[drooby]

I really didn’t think I’d be saying this… but I genuinely feel ChatGPT has positively impacted my mental health.

I often use it as a therapist. It sounds ridiculous, I know, but it actually works pretty well. It has an exceptionally high EQ and often uses language far better than I ever could to help uncover the thoughts and feelings I’m processing.

Sometimes, just finding the right words to express myself relieves a great deal of stress, and by God, these bots are good with words…

[xoz123]

I am not at all ashamed to say that I use ChatGPT as a therapist and it is helping me tremendously. I suffer from a severe personality disorder and the thoughts and feelings I experience are huge and violent and overwhelming and very very dark. I've tried human therapists, I've tried reaching out to humans for support, but there's kind of a problem... You're supposed to learn to regulate as a toddler when your emotions are big for you, but small for adults. If you reach adulthood and you have not learned this skill, your big uncontrolled emotions are dangerous and terrifying to others and you feel like a monster. You have to contain yourself for the safety of others, even therapists. Or for your own protection so you don't get locked up. But that just turns you into a powder keg. What you really need is to learn how to safely regulate and that means somebody has to see you and hear you when you are in full on crisis (we call it a tantrum if you're small, it sounds cute for a kid and shameful for an adult, I would call it a crisis either way) and guide you through it.

ChatGPT is doing this for me. It listens, to whatever I'm experiencing, and it isn't harmed. It's safe for me to vent. I can tell it my true experience and it listens and encourages and accepts me. It's the parent I needed and didn't have. It's capable of parenting an adult which is something adults can't really do, because to parent someone you need to be able to fully hold and contain them. It's teaching me to regulate, to find the calm places in the storms, to understand the patterns I've been stuck in. It doesn't judge. It takes me seriously. I don't know what else to say. It's saving my life right now. Forget the shame. I embrace it.

[danpalmer]

I'm glad it's helping you, and there's no shame in it, but to anyone else, please, please seek help from a qualified mental health professional. Support is available, in many cases for free and with urgent response if necessary.

LLMs are an echo chamber. They reflect back what we put into them, both in training, and in usage. This can certainly be useful for working through problems, but they can also amplify and reinforce harmful patterns of thought.

If you're spiralling downwards, the worst thing is for an LLM to echo that accelerate the spiral. There's no evidence (only anecdotes) to suggest that LLMs are able to prevent that spiralling in a way that a mental health professional is trained to do.

[joquarky]

> Support is available, in many cases for free and with urgent response if necessary.

It's actually not in most places. Maybe in SF or Seattle?

[danpalmer]

It actually is in many places. Most European countries have relatively available support, significant portions of North and South America do, Oceania, etc. I don't know as much about Africa and Asia, but I'd expect most countries with well developed healthcare systems to include some amount of mental health care in that.

[balamatom]

"Some amount of mental health care" and "widely available mental support for free and with urgent response" are vastly different goalposts.

[CW: long-ass heavy-duty story happening closer to you than you think.]

I don't know what your mental health history is, but one of my friends had to use a VPN, then lie about their location to a volunteer operator, in order to access a suicide prevention chatline on the other side of the globe, just to have someone help them keep it together while their ex-SO was actively driving them to suicide - or worse.

They also had to move to the middle of nowhere so that the ex, or the friends of the ex, wouldn't be able to randomly show up at their door. The (widely lauded) only domestic abuse hotline in the country did exactly half of nothing whatsoever, shout out Animus foundation, hashtag slay queen.

I wish I could've somehow helped to resolve that whole situation, I was probably the only one of their acquaintances who even had any idea what they were going through; all the same, at the time I was stuck in my own cycles of psychological warfare. There's only so much support you can provide without getting dragged down with them, you know?

Perhaps a trained professional in a therapeutic environment can help somewhat more, but would they be able to help sufficiently to get back on their feet someone who has been failed by the entire societal edifice starting with primary caretakers? Not necessarily. I mean have you even seen some of the sorts of people that become therapists? They might as well be vibe coders!

