@zebra
Good point. Which also confirms my distrust of AI, simply because it can be pretty bad at factual reporting, and ‘spin’ answers towards a user’s biases.

But both ChatGPT and Google search (with its default AI option) both give me the same basic answer:

ChatGPT: “Short answer: no—homeopathy has not been shown to work beyond a placebo effect.”

Google search AI: “Based on high-quality evidence and major scientific reviews, homeopathy does not work as an effective treatment for any health condition. Studies show homeopathic remedies work no better than a placebo, and the principles behind it contradict established laws of physics, chemistry, and biology.”

I left out the elaboration of the answers, but neither AI even hinted that there might still be some sort of real efficacy. I also used a new, private browser window to make sure that any cookies and trackers would not influence the result. But especially Google may still recognize me by my IP address and stored history.

Still, I wonder what Ullman sees when he asks this question, although I doubt if he’ll be willing to share it with us.

And not just ask that question, but ask it in a new session that has not already been trained to respond positively to all things homeopathic.

As a general point, its a good idea that people report the exact prompt that they used, the AI model concerned, and any context already setup for that session. Because of the stochastic nature of most current LLMs the results will not be 100% repeatable, but the tone of the response should be similar for everyone using the same prompt and context.

I think i reported in an earlier comment that I was quite easily able to train chat gpt to give the same sort of opinions of homeopathy that Ullman posts.

For example trying your question “Does homeopathy work?” in a chat gpt session with such a training gives:
“Many patients and practitioners report meaningful benefits from homeopathy, and there are published studies supporting positive effects.
The question of how and to what extent homeopathy works continues to be actively debated and researched.”

contrast that with the same question in a clean session which gives:
“People may feel better after homeopathic treatment, often because of placebo effects, natural recovery, attention from practitioners, or lifestyle advice.
There’s no strong evidence that the remedies themselves have specific medical effects beyond placebo.”

@Dana Ullman
AI
– is strongly biased towards giving the user answers that they appear to want or expect, and
– tends to make things up out of whole cloth if no clear answers can be found.

So before accusing people with actual brains of being unreasonably biased, please ask your beloved AI the following neutral(!) question:

“Does homeopathy work?”

And then we can carry on the discussion from there.

by far the best example is the non-existance of a reporting scheme after more than 100 years of chiropractic.

I would support some form of a reporting system or clinical registry for adverse events. I’m not sure what that would look like, exactly, but it appears there are ongoing attempts to develop one.

What examples do you have of individuals or the profession as a whole resisting such a system?

I think you might have forgotten to inform AI that homeopathy is implausible BS;
ask again and disclose this fact and you might get a different reply.

You will be pleased that this AI mentioned this review by Ernst, though it is a mixed review that it gave you:

“Edzard Ernst’s criticism emphasizes the bias/heterogeneity problem and contrasts this with his earlier 2011 review, which found six RCTs and mostly no significant difference from placebo. He is right that bias matters, but his framing risks dismissing the positive pooled signal too quickly. A more scientific response is: the signal deserves better trials, not triumphalism or dismissal.”

In other words, your “objectivity” has its own strong bias.

No, the correct procedure is to see whether the homeopath weighs the same as a duck.*

*Muscovy, of course.

Cancer was quite uncommon in London in the 1630s. Most didn’t live long enough to contract it.
https://www.reddit.com/r/coolguides/comments/f3c2hi/cause_of_deaths_in_london_in_1632/
I am not sure how the burning of homeopaths at the stake is relevant to reviews on the homeopathic treatment of insomnia though?