play a deranged psycho giving out life advice. To the extent that it isn't doing that, it's just adhering to whatever the company thought will land them in the least (legal and media) trouble. They had no way of actually tying it to life outcomes. It's more about what the user will like best in the moment. Which is a lot of affirmation and glazing. This is not unique to AI though, self help books can have the same failure mode especially when picked based on what feels nice to read.
Baseline epistemic hygiene would tend to preclude taking llm advice seriously, on any personal matter of consequence, in my view.
However "baseline" epistemic hygiene seems to be a somewhat distant goal for the vast majority now we live in an infosphere essentially comprising a Darwinian nightmare of competing agnotological agendas.
Fancy way to say conformist ?
I'm not a psychologist or a mental health professional, but I think that this might serve a similar purpose as journaling. It's obvious that AI can't "fix your problems," but just writing stuff down can help us process.
>>It's obvious that AI can't "fix your problems," but just writing stuff down can help us process.
Nobody can fix your problems, but having some one to talk to who listens without judgement and advices accordingly with empathy matters.
Thats what this is about.
Having said this, AI can write out helpful mini booklets on stressful situations in which humans have to work with each other.
I call it "interactive journaling"
agree. This cant do as much as genuine therapy (which requires another person, no way around it). But it is helpful in some ways. It helped me talk through something last week and it genuinely enriched my life in that way.
Does therapy actually work? Not by the numbers. tons enrolled, none cured.
It works at least as well as placebo!
Worked great for me. Big recommend. "Cured" is mostly an unspecifiable state, and while certainly there's lots still wrong with me, I am healed far beyond my expectations at the outset, so increment your count by one.
Cured would mean you are healed . No more therapeutics , no more abnormalities
the psyche is not the same as the body, the model of “disease”/treatment/cure is not the same and analogies are of limited utility
With most therapy the goal isn't to cure, but to manage and help cope in a healthy way. There are also plenty of mental illnesses or disorders that have no cure, and a few that are in the DSM because they cause problems with how our society is structured, not necessarily because its a true disorder (its only disorder because it causes issues functioning within societal systems).
They are made up. None of them are empirically or objectively defined
I think it's neither made up, nor actual health care.
It's basically a replacement for having cultural and social guard rails, lots of community, lots of straightforward expectations of roles etc. Today, people are atomized and lost, so they need a friend-and-priest-replacement who has authority but also patience to hear you out and has experience with judgement calls about people's lives.
But all this talk about "healthy" ways of coping etc is basically just there for medicine-envy and insurance reasons.
I think having one's mind/spirit sorted out is quite important, but the specific textbook strategies of today may be indeed mostly hogwash through a strict healthcare lens.
Therapy keeps people alive that wouldn't otherwise be, and people coping that wouldn't otherwise be able to.
I've noticed that the human tolerance for extreme suffering leads sometimes to binary thinking. "Well they're still going to work even though they're made to piss in bottles, they must be fine with it!" Human experience is a wide array of emotions and states, I don't think we should try to separate into "cured/healthy" and "unhealthy/requiring adjustment by a mental health professional." Improving quality of life is also good.
The suicidality blackmail
Pretty gross comment. Just say you're cool and awesome cause you don't care about people. Virtue signal and move on, why make try to make a fake discussion?
I was going though a problem I'm having parents - they are aging, decisions need to be made, that sort of stuff. Writing it out, thinking about it, reflecting further - I probably spent at least and hour just typing in my thoughts as they came to me and honestly I often didn't read the AI responses.
All of that got me to realize that the problem wasn't that I wasn't explaining myself well. I kept thinking that if I'd just found the right words they'd change their minds. The process of digging into not just where they are now but who they've always been, how they've always been. I need to accept that and move forward.
That's rubber ducking. Or talking to yourself in the shower. Just feels less silly, because in theory you could have read the response.
I recently had to do a very hated, very tedious, very long, very boring task at work and I asked Claude to hype me up and give me a pep talk to conquer the task and it really did a great job.
Then I got nervous if IT reads our prompts and felt very sheepish if they would be seeing my asking it that.
Heh. Sometimes when I'm stuck at work and feeling unmotivated/catch myself procrastinating, I'll write out my thoughts in my OneNote notebook that I take all work notes in. I do kinda cringe about the idea of IT reading those notes sometimes, but oh well, and it's helped a lot.
This is connecting to another commenter's idea that AI can be sort of a stand-in for journaling.
fwiw I asked my friend in IT who worked on the Claude rollout if they routinely read out prompts and he said no, they watch category of use and can usage stats then have a few alert flags if it is asked specific things such as security breaches, explicit keywords or about violence but otherwise they don't read it.
