Frightening stuff. I’ve noticed that and no longer visit Reddit or use Wikipedia. The bias is apparent. What solutions? How can we make change aside from not using it? When we use AI can we ask it not to use Wikipedia as a source or is it’s training too deep into Wiki already for that not to make a difference?
You're just going to have to do more research to understand the world, and get a bunch of different perspectives, in some cases probably have to average them out to understand what's really happening. Gotta fact check your AI answers, too
Honestly, Wikipedia editors are a self-selecting sample. People with that much motivation to do voluntary work for what they consider a public service are gonna skew left, they just are. You complain that they control things becuase they're the ones that show up - but, *you* didn't show up! You and I and anyone else who uses Wikipedia to find information is freeriding on their efforts.
Ultimately they dedicate their free time to maintaining a massive and gloriously non-commercial website. Honestly complaints of political bias seem like looking a gift horse in the mouth. The bias undoubtedly exists, but the counterfactual here is not a free encyclopedia that is also the embodiment of political neutrality; the alternative is information locked behind paywalls, Elsevier-style.
Just want to point out I have tried to update Wikipedia articles on several occasions, and 100% of my edits, which were factual and appropriately sourced following the rules were immediately reverted. It is definitely a situation not anyone can edit now, despite what the branding is. It's very centralized and controlled
1. You cited a study by the National Review on how Wikipedia has bias?
2. That quote about the truth was more nuanced than you are considering. Many of these topics that you cite don't have an objective truth. There is no specific, reference-able primary source. These are topics where opinions are widely diverse, and represent the spectrum of political positions. That's what she was saying.
3. The bureaucracy of Wikipedia has existed since the beginning, and has been studied in depth. There is no neglect, this is how it's always worked. You use an anecdote about whether a person is notable, but that has been happening since the very beginning of Wikipedia. That's not new, and there is not some kind of new editorial class changing things.
We've debated this before, and I don't have any interest in doing so again. I will just point out that if you wanted to prove Wikipedia isn't trustworthy it's not enough to point out there are a few articles that are incorrect. There is no source of information that is perfect. You'd have to prove that it's incorrect so often that in practice an average user is likely to see an incorrect post.
2. Yes, some things are philosophical questions but also there is plenty of objective truth, and there's many issues which do in fact have this. Our pursuit of the truth should pretty much always be our highest priority, and I personally find it striking someone who works in the information business in particular to say this. She isn't a religious leader.
3. The bureaucracy part having ongoing issues for me does illustrate an existential flaw in the idea of an internet encyclopedia anyone can edit. The flaw is endemic so it still gets discussed, I think it's probably worse than any of the political astroturfing since many people continue to note it. Not new, worth repeating to remind people.
You're right though we don't need to re-litigate Wikipedia again, I do hope they read what people point out and take to heart, they're an open internet project after all so they should be open to feedback. If anything it's motivation to do better. Maybe I'm just disheartened because I think the project could be much better.
One interesting aside: I asked Claude about the Osler Wikipedia page and it also thought some of the inclusions were questionable. That’s anecdotal, but it does raise an interesting angle, LLMs may end up acting as a kind of secondary consistency check on Wikipedia-style aggregation (prob works best in domains where there isn’t a large adversarial editing or sourcing ecosystem, no one is trying to put out much new information about that topic)
2. But she wasn't talking about everything. She was talking about those specific issues where there isn't an objective truth. This discussion is a great example of her point.
For the topics you cite (most of which are political questions), LLMs won't be any better than people in answering them because there is no "truth". Just opinions.
the problem is that framing bleeds into domains where there actually are right answers, and honestly it's a big waste of time when people can't agree 2+2 = 4 etc ...I think reverence for the truth is always the right way to process the world, even if we don't yet know, because it will help motivate us to find it ...I understand the context, I just still really disagree with it
On LLMs, you're right they won't resolve political disagreements. The Osler angle was much narrower: in domains with clear historical record an LLM noticing internal contradictions on a page is a useful consistency check
Most universities I have been associated with will not accept Wikipedia as a reference. You can ask AI queries to preclude Wikipedia or Reddit. It's often interesting to do a query with them and then without them as sources. In the end, AI is just like any computer: the input affects the output. This is why schools used to push critical analysis and why critical analysis is needed more than ever.
"they did not become captured through conspiracy. They became captured through neglect."
