Talk:Epilepsy

Article (edit | visual edit | history) · Article talk (edit | history) · Watch • Watch article reassessment page • Most recent review

Result: No consensus. ~~ AirshipJungleman29 (talk) 21:31, 23 October 2025 (UTC)

This was listed as a GA in 2014. Unsurprisingly, a lot of changes have been made since then, and unfortunately a lot of the recent ones introduce problems that -- while I don't have proof -- are very characteristic of LLM use. While there is no policy against AI-generated text, the issues here are specifically the ones like editorializing, undue emphasis on symbolism, superficial analysis, etc., that imply inherent WP:OR and WP:SYNTH problems that violate Good Article criterion #2. I'm not reviewing every diff -- and can't, because a giant swath of edits from 2021-2024 has been revdelled -- but here are some with issues.

• The edits by @Pwoli show consistent issues. To name a few: This one inserts a large wad of unreferenced, editorializing AI slop under the "Society and culture" section. This states that animal models play a central role in epilepsy research, an assertion of importance over other kinds of research that doesn't seem to be borne out by the source or in general. This inserts a lot of unsourced material like the assertion, with no citation, that delays in diagnosis are highlighting the importance of careful clinical assessment and appropriate use of EEG and video documentation, which may well be true but is still original research.
• The 2024 edits by Saraakwong also show signs of AI use. Some of these were reverted in the edits above (which is for the best since I found a hallucination in one of them), but at least one, this, may be an overly close paraphrase of the source material, judgment call.
• In general I see several uncited statements ending paragraphs. I'm not tracking down the origin because, again, it's entirely possible that I can't.

— Preceding unsigned comment added by Gnomingstuff (talk • contribs) 19:04, 9 September 2025 (UTC)

A couple of comments:

• Animal models are hugely important in drug research. I think that "central role" is a fair and neutral description in this case. The cited source (which, if you would like, may be fairly criticized on WP:MEDDATE grounds, being ~12.5 years old) uses words like "valuable" and "urgent" and "needed" to describe animal models. If your personal priority for epilepsy is more like prevention or figuring out how many people have it, then I could imagine animal models seeming less important to you, but within the biomedical research community, this really is of central importance. No new AEDs should realistically be expected until there are useful new animal models.
• What you deride as a large wad of unreferenced, editorializing AI slop under the "Society and culture" section appears to me to be an attempt to comply with the WP:UPFRONT guideline, by giving a general summary of the long section. For example, the second sentence is People with epilepsy may experience social stigma, legal restrictions, economic disadvantage, and barriers to education and employment., and the next three subsections in this section are: "Stigma", "Economic impact", "Driving and legal restrictions". The fourth sentence says Efforts by advocacy groups and international organizations aim to improve public understanding, reduce stigma, and promote access to care and the fourth subsection is "Advocacy and support organizations". I doubt that this is best explained by assuming LLM use.
• In terms of "editorializing", I don't see anything in that paragraph that isn't just the plain facts. The plain facts are that people with epilepsy sometimes do experience stigma. They do get fired from their jobs. They do get kicked out of school (maybe not in your country, but in many). They do get rejected socially. Editorializing sounds like "Society should be more supportive of people with epilepsy". Editorializing does not sound like "Social consequences, such as educational exclusion, unemployment, and social isolation, further compound the impact on quality of life". The sentence in the article is just stating the facts. There are social consequences, and those social consequences hurt people with epilepsy.

I do think this article needs work, but I don't think that "AI slop" is actually one of its faults. WhatamIdoing (talk) 18:57, 11 September 2025 (UTC)

Leaving aside that the entire first paragraph of this is redundant, vague, and unsourced, and that the treatment gap is mentioned twice and only defined the second time, by editorializing I mean stuff like the treatment gap underscoring the need for strengthened health systems and public health interventions, which is written from an advocacy perspective. Stating that a specific societal change is a "need" is an opinion, not a fact. Other people may have different opinions about what "the need" is: loosened regulations on pharmaceutical companies, or decoupling health insurance from employment, or research into alternative medicine, or abolishing capitalism entirely, or patiently staying the course with no policy change, or even that the treatment gaps is not a problem so no change is necessary. The same goes for cost-effective treatment strategies -- if someone can't afford treatment then by definition it's not "cost-effective" for them. Gnomingstuff (talk) 23:11, 11 September 2025 (UTC)

@Gnomingstuff Seeing as Whatamidoing has already debunked the editorializing claim, can you expand on undue emphasis on symbolism, superficial analysis, etc., that imply inherent WP:OR and WP:SYNTH problems that violate Good Article criterion #2. Specifically can you identify the the cases of OR and SYNTH that led you to open the reassessment. I'm not sure how to help here when you haven't pointed out the inherent OR/SYNTH. IntentionallyDense ^(Contribs) 08:27, 14 September 2025 (UTC)

I literally just explained it. I don't know how else I am supposed to get across, for instance, the fact that stating something is a "need" is an opinion. Gnomingstuff (talk) 03:48, 15 September 2025 (UTC)

Okay I can see what you mean by the wording being odd (as Whatamidoing pointed out not all the odd wording is editorialization). I think the other odd wording may come from the fact that medical articles do say things like “cost effective treatment strategies” and “the need for…” and many people not well versed in medical talk then use those phrases, not understanding that the audience of a research paper is different than the audience of a wikipedia article.

Please correct me if i’m wrong here as I’m just trying to understand the issue at hand, but is your issue the odd wording (i haven’t looked at it enough to call it editorialization)? because if that’s your main concern then that’s a pretty easy fix and i can get started on that.

However you also mentioned that there is WP:OR and WP:SYNTH which you still haven’t shown evidence of. To make an claim of OR or SYNTH i’m assuming you looked at the sources and found that they did not support the claims being made, it would be helpful if you could just point out which sources and areas of text were the ones that led you to the conclusions of OR/SYNTH. I’m not asking for every instance of OR/SYNTH that you found, just whatever one led you to making that claim. IntentionallyDense ^(Contribs) 07:46, 15 September 2025 (UTC)

@Gnomingstuff could i please get an update on this? I don’t want to get too into cleaning up the sources without knowing which ones had inappropriate SYNTH or OR attached to them. IntentionallyDense ^(Contribs) 21:28, 17 September 2025 (UTC)

I'm really sorry, as English is not my main language, I have used LLM in order to translate the things. If this had been AI slop, I would have simply copied and pasted directly from a language model. Instead, I used it as a tool to assist with rewriting, and I reviewed and adjusted the text myself. At no point did I copy and paste output directly. While I cannot provide proof of this process, I am being transparent in stating that I used the tool solely to support rewriting. Given that Whatamidoing has already debunked the claims, could we please remove the notice? You will probably think I used AI for this too, but I am just autistic and verbose. Sorry again, I am very ashamed for this whole situation. Pwoli (talk) 01:05, 15 September 2025 (UTC)

I never said that you needed to be ashamed of anything, I was just citing the specific diffs that I was referring to. Gnomingstuff (talk) 03:52, 15 September 2025 (UTC)

Pwoli didn’t ever claim that you said they needed to be ashamed either. @Pwoli seems like they were just trying to help build the encyclopedia and because of english not being their main language and maybe not realizing they needed to adjust their language for Wikipedia may have used some wording that is indicative of AI. However just because something sounds like it could be AI doesn’t mean it is (which i know you know)and since we have now gotten confirmation that they did not copy and paste from AI I am going to go ahead and remove the banner until there is some evidence of AI usage. IntentionallyDense ^(Contribs) 07:51, 15 September 2025 (UTC)