I love this and think it is a great use of the tech
now you have a great story for future job interviews asking about how you deal with hardship
I haven't done it yet due to privacy concerns, but I would totally do it with a private local model that's as intelligent as current frontier models and is not sycophantic and perhaps is finetuned on the psychology literature.
My reasoning is that, if therapy is a well-understood science, then I trust a big finetuned LLM more than a run-of-the-mill human therapist. I will not be able to afford a Harvard trained psychologist.
If therapy is more of an art and needs the human touch and mojo, then again, then again I'm not going to be able to afford Sigmund Freud or Carl Jung.
The few times I've tried human therapists, my impression was that the questions and answers were fairly standardized, which I think LLMs can excel at. Not to mention I'm more at ease talking to silicon- than carbon-based creatures.
My experience with therapy has been that it's more of an art than a science. It's hard to say what good therapy even is - different people need different things. I would not want Freud or Jung to be my therapist.
I think your ideal AI therapist doesn't exist and may never exist. Given current models, I have a hard boundary where I will not rely on AIs for therapy or companionship. There are just too many stories of AI psychosis, and it's too easy to see myself becoming dependent on them.
> then again I'm not going to be able to afford Sigmund Freud or Carl Jung.
Just spin in place chanting “it was my mother’s fault” and you’ll get 99% of the effect.
The problem is that all the frontier models tend to be more sycophantic when confronted with emotional support issues.
I believe sycophancy is a side effect of RLHF and whatever reward function it explicitly and implicitly optimizes.
"AI is replacing bartenders as everyone's therapist. On the bright side, the bartenders finally have someone to talk to about losing their jobs." -- Claude
Well it seems obvious that would happen. In fact, I think many more people are using them as doctors. You only get 30mins with a therapist and about 5 minutes in a crowded doctor’s office. It’s so great to go into detail and have the LLM cross reference everything better than using a search engine picking up every single detail.
Is anyone actually surprised? The baseline empathy and emotional intelligence of people, both online and offline, has absolutely tanked--I think Covid was the epoch, but arguably going a lot farther back since then. People are just shitty.
I’m surprised that 6 out of 10 people even use AI, let alone that many people use it as a therapist.
Yeah. I simply don't believe it. They do describe their methodology at the end of the article. It's a single sentence.
It increased generally with social media adoption and the installation of a screen between a person and their people.
Being more loosely connected may not be a substitute for the meaningful connection, both as a skill to learn, and participate in
retail facing acquaintances have confirmed to me that this started long before covid. personally I'm much more of a misanthrope than average, still to me it's more important to focus on systems (like the modern western system of no values no books no community no nutrients no free time no third space no thinking only consumption) as opposed to focusing on symptoms (eg. people behaving animalistically when confronted with a system that does not align with the needs of its occupants). we are all living creatures trying to optimize, therefore outside of edge cases there's functionally no difference between you, I, and 'shitty' people
Surprised and suspicious. A tiny fraction of LLM queries are what anthropic calls 'affective' queries. Very little LLM traffic is for anything other than factual uses, and of that a vanishingly small fraction is something like psychological advice
https://www.anthropic.com/news/how-people-use-claude-for-sup...
I suspect this company is deliberately using terrible data to drive eyeballs, which, mission accomplished I guess.
As much as I can be critical of LLMs I must recognize they help a lot putting a different (and generally healthier) perspective to things, I don't have to book a psychologist and can just "empty my bag".
Some rules I have for myself: limit the chat length, don't go too far down the rabbit hole, put one prompt with as much detail as possible and keep the LLM focused on one specific topic or issue I want to resolve, make sure I ask it to give me a nuanced take and not just validate my opinion.
Can see a place using it here. Most human psychologists patients see every week or so at a scheduled time, not because that is what is most effective in terms of treating your psychological issues but because of scheduling and cost reasons. Being able to get support at more regular intervals if done correctly may improve healing. The scheduled time may also not be the most effective time either. Events can happen on the day that reduce treatment effectiveness. The scheduling and resource allocation is a human problem (not enough psychologists, psychologists need their own personal boundaries and recovery time etc). A tool that can extend psychologists usefulness may be able to get the treatment that people actually need.
Then there is the money aspect. Too many people don't have the money for psychological treatment. Having a cheaper alternative will help. Unless we actually start thinking outside the box here we won't solve mental illness and so all tools should be used.
Is it really any surprise with how expensive psychologists and therapy are?
