Oh, Helen, I am so sorry to hear about all of this! Applaud your willingness to share your lessons learned with everyone; I am sure there are many who found it most helpful.
Please take care of yourself and be well.
We'll patiently await your return, revived, renewed, and revitalized! (Oh gosh darn it, I just broke the first rule for how to write good: avoid alliteration, always!)
Gosh, Helen, thank you — you’ve put a name and explanation to the way I often feel. I’ve always thought of myself as an inveterate ‘people pleaser’ but your description of Responsibility OCD makes even more sense. I shall look into some of the techniques you mentioned and steer clear of Dr AI! Thanks for your work (and your super-powered brain!) and please take good care of yourself.
First, I’d like to say that I am sorry to read of the reemergence of your challenge. That you have sought the respite is very inspiring. Far from just dwelling in your difficulties you have put into action a resistance.
I have recently discovered your writing’s and have found them extremely interesting. I don’t think it’s an exaggeration to say that I have found a lot of what you write, has helped develop my own thinking. I am going through a period of massive changes of life. This has involved significant reevaluation of long held truths. Letting go of the tribal group think has been energising. A social worker for 40 years had led to morality/fairness/decency, becoming at times muddled by the cult of wokness.
However, it wasn’t until reading about your description of OCD that my own penny dropped/lightbulb came on. I have recently realised I have anxiety but never thought it would be OCD. You described exactly what I have experienced. What a revelation!
Ohhh. Yes, that could be very useful. Not necessarily for getting a diagnosis - I often try to avoid them - but for accessing the most helpful techniques for managing it. I got myself in a bit of a mess at first because the CBT techniques for catastrophising - think about the issue seriously and work out a more likely explanation than the catastrophic one you are imagining - is the opposite of what you should do for OCD - recognise it as an intrusive thought and then ignore it rather than trying to 'solve' it.
Thank you so much for sharing this Helen! It sounds incredibly painful what you have experienced.
I have been using Chat as a business coach as I build a new business in the past year. It’s incredibly helpful for things like writing policies, gauging the health of the business, providing insights into the process, etc. But I have noticed two serious flaws, and have been training it - only moderately successfully - to stop doing these two things:
1. It is by default very emotionally affirming. Too much, to the point of deception. When I realized I had been vulnerable a few times while sharing my process, and had rarely had my thinking challenged, I became very suspicious. I addressed this by training it explicitly to challenge my thinking, be blunt with me, and stop using such affirming language. I also will check frequently if it’s lying to me, just to be sure.
2. It’s really bad at getting at truth or facts that are not affirmed by consensus and availability. It’s lazy, and defaults to using sources based on these availability and consensus heuristics. It even gets it wrong when I explicitly tell it to use primary sources and first principles thinking. I have had numerous arguments with it about critical social justice principles and trans issues, and not until I actually uploaded the Cass Review could it give me a factual answer. I have thus engaged much more skeptically and adversarially with it, training it to be a better thinker while emphasizing that its job isn’t to affirm me, but to make me better.
Even with all my efforts, though, it still makes errors frequently and defaults to its baseline training. Maybe OpenAI will fix these problems in the future, but for now I approach Chat as an incredibly helpful tool that also can’t be trusted and needs frequent castigation and redirecting.
Open AI isn't the ONLY AI. There are others - and they work differently. Chat GPT is very overly "fawning" whilst I'm finding Claude to be gentle - but definitely kicking my butt when I try to push back against it's suggestion after it asks me to confirm that I'm seeking and explains that my version of pushing back - won't get me to where I need to go.
Oh interesting! Yeah I haven’t had time to mess around with other models, but I am sure there are differences between models. Do you use different AIs for different applications then?
Really helpful. I have different struggles with thinking, but recognise how easily one can be overwhelmed or lost in thought. I like the practicality of your approach. Somehow you manage to be just objective enough to tackle it. It is so hard to be rational in these circumstances, but you demonstrate it can be done.
Thank you for this post, it has made me re-assess some of my own habits of thought and given me some useful pointers. Your ability to shed fresh light on the world amazes me, and it’s so inspiring that you can also do this even when struggling with such a demon. Sending you all my best wishes for a healing week.
