Timeline for How are the ideas of Karl Weick on sensemaking and George Kelly on the "naive scientist"/Personal Construct Theory regarded currently?
Current License: CC BY-SA 4.0
10 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Apr 18 at 9:43 | comment | added | iSeeker | @BryanKraus I quite agree. Sensible use involves cross checking and not relying on it exclusively. Most informed people know to look out for its errors. And is it significantly worse than many other free information sources that can also mislead? At least one can interrogate C-GPT. [Does P&N have a "Discussion section" like some other SEs in case we both want pursue this?] | |
| Apr 17 at 22:54 | comment | added | Bryan Krause♦ | @iSeeker But if you rely on it and it misleads you, you won't know you were misled. | |
| Apr 17 at 22:38 | comment | added | iSeeker | @BryanKraus What I find useful about Chat-GPT, even though I find errors quite often, is that one can drill down and pursue a question in a way that you cannot do with e.g. Wikipedia, or the medical literature, and can then cross-check with other sources. I use it mostly for STEM subjects in which it acts more like a personal tutor. Its interpretation of a detailed and quite dense radiologist's report (when I'd already been well informed on my basic condition for some years) made good sense and the understanding I gained from it was confirmed by the surgeon. It was helpful. | |
| Apr 17 at 22:22 | comment | added | Bryan Krause♦ | @iSeeker Most of the support of the usefulness of LLMs is exactly the sort that you've provided: anecdotes where they got the right answer or cute examples where they seem to be especially insightful. There are also some embarrassing failures of the same models given similar prompts. The problem is that until you actually got the specialist's recommendation, you wouldn't have known which category you were in, and unless your specific question has been carefully posed to LLMs and studied we wouldn't know exactly which cases it would fail in or how often. | |
| Apr 17 at 19:25 | comment | added | iSeeker | Thanks, I'll try that. [In defence of LLMs, however, I used ChatGPT to interpret the highly technica radiologist's report on an MRI scan, and to explore surgical options, risks and expected recovery, before seeing a UK spine specialist, privately, whose recommendations were virtually identical to the LLM's best recommendation - and I'm now 14 weeks recovered from lumbar decompression. (Same hospital as King Charles and Princess of Wales attended, just after me.)] | |
| Apr 17 at 19:03 | comment | added | Arnon Weinberg♦ | It's not my area of expertise, and it may be that few others here know. Personally though, I'd rather have no answer, than a plausible-sounding answer whose accuracy is practically equivalent to a toss of a coin. If you are in a hurry, then you could try posting a bounty. | |
| Apr 17 at 18:48 | comment | added | iSeeker | @ArnonWeinberg Since no-one here seemed willing to answer I was surprised to get some sort of response from an LLM in Quora. I'm well aware of their limitations and would greatly prefer some refs and other aspects you mention, but the suggestion that Weick's and Kelly's ideas might not be completely outdated/discredited is encouraging - unless you can put it right...? | |
| Apr 17 at 18:14 | comment | added | Arnon Weinberg♦ | Our experience here is that LLMs are good at generating responses that sound plausible, but are rarely accurate or useful. Notice in the response that no sources are provided, statements are vague and meaningless, and there is no discussion of replication, competing theories, refinements, outstanding questions, on-going research, or practical applications. If this was all you wanted, then LLMs are indeed a better choice than SE. | |
| S Apr 17 at 11:04 | review | First answers | |||
| Apr 17 at 18:17 | |||||
| S Apr 17 at 11:04 | history | answered | iSeeker | CC BY-SA 4.0 |