Agree, but individual editorial choices of Wikipedia impact the site more than the overall LLM training set. A quick Gemini search on "gaza" cites several sources (including Wikipedia). That is not a simple "fairness check", instead an intentional signal-processing decision at the engineering level.
The bad news is that training controls for most frontier models are less transparent today than they were a few years ago. Coupled with the tendency to hallucinations citations, and consumers are correct to doubt anything that the token prediction machine regurgitates. Garbage-in / Garbage-out
Yes to all of this—it’s disheartening. I used to teach a Wikipedia unit—contrasting it with search, which produces unvetted results. I would assign students to analyze the talk and history pages of controversial entries to demonstrate the process of debate, consensus, and a search for verifiable facts. I was trying to convince them to rely more on Wikipedia than Google.
But now that the cabal deletes all citations of publications they don’t like, now that it’s essentially run by AGP (autogynephile) men who identify as women, no biographical entry is reliable any longer. I’m sorry if that sounds conspiratorial, but it’s unfortunately too true. Those men, with their attack dog supporters, have suppressed not only Mideast perspectives but anything to do with sex and gender that differs from their beliefs. It has been the same process as with Reddit—dissenters punished, pushed out, and sidelined. Open debate is no longer tolerated.
Yes, I read Wikipedia for old movies to find out cast lists and production but no longer use it to learn about most other topics.
Yes, very sad. Look if they presented all perspectives on controversial things it would be one thing, but in so many cases one side is basically unwelcome, or is given 98% of the page and maybe 2% at the bottom shares the other side (of course, it must be referencing a certain type of media so you will only get what they are willing to share)
I used to really believe its consensus-based truth seeking was the future of the internet. Larry Sanger, one of its founders, has spoken about the decline of Wikipedia but is today smeared as beyond the pale probably because he is also religious.
Wikipedia: articles must be well-sourced, meet notability guidelines, etc., etc.
Also Wikipedia: has an article on 'Toilet paper orientation'.
I'm not saying everyone should be dead serious, rather the opposite; but a little more consistency wouldn't hurt.
Who are you to impugn the notability and encyclopedia-worthiness of toilet paper orientation?
I happen to believe that toilet paper orientation is encyclopedia-worthy, and so is everything else.
Wikipedia in fact does have some useful information, for example:
https://en.wikipedia.org/wiki/List_of_people_who_have_been_pied
"Consensual pieing is excluded from this list"
Frightening stuff. I’ve noticed that and no longer visit Reddit or use Wikipedia. The bias is apparent. What solutions? How can we make change aside from not using it? When we use AI can we ask it not to use Wikipedia as a source or is it’s training too deep into Wiki already for that not to make a difference?
You're just going to have to do more research to understand the world, and get a bunch of different perspectives, in some cases probably have to average them out to understand what's really happening. Gotta fact check your AI answers, too
Honestly, Wikipedia editors are a self-selecting sample. People with that much motivation to do voluntary work for what they consider a public service are gonna skew left, they just are. You complain that they control things becuase they're the ones that show up - but, *you* didn't show up! You and I and anyone else who uses Wikipedia to find information is freeriding on their efforts.
Ultimately they dedicate their free time to maintaining a massive and gloriously non-commercial website. Honestly complaints of political bias seem like looking a gift horse in the mouth. The bias undoubtedly exists, but the counterfactual here is not a free encyclopedia that is also the embodiment of political neutrality; the alternative is information locked behind paywalls, Elsevier-style.
Just want to point out I have tried to update Wikipedia articles on several occasions, and 100% of my edits, which were factual and appropriately sourced following the rules were immediately reverted. It is definitely a situation not anyone can edit now, despite what the branding is. It's very centralized and controlled
Here we go again...
1. You cited a study by the National Review on how Wikipedia has bias?
2. That quote about the truth was more nuanced than you are considering. Many of these topics that you cite don't have an objective truth. There is no specific, reference-able primary source. These are topics where opinions are widely diverse, and represent the spectrum of political positions. That's what she was saying.
3. The bureaucracy of Wikipedia has existed since the beginning, and has been studied in depth. There is no neglect, this is how it's always worked. You use an anecdote about whether a person is notable, but that has been happening since the very beginning of Wikipedia. That's not new, and there is not some kind of new editorial class changing things.