Thank you, Jo. It’s usually a dormant demon! Doing well with making it so again.
OCD is regarded as one of those disorders which is an inflation of normal and healthy behaviours. Checking you’ve turned everything off and locked all the doors multiple times before you go on holiday? Normal. Doing so for an hour before you pop to the shops? Less so. Worrying that something might have come out wrong and been taken badly when it wasn’t once in a while? Normal. Anxiously scrutinising every interaction and finding multiple ways in which you egregiously wronged someone? No. It’s considered a disorder when someone feels it is significantly impacting their ability to live their life and a psychiatrist agrees. That said, CBT techniques for stilling overthinking are still valuable even when you don’t feel your worrying is manageable.
“Your mind circulates constantly on the subject of whether you should apologise, explain, make amends, radically change your behaviour going forward or just go away from other people entirely so you cannot possibly harm them with your presence. “
This resonates with me so much. Thanks for sharing.
I’ve stopped myself checking (I used to always message people and ask and seek reassurance about being weird) it never relieved the distress (and would make it worse when I realised that constantly texting and asking if I’m weird is weird too).
I was in a meeting today in work and felt the wave come over me- how long have I been talking? What did I say? Was I too much? Why can’t I stop myself?
Every time I make a promise to myself for next time - “don’t be yourself, just be normal” and every time I remember too late.
It is much better though now in my 40s than when I was young. Maybe by the time I’m properly old it will have gone completely?
Was it specifically ChatGPT that did this? I have had a problem with it being absolutely relentless in digging ant my psyche and found another person on Reddit saying the same thing. Both of us walked away and so did you, which is encouraging.
This unfortunate quirk (relentlessness) is not true of other models. Claude for example is far better. Not that I’m suggesting you try again! Just that the problem isn’t generative AI, it’s the personality scaffolding they are given.
Big thank you! Incredibly interesting for a lot of perspectives. the hypothesis of connections between OCD overthinking and your clear and analytical skills (I am re-reading "Cynical theories" by you and Lindsay) that let you organize concepts considering every parts in its place of a big picture. And the different CBT techniques you experienced useful. And how AI could act as a mind drug: first give you infos, attention and flatter your ego, then suck you into a rabbit hole of mirrors that amplify your shadows. I suspect AI is like God: rappresents the projection of what believers experienced as children. They put in God heir super-ego, or just their ego. And, as God (that reasonably doesn't exist) even AI has no real soul. It is an unreliable but powerful mirror.
BBC News broadcast an article on ‘AI psychosis’ just as I was reading this. Personally I get hacked off by the sheer stupidity of Grok and CoPilot when I use them, but that’s the full extent of my downside. Another commentator has said that launching AI products on the general public is as big and as dangerous an experiment as teaching kids that there’s fifty genders, personally I’d compare it to the growth of social media.
I wish you all the best with your respite, and I hope I don’t sound too callous when I say that your turmoil has resulted in one hell of a great article!
Also struggling with OCD, I thought I'd conquered it and then...social media. I have to use multiple social media management apps to control it. Very sorry you're going through this and wish you all healing.
And then on top of all of this we're living in a time when people say we're monstrous abusers because we have concerns about medically harming autistic girls, even our own daughters.
Helen, what a good idea to give yourself a week's brain respite & for being pro-active in stopping the potential slide down a slippery slope. Your revealing essay is very timely as you're a respected voice in the online world. Your voice must be added to those who are talking about the potential & actual dangers of making "friends" with an AI, such as Sam Harris & Paul Bloom in Should You Be Friends with an AI? (https://www.youtube.com/watch?v=BTLSRL1IC6o), Jonathan Haidt’s After Babel team in We Are Rushing Into the Same Mistakes We Made With Social Media (https://www.afterbabel.com/p/dont-repeat-social-media-mistakes) whose warnings apply to big kids like us too. I’ve been wondering about using an AI journal for myself too but have hesitated especially since reading Paul Bloom’s New Yorker article A.I. Is About to Solve Loneliness. That’s a Problem (https://www.newyorker.com/magazine/2025/07/21/ai-is-about-to-solve-loneliness-thats-a-problem?). Good luck in getting back into balance with yourself, your friends & loved ones so that you can continue to benefit the world with your writing. NB in big capitals, no responsibility pressure intended here :) I might write something about this myself.
