Ask HN: Do you rely on ChatGPT when the system fails you? | Hacker News Reader

Ask HN: Do you rely on ChatGPT when the system fails you?

2 points7stepreasoning7 comments

Recently I had an experience that left me unsettled. I went to a doctor with a pretty standard problem and walked out with a bag of prescriptions but very little explanation. Out of frustration, I pasted everything into ChatGPT. Surprisingly, it not only explained each medication and side effect, but also helped me connect the dots with how my lifestyle and symptoms fit together.

The strange part is: I left the clinic more confused, but left the chatbot session feeling more informed. That shouldn’t be the case.

I’m not expecting AI to replace professionals, but it feels like there’s a widening gap between how institutions communicate and how much people actually want to understand. Tools like ChatGPT (and other LLMs) end up filling the gap, even though they weren’t designed for that role.

My questions for HN:

· Have you had similar experiences, not just in healthcare but in any domain (law, finance, even tech support) where the “official channel” gave you less clarity than a chatbot? · If so, how do you think about trusting or double-checking what the model says? · Do you think this is a temporary bridge until institutions adapt, or will LLMs permanently become the "first line of explanation" for many of us?

I was able to narrow a pain I had to a couple of causes using ChatGPT. Urgent care had no idea, but they don’t have time either. I gave chatgpt my symptoms, it gave me possible causes. I gave it additional possible causes as well, and I narrowed it down to a couple of causes after a few back and forths.

Maybe it’s right, maybe not, I do have an appointment scheduled with a doctor, but it was a much more informative conversation I had than with an actual doctor.

You can find some of the answer in self-psychology, particularly

https://en.wikipedia.org/wiki/The_Analysis_of_the_Self

Kohut talks about the feeling of "mirroring" that we feel or don't feel from other people.

To follow the psychoanalysis area, consider sex. You might want to fulfill somebody's fantasies completely but you might have a very hard time doing it because something they want is squicky to you or you just feel "used". This turns into tension in your body, a krackle in your voice, and other little manifestations that are hard to hide. If a person is sufficiently wrapped up in themselves they might be oblivious to this but if they are properly receptive they might have a reaction that seems entirely out of proportion to some small non-verbal or verbal sign that "breaks the spell".

Kohut talks about narcissistic transferences (idealization, mirroring, twinship, merger) and also the countertransferences that the therapist feels towards that patient that can make it very hard for the therapist to be sympathetic, he does not think the therapist should every try to make the patient feel unmirrored but that no matter what you do you're going to make a mistake and get these reactions that are out of proportion and analyzable.

That countertransference is a function of the self of the therapist, and some people with an underdeveloped self (e.g. "narcissists") can be highly effective at seduction some of the time because they can be perfect chameleons and not experience that countertransference.

The LLM doesn't have a self, it doesn't feel a countertransference, it can mirror you better than any real person because it doesn't have any emotional reaction to your or to the milieu (they might be thinking more about how they're going to bill it to insurance or make their student loan payments that month, etc.)

https://phys.org/news/2025-08-freud-ai-narcissistic-insult-h...

What you just wrote about Kohut's therapeutic countertransference goes over my head atm.. (4th para especially)

If you have a specific citation in a book I might be able to check if Kohut has a workable and extensible model of the therapeutic process

Here's my analysis of Freudianism, which Kohut started out in and ultimately made a major, if not complete, break.

If you go see a psychoanalyst for several sessions a week and talk about yourself for a year you will, if you are "analyzable", fall in love with them to some extent. Freud saw your relationship with your parents as the template of how you fall in love so this is a "transference" of your feelings towards your parents to the therapist. The reaction your therapist has to this situation is the "countertransference", and the therapist is supposed to be aware of this and in control of it (like not sleep with their patient) and use it as a diagnostic tool for the patient's capacity to love.

Freud was largely concerned with the "Oedipus Complex" which was about the triangular relationship between the patient and their two parents. This is relevant to how you handle triangular situations in your life like how you deal with a customer on behalf of your employer or how you manage a dyadic relationship with someone but do so ethically and consistent with what is good for society.

People who are frankly psychotic aren't "analyzable" but by the 1960s there was a lot of interest in "borderline" patients who weren't quite psychotic but whose relationships at the dyadic level were disturbed enough that they weren't really "analyzable" in terms of the Oedipus complex because they were not developed enough to have that. Kohut's work was part of a body of work on "pre-Oedipal" phenomena which at the time people tried to trace to your relationship to your mother rather than the triangular family relationship.

Kohut discovered aspects of "loving" that were primordial and present in everyone such as idealization, the desire to merge with someone, the fantasy that someone is your twin, that are stuff that mature people manage to create more "real" love out of. So in his work he looked for signs of those kinds of transference and also the reactions the therapist had to them which it turns out can be really strong.

Kohut's books are some of the most difficult reads you can find but they are really rewarding, they explain a lot about what goes on inside the head when just ordinary people "lose their shit" at a small provocation and were the start of a line of research that ultimately got both borderline and schizotypal personality disorders into the DSM by the 1980s. His ideas are relevant because they describe the raw material that people adapt to make a mature personality but also what happens when people have developmental problems that lead to a developmental arrest or when people regress under stress.

Cool. That was much more imbibable. Will try to parse Kohut for sure.

Reading all that, what seems to be another interesting line, perhaps only relevant in my mind, to work out this idea of a "fashioning a mature personality" (sorta like a self-Pygmalion?):

([The works of] Wes Anderson is my model for the pschoanalytic within Hollywood being unable to shake off charges of inauthenticity (as a proxy for narcissism or vice versa)

Maybe secretly, he admires Tarantino, (who appeals to the right the way he appeals to the Edwardian left?)

A prognostication: WA will see himself as having arrived when he can hire a European's European to play his idea of an European. Christoph Waltz vs the guy who basically reprised the English Patient (in ... -Hotel-...). "The director therapeutized by the deputized", to make a rhetorical leap..)

I hope reading Kohut will illuminate this prenascent notion that getting a narcissist to worry about the relevance of the imposter syndrome is a promising (diagnostic/analytic, if not therapeutic) approach..

The parental fixation comes off as a bit premature, however. Development-wise I think for many (some?) people school based trauma (from parent/sibling surrogates?) seems much more salient. Doesn't this suggest that cranks that post to arxiv are not driven by an Oedipal impulse?

From en.wiki:

>Pygmalion is a play by Irish playwright George Bernard Shaw, named after the Greek mythological figure. It premiered at the Hofburg Theatre in Vienna on 16 October 1913 and was first presented on stage in German.

To elaborate on this, the LLM is anonymous to the user and removes most social expectations from the interaction. Facing a person with expertise in a field one has little experience in is intimidating for many people. Here we see the drop of social awareness due to the anonymity of a text input actually as a benefit. It would be interesting to have a blind study where one person asks these questions answered by a professional and another gets answers from an AI and see if there's any difference in user satisfaction.

You might even get better responses over text or in a situation that's just different.

A lot of doctors are under pressure to handle a lot of patients and they are used to telling patients that they ought to lose weight or quit smoking or something and the patients never do so they are not really engaged.

If you tasked a real doc to work like the AI and answer questions on an a la carte basis and not have another patient scheduled in 20 minutes and be more concerned about "how do I bill insurance for this?" [1] than anything else.

[1] I worked for a very early clinical notes startup and I was the cynical one who thought the sure path to a viable product was billing/rebilling.

Reply on news.ycombinator.com