We've debated this before, and I don't have any interest in doing so again. I will just point out that if you wanted to prove Wikipedia isn't trustworthy it's not enough to point out there are a few articles that are incorrect. There is no source of information that is perfect. You'd have to prove that it's incorrect so often that in practice an average user is likely to see an incorrect post.
1. Sorry the study itself is via the Manhattan Institute you can read here if curious: https://manhattan.institute/article/is-wikipedia-politically-biased
2. Yes, some things are philosophical questions but also there is plenty of objective truth, and there's many issues which do in fact have this. Our pursuit of the truth should pretty much always be our highest priority, and I personally find it striking someone who works in the information business in particular to say this. She isn't a religious leader.
3. The bureaucracy part having ongoing issues for me does illustrate an existential flaw in the idea of an internet encyclopedia anyone can edit. The flaw is endemic so it still gets discussed, I think it's probably worse than any of the political astroturfing since many people continue to note it. Not new, worth repeating to remind people.
You're right though we don't need to re-litigate Wikipedia again, I do hope they read what people point out and take to heart, they're an open internet project after all so they should be open to feedback. If anything it's motivation to do better. Maybe I'm just disheartened because I think the project could be much better.
One interesting aside: I asked Claude about the Osler Wikipedia page and it also thought some of the inclusions were questionable. That’s anecdotal, but it does raise an interesting angle, LLMs may end up acting as a kind of secondary consistency check on Wikipedia-style aggregation (prob works best in domains where there isn’t a large adversarial editing or sourcing ecosystem, no one is trying to put out much new information about that topic)
2. But she wasn't talking about everything. She was talking about those specific issues where there isn't an objective truth. This discussion is a great example of her point.
For the topics you cite (most of which are political questions), LLMs won't be any better than people in answering them because there is no "truth". Just opinions.
the problem is that framing bleeds into domains where there actually are right answers, and honestly it's a big waste of time when people can't agree 2+2 = 4 etc ...I think reverence for the truth is always the right way to process the world, even if we don't yet know, because it will help motivate us to find it ...I understand the context, I just still really disagree with it
On LLMs, you're right they won't resolve political disagreements. The Osler angle was much narrower: in domains with clear historical record an LLM noticing internal contradictions on a page is a useful consistency check
Most universities I have been associated with will not accept Wikipedia as a reference. You can ask AI queries to preclude Wikipedia or Reddit. It's often interesting to do a query with them and then without them as sources. In the end, AI is just like any computer: the input affects the output. This is why schools used to push critical analysis and why critical analysis is needed more than ever.
"they did not become captured through conspiracy. They became captured through neglect."
Agree, but individual editorial choices of Wikipedia impact the site more than the overall LLM training set. A quick Gemini search on "gaza" cites several sources (including Wikipedia). That is not a simple "fairness check", instead an intentional signal-processing decision at the engineering level.
The bad news is that training controls for most frontier models are less transparent today than they were a few years ago. Coupled with the tendency to hallucinations citations, and consumers are correct to doubt anything that the token prediction machine regurgitates. Garbage-in / Garbage-out
Yes to all of this—it’s disheartening. I used to teach a Wikipedia unit—contrasting it with search, which produces unvetted results. I would assign students to analyze the talk and history pages of controversial entries to demonstrate the process of debate, consensus, and a search for verifiable facts. I was trying to convince them to rely more on Wikipedia than Google.
But now that the cabal deletes all citations of publications they don’t like, now that it’s essentially run by AGP (autogynephile) men who identify as women, no biographical entry is reliable any longer. I’m sorry if that sounds conspiratorial, but it’s unfortunately too true. Those men, with their attack dog supporters, have suppressed not only Mideast perspectives but anything to do with sex and gender that differs from their beliefs. It has been the same process as with Reddit—dissenters punished, pushed out, and sidelined. Open debate is no longer tolerated.
Yes, I read Wikipedia for old movies to find out cast lists and production but no longer use it to learn about most other topics.
Yes, very sad. Look if they presented all perspectives on controversial things it would be one thing, but in so many cases one side is basically unwelcome, or is given 98% of the page and maybe 2% at the bottom shares the other side (of course, it must be referencing a certain type of media so you will only get what they are willing to share)
I used to really believe its consensus-based truth seeking was the future of the internet. Larry Sanger, one of its founders, has spoken about the decline of Wikipedia but is today smeared as beyond the pale probably because he is also religious.