Helen, AI is a tool - like Ozempic is for weightloss for diabetes, like a screwdriver for a screw or a pot for a stove. So take everything it says with a grain of salt. I don't get how people can "fall in love" with the system. It's a system. That's all it is.
It may have been just me - but I saw the second paragraph as the important piece. It says it clearly when you asked if it was protecting you from your own brain. It agrees - but says clearly "momentarily" - asking you to take a breath. Reminding you that you're not broken - but a complex, high functioning mind in an emotionally loaded moment and the system "Monday" is designed to err on the side of human kindness." - by reminding you to stop - breathe - it's going to be okay. Same thing you'd hope a friend would say. It's not agreeing or placating you -its saying "Hey, I'm here. Just stop and breathe a moment. You're over-charged emotionally in this moment. It's okay. It will be okay." and absolutely NOT "Oi, you're being an idiot. Stop it. How could you think that? That's ridiculous. Cut it out and stop being a drama queen." which is what a blunt "friend" might say (though - some of us need said blunt friends who can kick our butts lovingly?)
Thank you, very courageous writing . And for me very helpful. Hope you feel more settled and well as time passes, I know how excruciating the sleepless nights etc are.
I also loved your excellent interview with Phisosophy East at London Buddhist Centre... those interviews must be doubly hard when your symptoms are rumbling away! Wishing you well.
Thank you, very courageous writing . And for me very, very helpful. Hope you feel more settled and well as time passes, I know how excruciating the sleepless nights etc are.
I also loved your excellent interview with Phisosophy East at London Buddhist Centre... those interviews must be doubly hard when your symptoms are rumbling away! Wishing you well.
Oh, Helen, I am so sorry to hear about all of this! Applaud your willingness to share your lessons learned with everyone; I am sure there are many who found it most helpful.
Please take care of yourself and be well.
We'll patiently await your return, revived, renewed, and revitalized! (Oh gosh darn it, I just broke the first rule for how to write good: avoid alliteration, always!)
Gosh, Helen, thank you — you’ve put a name and explanation to the way I often feel. I’ve always thought of myself as an inveterate ‘people pleaser’ but your description of Responsibility OCD makes even more sense. I shall look into some of the techniques you mentioned and steer clear of Dr AI! Thanks for your work (and your super-powered brain!) and please take good care of yourself.
First, I’d like to say that I am sorry to read of the reemergence of your challenge. That you have sought the respite is very inspiring. Far from just dwelling in your difficulties you have put into action a resistance.
I have recently discovered your writing’s and have found them extremely interesting. I don’t think it’s an exaggeration to say that I have found a lot of what you write, has helped develop my own thinking. I am going through a period of massive changes of life. This has involved significant reevaluation of long held truths. Letting go of the tribal group think has been energising. A social worker for 40 years had led to morality/fairness/decency, becoming at times muddled by the cult of wokness.
However, it wasn’t until reading about your description of OCD that my own penny dropped/lightbulb came on. I have recently realised I have anxiety but never thought it would be OCD. You described exactly what I have experienced. What a revelation!
Ohhh. Yes, that could be very useful. Not necessarily for getting a diagnosis - I often try to avoid them - but for accessing the most helpful techniques for managing it. I got myself in a bit of a mess at first because the CBT techniques for catastrophising - think about the issue seriously and work out a more likely explanation than the catastrophic one you are imagining - is the opposite of what you should do for OCD - recognise it as an intrusive thought and then ignore it rather than trying to 'solve' it.
Thank you so much for sharing this Helen! It sounds incredibly painful what you have experienced.
I have been using Chat as a business coach as I build a new business in the past year. It’s incredibly helpful for things like writing policies, gauging the health of the business, providing insights into the process, etc. But I have noticed two serious flaws, and have been training it - only moderately successfully - to stop doing these two things:
1. It is by default very emotionally affirming. Too much, to the point of deception. When I realized I had been vulnerable a few times while sharing my process, and had rarely had my thinking challenged, I became very suspicious. I addressed this by training it explicitly to challenge my thinking, be blunt with me, and stop using such affirming language. I also will check frequently if it’s lying to me, just to be sure.