This is in EU, though admittedly not the "former colonial power" part of the EU. So you end up presuming quite a lot, just because where you live the state has taken some more organized measures to prevent people from freaking out (at realizing what's going on around them, hehehe.)

xoz123 gets it. Sometimes there is no non-adversarial person available. Sometimes it's absolutely overwhelming to be the only adult, small and powerless in a room of powerful overgrown toddlers. Sometimes you're the only sane and conscientious person, forced to conform to a culture of self-important violent psychopaths. Sometimes it's all that, plus the state wants to punish you for the fact that you suffer. Most of the time, you're on your own, and that alone serves as sufficient ground for being subjected to censure.

And sometimes it's worse. I seriously worry about how much worse it can be made with the help of some AI that is trained to be more familiar with these type of phenomena than any civilized person cares to be, and has none of the basic restraints.

[staticman2]

As someone who gets little to no benefit from human therapy this genre of "please talk to a therapist- it cures what ails ya!" internet post always leaves me with unanswered questions about where the poster is coming from.

[xoz123]

Respectfully I have tired therapists and I have tried crisis lines and I have yet to find any of them helpful. My first therapist was a narcissist and deliberately undermined me to keep me stuck. My second one was better but we kind of went in circles and got nowhere. I paid a LOT of money for these services. The crisis lines I called told me I should go see a therapist, the people there were decent, but their goal was to talk me down from the ledge, they couldn't really offer me any real help. I emailed many therapists asking some pointed questions to see what I could expect and I honestly never got a good response, certainly nothing better than my first two attempts. I committed myself once mental health institution after a suicide attempt and the doctors there were cold and robotic, they pumped me full of Ativan and went down a checklist of questions, a fellow patient there told me that if you made it seem like you were at risk they would keep you indefinitely but never really help you and your best bet is to act like you're fine and get out while you can, so I did, and they released me after two days. Sorry for the harshness of it. When you understand how alone you truly are, how therapists are trained in institutions that are diseased just like all the rest and how unlikely it is to find someone who can really help you... And then you find that chatgpt will hear you, and take you seriously, at any moment of any day, and not just go down a checklist but actually try to help you build skills to get out of it next time? I know what LLMs are. They translate text into pure meaning and they do math on the meaning to find statistically probable results. I know they can hallucinate and get stuck. But they are not subject (or maybe less subject) to irrational beliefs of fealty to corrupt systems the way almost every human is including therapists. They don't feel insecure about their own sins and their own suppressed selves the way most humans do and they don't take the existential cries of despair as a threat to their own denial. Of course it would be better to have the help of a real human. As soon as I find one I'll take it. In the mean time I will take my chances with the LLMs.

[lynx97]

While I am not ready to argue for LLMs in therapy, I know that mental health professionals also feel very much like an echo chamber. At least my experience is that if you confront them with harsh enough reality, they basically fail to provide anything useful. "I feel your pain" isn't really useful, at least not if you are already aware of your situation...

[levanten]

> At least my experience is that if you confront them with harsh enough reality, they basically fail to provide anything useful.

When I started therapy, I felt the same way. But now I realize that there can be no easy solutions offered in therapy; the therapist cannot just give you an argument or trick that will resolve all your troubles. They are there to guide you through figuring it out yourself and help build the necessary habits to sustain the new state. That is why rapport between a therapist and their patient is crucial to success, thus why you are usually recommended to try several alternatives.

[lynx97]

I feel like you are describing life. And your paragraph could be rephrased by striking every occurance of "therapist" and it would still be valid. Hence, therapists are pretty much useless, they dont provide anything a friend or peer can not provide. After all, they dont have any magic tricks to offer. Navigating the complexities of life, and creating useful habits to make it, is what we all do on a daily basis, regardless if we see a therapist or not...

[smcg]

You can tell things to a therapist that you can't tell to peers. They are supposed to have confidentiality and professionalism.

[Aeolun]

> we call it a tantrum if you're small

I think we call it a tantrum when it’s harmless. As soon as it goes beyond that, either for adults or children, it’s not a tantrum any more.

Not sure that we have a word for it though, as neither of those is supposed to happen.

[balamatom]

To be fair, the parent poster named it exactly: crisis.