They’re trained professionals. And we’ve sent demand through the roof by convincing everyone and their dog that they need therapy. Despite it seemingly doing way more harm than good for a lot of people who probably don’t need therapy.
It's helpful tbh fam but I have no diagnosed issues, just ask for casual wellness ideas like a search engine (when compared to a search engine the headline is less surprising to me - "more than 6 out of 10 people search online for psychological support")
This is absolutely a good thing for the world.
Given how psychotic and demanding work environments have become, I definitely use it as a crutch.
This just makes sense. A normal day contains dozens of experiences that could be bettered by cheap actions that I am awful at predicting or imagining. I had an argument with a partner at one point where I was baffled and basically at a loss, asked chatgpt, and it spit back a response that seemed... okay. I adapted it into my own voice, keeping only what was sincere, etc (not just dumping LLM slop at another human, which is fucked, more like using it to coarsely choose a vector/filter through a big cloud of things I actually believe). My partner's response was incredible. It completely diffused the situation and my they were pleasantly surprised. Without the LLM, I would have been entirely unable to conceive of and walk that happy path.
The problems we have with our psychology often involve deviation from the normal or desirable state, so a robot that spits out a cheap reversion to the mean can be really helpful.
My flavor of this is somewhat autism-coded, but it generalizes well. EG people who aren't used to negotiating, valuing themselves, etc. Obviously LLMs output hallucinated dogshit and occasionally dangerous nonsense.
But it must be admitted: a lot of our psychological hiccups can be solved by the thoughtless, typical advice.
38% putting them over professionals is nuts though. I would much rather have the real thing, but it's $200/hr and asleep at 3am.\
Oh hell no. That's horrible.
I think there are situations where AI is good for quick advice, provided you _also_ have a professional to talk to if needed. I sometimes seek advice about dealing with difficult colleagues, or ask for opinions about things I said at work while a little emotional. These don’t affect my personal life that much so the stakes are lower.
I’ve been using grok voice chat for mock dialogs to help practice diplomatic and candid conversations. I’ve found people increasingly dismissive and aloof in conversation. I prompt grok (it’s in my car) to role play as a medical billing administrator or similar reluctant authority to help practice resolving the disagreement.
This is a little bit sad, but not that surprising.
For a while now I've wondered how valuable this really is for crowdsourcing of sentiment and opinions. We went from yahoo answers to reddit and now to this. Those previous ways of getting input were notoriously full of trolls and ulterior motives, but maybe a one-on-one conversational format with no distractions is a higher quality source? Is it a feature or a bug that the LLMs are biased in favor of whatever junk their owners want?
It's certainly a lot more satisfying than asking Reddit most of the time, whether you are getting the truth or not. I don't know how many times I've posted a question on reddit, not necessarily about something psychological or relationship related, even about things like mortgages or landscaping, come back to check the comments and ya, have my day totally de-railed by trolling strangers with an ax to grind.
>"Respondents report spending an average of 5.1 hours a day on screens during the week (excluding working and studying hours and excluding weekends), with screen time rising to 6.4 hours in the Philippines and Thailand.
Two out of three people believe that this exposure has several negative, even if moderate, effects on their mental health."
5-6 hours excluding work and study is mental. I know "touching grass" isn't exactly a professional treatment plan, but instead of spending more time in front of a screen to fix mental health issues have we tried prescribing people to actually go out?
If you're working eight hours, sleep seven, and maybe spend an hour or two cooking and doing daily chores, there's not even enough time left to exercise. no shit it's having a bad effect on their mental health, most of these people don't need a therapist, they need sunlight and their phone taken from them
It partly depends on whether "time spent on screens" is counting time where a person's attention is primarily on the screen, or time when a screen is actively directed at a person.
As a few examples from my life (I'm sure there are plenty of other such scenarios too):
- There are quite a few hours in a typical week where my phone screen is showing a video, but just because I can't have it playing in the background (eg YouTube without premium, although actually I've just installed a third party app to get around that for YT). I'm actually just listening with a wireless earpiece or two while doing something else.
- Time spent with a friend where we're sort of watching TV, but more than half the time our attention is on our conversation not on the screen.
- Time spent multitasking, whether that's doing a hobby while also watching TV, or texting people while also eating, or whatever.
Those types of things can make the difference between a certain amount of screen time being a much smaller or much bigger part of a person's day.
honestly the bar isn't "is AI a good therapist." the bar is "is it better than staring at the ceiling at 2am." for most people, it is, and that's enough to explain the numbers. I know im curious to see how it responds.