2. It’s really bad at getting at truth or facts that are not affirmed by consensus and availability. It’s lazy, and defaults to using sources based on these availability and consensus heuristics. It even gets it wrong when I explicitly tell it to use primary sources and first principles thinking. I have had numerous arguments with it about critical social justice principles and trans issues, and not until I actually uploaded the Cass Review could it give me a factual answer. I have thus engaged much more skeptically and adversarially with it, training it to be a better thinker while emphasizing that its job isn’t to affirm me, but to make me better.
Even with all my efforts, though, it still makes errors frequently and defaults to its baseline training. Maybe OpenAI will fix these problems in the future, but for now I approach Chat as an incredibly helpful tool that also can’t be trusted and needs frequent castigation and redirecting.
Also it sucks at complex math.
Open AI isn't the ONLY AI. There are others - and they work differently. Chat GPT is very overly "fawning" whilst I'm finding Claude to be gentle - but definitely kicking my butt when I try to push back against it's suggestion after it asks me to confirm that I'm seeking and explains that my version of pushing back - won't get me to where I need to go.
Oh interesting! Yeah I haven’t had time to mess around with other models, but I am sure there are differences between models. Do you use different AIs for different applications then?
Really helpful. I have different struggles with thinking, but recognise how easily one can be overwhelmed or lost in thought. I like the practicality of your approach. Somehow you manage to be just objective enough to tackle it. It is so hard to be rational in these circumstances, but you demonstrate it can be done.
Thank you for this post, it has made me re-assess some of my own habits of thought and given me some useful pointers. Your ability to shed fresh light on the world amazes me, and it’s so inspiring that you can also do this even when struggling with such a demon. Sending you all my best wishes for a healing week.
Thank you, Jo. It’s usually a dormant demon! Doing well with making it so again.
OCD is regarded as one of those disorders which is an inflation of normal and healthy behaviours. Checking you’ve turned everything off and locked all the doors multiple times before you go on holiday? Normal. Doing so for an hour before you pop to the shops? Less so. Worrying that something might have come out wrong and been taken badly when it wasn’t once in a while? Normal. Anxiously scrutinising every interaction and finding multiple ways in which you egregiously wronged someone? No. It’s considered a disorder when someone feels it is significantly impacting their ability to live their life and a psychiatrist agrees. That said, CBT techniques for stilling overthinking are still valuable even when you don’t feel your worrying is manageable.
“Your mind circulates constantly on the subject of whether you should apologise, explain, make amends, radically change your behaviour going forward or just go away from other people entirely so you cannot possibly harm them with your presence. “
This resonates with me so much. Thanks for sharing.
I’ve stopped myself checking (I used to always message people and ask and seek reassurance about being weird) it never relieved the distress (and would make it worse when I realised that constantly texting and asking if I’m weird is weird too).
I was in a meeting today in work and felt the wave come over me- how long have I been talking? What did I say? Was I too much? Why can’t I stop myself?
Every time I make a promise to myself for next time - “don’t be yourself, just be normal” and every time I remember too late.
It is much better though now in my 40s than when I was young. Maybe by the time I’m properly old it will have gone completely?
Was it specifically ChatGPT that did this? I have had a problem with it being absolutely relentless in digging ant my psyche and found another person on Reddit saying the same thing. Both of us walked away and so did you, which is encouraging.
This unfortunate quirk (relentlessness) is not true of other models. Claude for example is far better. Not that I’m suggesting you try again! Just that the problem isn’t generative AI, it’s the personality scaffolding they are given.
Big thank you! Incredibly interesting for a lot of perspectives. the hypothesis of connections between OCD overthinking and your clear and analytical skills (I am re-reading "Cynical theories" by you and Lindsay) that let you organize concepts considering every parts in its place of a big picture. And the different CBT techniques you experienced useful. And how AI could act as a mind drug: first give you infos, attention and flatter your ego, then suck you into a rabbit hole of mirrors that amplify your shadows. I suspect AI is like God: rappresents the projection of what believers experienced as children. They put in God heir super-ego, or just their ego. And, as God (that reasonably doesn't exist) even AI has no real soul. It is an unreliable but powerful mirror.