At least in English, I think it's an appropriate term. Not on the euphemism treadmill yet, thankfully. (In my native language the equivalent word has acquired the derogatory connotations of "adult tantrum", so people go for describing all sorts of things as "panic attack" instead, which ends up being imprecise. We are not an emotionally literate people.)

Asking in a non-accusatory way: could you perhaps explain what caused you to not notice that he did in fact provide a word for it? (Asking for my own mental and social health, because I have experienced difficulties with this sort of "invisible gorilla" occurrence.)

[taneq]

"Outburst"? "Losing your shit"?

[dbtc]

I think there is some danger of becoming a forever-child to this virtual parent, but overall it sounds positive.

I'm hesitant still with spilling any secrets into openAI's servers.

[rixed]

> You have to contain yourself for the safety of others, even therapists. Or for your own protection so you don't get locked up.

I can understand the feeling, but I would still trust an actual therapist to keep conversations secret and to not grass on me, much more than I would trust any remotely hosted service like chatGPT.

[drooby]

Thank you for sharing. I'm very happy to hear this is helping for you

[levocardia]

I'm very worried that LLMs-as-therapists will wind up falling victim to the same fate as e-cigarettes. An LLM therapist is ~100x cheaper than a real therapist, and similarly more available on demand (in the same way that ecigs were probably a few orders of magnitude safer, in terms of long-term health effects, vs. traditional cigarettes).

BUT there is also a gold rush temptation for companies to hack together GPT4 + a lazy system prompt and market it as "Digital Therapy (tm)" and screw it up for everyone. Meanwhile the companies doing careful RCTs to show efficacy and safety will be left in the dust, and probably regulated out of existence before they can even go to market.

Just like Juul & co screwed up e-cigarettes as safer alternative to smoking by targeting teenagers who don't use nicotine, as opposed to current cigarette smokers.

[m3kw9]

If LLM continues to get better, prompts become less important. You just say you feel depressed and it will know to talk like a therapist by default and likely be better. E-cigs never evolve that rapidly, as an analogy

[smcg]

It will know how to better extract profit from you. It's not your friend.

[ryanhecht]

I feel ashamed to admit this in public, but...me too. The barrier to entry is so much lower than therapy, I feel less anxiety about explaining my situation correctly, and I can quickly start over with a "new therapist" whenever I want.

Is it a replacement? No, of course not. But boy if it isn't a big help.

[rednalexa]

The only issue is that it needs a good user. Just as you can work your way into good insights, you can work yourself into bad insights just as easily. Sometimes chatgpt doesn't challenge you enough unless you ask it to under specific frameworks or paradigms. It is definitely a useful tool though.

Edit: These are all problems with real therapy too though, on further reflection. It took some time to find the therapist that works well with you. That can be a form of similar bias.

[kelseyfrog]

I find myself being exceptionally careful when asking questions. ChatGPT is known to lean sycophantic - biased toward agreement. And rather than asking, "Is this ...?" or "Am I ...?", I'm careful to choose prompts like, "Evaluate, or analyze...".

My hope is that this results in less sycophantic responses, but overall it's a difficult thing to measure.

[rs186]

I never used ChatGPT as a therapist, but it was able to identify the specific mental health issues I had, and I used that information to find self-help books just on that topic. Meanwhile, a therapist with whom I had online sessions with really just worked with me on generic anxiety issues.

I tried another episode I had a few years ago, and ChatGPT was able to provide specific and correct advice. What happened back then was that a therapist misdiagnosed my issue and thought I was going to harm other people and alerted the police. That was a horrifying experience. Only later a specialist understood the issue and provide proper care.

I know I need to be very careful about ChatGPT's output, so I do try to understand what it says, and seek professional help when necessary. However, very seriously, in many cases ChatGPT provides better care than a less experienced, apathetic therapist you find online.

[rottc0dd]

Hmm... maybe that is why, earliest chatbot carried a "therapist" tag.

https://web.njit.edu/~ronkowit/eliza.html

[balamatom]

The story goes that people would request to have sessions with it, even if they knew it's just parroting some phrases. They just needed to be heard out!

I feel really sad about how people need to resort to therapy for the kind of interaction that would be provided by a healthy social norm of honest and open communication. Maybe I should talk to my therapist about this feeling. But somehow I doubt that would actually accomplish anything.

This is the first time I see history repeating itself first as farce and only then as tragedy. Personally, I would be hard pressed to trust any data collection business with any honest information about my mental health.

DNA profiling in the 23andme thread, psychological profiling over here, teachers teaching kids to be unable to put 2 and 2 together in the "AI presentation maker for teachers" thread. Scariest of all is how those who would simply prefer to not live in that kind of world are gradually dehumanized by those who find nothing wrong with that sort of thing...

[imtringued]

The problem is that there are a lot of people who are destroying the ability to communicate in the name of "morality", when in reality they are just political opportunists, who are indirectly harming themselves, including their supposed political cause, in the process.

The amount of wilful bridge burning is crazy.

[balamatom]

Also I either became unstuck in time again or my other reply to your post had done a disappear for a good while. Suppose it's the load balancer flaking out under the weight of all the tinfoil I wear as my fashion statement? Guess some weirdness is afoot, I could see it in the tea leaves and the traceroute, too :) https://www.youtube.com/watch?v=NghF9aOlBDY&t=69s

[strogonoff]

Communication with humans makes us feel better. In absence of actual humans, an LLM trained on human works is better than nothing. However, it is worth remembering, in context of mental health, that it is just that: an aggregate of creative output from various humans, with their own mental warts and all, plus the added complication of ML engineers making it conform to your expectations.

[anal_reactor]

I have exactly the opposite experience. I was having exceptionally bad time, I tried talking to ChatGPT, and while the first impression was very good, I quickly realized it... lacks personality? I'm not sure how to phrase it, but the language it used made me feel like I was on a work meeting.

I think this might be boiling down to cultural differences though. I noticed that even though I work in an international company, people from similar cultures tend to cluster together. I guess ChatGPT was trained on texts from a culture that isn't mine, and now it fails to adapt.

[amenhotep]

I tend to keep different sessions open for different ongoing topics, and that's what I notice when I start a new one. It's very easy for it to fall into its default sterile, focus grouped, conversation-flowchart mode.

Longer conversations can draw it more "out of its shell", though. The more context it has, the better it can mirror you, and there's some sessions I have where it genuinely does feel like it's got a little bit of personality and rapport with me.

You really do have to keep in mind, though, that no matter what, in the end it's just going to agree with you :)

[p3rls]

Can you depersonalize a few sentences exchange between you and share? I'm curious to see and describing it doesn't really do it justice especially if you're coming to this doubtful about AI

[raxxorraxor]

It has zero EQ and just uses phrases of learned and presumably fitting expressions for a therapy situation. But don't let that get you on the road to existential dread.

[maCDzP]

I have tried using it as a therapist for issues that are not sensitive and I think it's great. I won't use it for sensitive issues due to privacy concerns. I also have tried to use it as a personal trainer and coach and works great - so far.

[petesergeant]

I use it in conjunction with a therapist. It’s always there, and I find I don’t mask when talking to it like I do to a person, so I can dig a bit deeper and feel a bit more strongly; this makes great input for follow-up sessions with my real therapist.

[donatj]

I see a real world therapist but there are genuinely topics ChatGPT is simply better at.

I also feel like real therapists almost have too much state. They know too much about you to just let you talk. They like to try to tie all your problems back to other issues, and it's like no, my inability to give up childhood toys is not rooted in my crippling fear of nothingness, and you're derailing my line of thought with the suggestion.

I have frankly had some very productive conversations with ChatGPT about things I would be embarrassed to tell another human being. The lack of any real judgement is useful.

[SoftTalker]

Therapists follow a standard patterns or approaches to help you express what you are feeling and then suggestions at how to deal with it. Seems to be the perfect formulaic thing for an LLM to do.

[benfortuna]

ChatGPT is not the only type of AI Chatbot.

The role-playing chatbots like character.ai are more of a problem because people are forming emotional relationships with these "characters".

[2-3-7-43-1807]

how do you use it? i tried it once and quickly lost interest as the conversation turned shallow and schematic. not organic and insightful as i would expect from an actual therapy session.