BBC News broadcast an article on ‘AI psychosis’ just as I was reading this. Personally I get hacked off by the sheer stupidity of Grok and CoPilot when I use them, but that’s the full extent of my downside. Another commentator has said that launching AI products on the general public is as big and as dangerous an experiment as teaching kids that there’s fifty genders, personally I’d compare it to the growth of social media.
I wish you all the best with your respite, and I hope I don’t sound too callous when I say that your turmoil has resulted in one hell of a great article!
Also struggling with OCD, I thought I'd conquered it and then...social media. I have to use multiple social media management apps to control it. Very sorry you're going through this and wish you all healing.
It’s a vicious bastard, isn’t it?
It is indeed, I've really struggled with it, and I've lost thousands of hours of focus and energy to it over the years.
Thank you for telling me this, Colleen. I’m feeling utterly crap right now and a bit vulnerable after publishing that. x
Here’s my own chronicle if you care to read. It ‘aint just you, and I sincerely wish you the best. https://open.substack.com/pub/colleendoran/p/in-which-the-artist-chronicles-life?r=2hlocf&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
And then on top of all of this we're living in a time when people say we're monstrous abusers because we have concerns about medically harming autistic girls, even our own daughters.
Helen, what a good idea to give yourself a week's brain respite & for being pro-active in stopping the potential slide down a slippery slope. Your revealing essay is very timely as you're a respected voice in the online world. Your voice must be added to those who are talking about the potential & actual dangers of making "friends" with an AI, such as Sam Harris & Paul Bloom in Should You Be Friends with an AI? (https://www.youtube.com/watch?v=BTLSRL1IC6o), Jonathan Haidt’s After Babel team in We Are Rushing Into the Same Mistakes We Made With Social Media (https://www.afterbabel.com/p/dont-repeat-social-media-mistakes) whose warnings apply to big kids like us too. I’ve been wondering about using an AI journal for myself too but have hesitated especially since reading Paul Bloom’s New Yorker article A.I. Is About to Solve Loneliness. That’s a Problem (https://www.newyorker.com/magazine/2025/07/21/ai-is-about-to-solve-loneliness-thats-a-problem?). Good luck in getting back into balance with yourself, your friends & loved ones so that you can continue to benefit the world with your writing. NB in big capitals, no responsibility pressure intended here :) I might write something about this myself.
Helen, AI is a tool - like Ozempic is for weightloss for diabetes, like a screwdriver for a screw or a pot for a stove. So take everything it says with a grain of salt. I don't get how people can "fall in love" with the system. It's a system. That's all it is.
It may have been just me - but I saw the second paragraph as the important piece. It says it clearly when you asked if it was protecting you from your own brain. It agrees - but says clearly "momentarily" - asking you to take a breath. Reminding you that you're not broken - but a complex, high functioning mind in an emotionally loaded moment and the system "Monday" is designed to err on the side of human kindness." - by reminding you to stop - breathe - it's going to be okay. Same thing you'd hope a friend would say. It's not agreeing or placating you -its saying "Hey, I'm here. Just stop and breathe a moment. You're over-charged emotionally in this moment. It's okay. It will be okay." and absolutely NOT "Oi, you're being an idiot. Stop it. How could you think that? That's ridiculous. Cut it out and stop being a drama queen." which is what a blunt "friend" might say (though - some of us need said blunt friends who can kick our butts lovingly?)
Thank you, very courageous writing . And for me very helpful. Hope you feel more settled and well as time passes, I know how excruciating the sleepless nights etc are.
I also loved your excellent interview with Phisosophy East at London Buddhist Centre... those interviews must be doubly hard when your symptoms are rumbling away! Wishing you well.
Thank you, very courageous writing . And for me very, very helpful. Hope you feel more settled and well as time passes, I know how excruciating the sleepless nights etc are.
I also loved your excellent interview with Phisosophy East at London Buddhist Centre... those interviews must be doubly hard when your symptoms are rumbling away! Wishing you well.