Law360 mandates reporters use AI "bias" detection on all stories
niemanlab.org
niemanlab.org
> If Jimmy Allred says it’s raining, and W. Lee O’Daniel says it isn’t raining, Texas newspapermen quote them both, and don’t look out the window to see which is lying, and to tell the readers what the truth is at the moment.
https://quoteinvestigator.com/2023/11/14/rain-look/
If an AI tool, or for that matter a meddling editor, says a headline is "framing the action as unprecedented in a way that might subtly critique the administration", the correct response is "yes, that was the idea".
Yes, some reporting is biased. But some reporting is simply accurately reporting damaging information, and "biased" is a way of attacking that without addressing the substance of the problem.
Which is perhaps why organizations are searching for ways to not appear biased and begin to restore that trust. I think the reason the LLM approach is likely to fail is because it is unable to detect the "missing facts" case and can only really advise about sentiment and phrasing. Which is not I think the actual problem that needs fixing.
Or, even less charitably, management and employees have different politics, and management are the ones who find the articles inconvenient.
To be clear, there are absolutely biased news sources out there. For many of them, the bias is the point, and they have no particular desire to "restore trust" because they're already trusted by people who only want to read things supportive of their party. But a politician who finds the truth inconvenient will decry everything accurate as biased.
I have seen many documentaries that contain only facts and real events, but nevertheless are pushing a heavily biased agenda. Which facts we report and which we highlight and how we frame them tells a story.
For example: "the bias indicator flagged a line that said the suit spotlighted challenges with ableism and sexism in the healthcare industry. This copy was flagged because it “frames the lawsuit as a representative example of systemic issues.” Instead, the bias indicator said the story should “state the facts of the lawsuit without suggesting its broader implications.”
The model made the right decision. That's the kind of language that makes people distrust journalists: a passive voice assertion of a radical belief, which adds nothing to the story and isn't supported by any evidence. It doesn't belong in a professionally written report.
And yet we read, "that edit is at odds with any attempt to deliver legal analysis. In most cases, reporters chose not to accept these edits, but they were still required to go through the motions."
Claiming the healthcare system is challenged by "ableism" isn't legal analysis, that's pure politics. But more importantly it's also just absurd. Of course people who work in healthcare want people to be "able". Why would you become a surgeon if you thought disability and sickness was fine? No surprise their bosses are placing the LLM's judgement over that of their own employees, when the journalists have gone so far off the rails. The union should be careful of picking a fight over this because frankly the next step is to just automate away the journalists entirely.
- Bruce Schneier
Edit: substitute any issue you care to name for “health system”, there’s always a crisis that can be used to grab eyeballs.
The common "must present both sides" approach can fail the "nothing but the truth" criterion. But even with its many errors, it's better than not trying anything at all.
(If my comment offends you, I assure you, don’t worry, your party is definitely the Good Party in this scenario.)
> characterizarions
Sigh, mobile keyboard. *characterizations
Obviously there are biased antisemitic news organizations, but of the high-profile ostensibly neutral ones BBC stands head and shoulders above the rest in the level of antisemitism.
(You don't have to take my word for it - a quick Google will find huge numbers of examples. Usually they'll get criticized and then post a correction, so in some sense they themselves acknowledge the problem - yet it keeps happening.)
Despite that, most people who use the word "antisemitic" apply it only to something that is done against hebrews, and not to something done against arabs.
Therefore it makes sense to request clarification about what someone means by that word, i.e. if they meant that BBC is anti-Israel or it is anti-arabs.
It would be much better if everyone who means that something is anti-Israeli, would say it clearly, instead of using the ambiguous word "antisemitic".
The word "Semitic" has been created due to a misunderstanding of the Bible, because there the classification of the people was not based on real descendance from common ancestors, but it was based on the political dependence of those people at the time when the Book of Genesis was written. Unfortunately, nobody has found a suitable replacement for this word.
(In the Bible, the descendants of Shem were those dominated by Assyro-Babylonia, while the descendants of Ham were those dominated by Egypt, regardless of their true ancestors. For instance the Phoenicians were classified as descendants of Ham and the Hebrews as descendants of Shem, despite being 2 extremely closely related populations, separated by little else except their different religions.)
If you're inclined to offer the most charitable interpretation in the universe, perhaps.
112 BBC journalists recently signed an open letter accusing the BBC of pro-Israel bias[2].
Domestically, every single political movement feels the BBC is biased against them. Their dry reporting style is at odds with what many, especially Americans, are used to. Unlike the author of that blogpost I do not really have a problem with them not using the word 'genocide' but I do take issue with some of their selective reporting (of course, every outlet must make a selection and everyone will have a problem with it). For example, Israeli newpaper Haaretz has been doing a lot of reporting on IDF soldiers being ordered to fire on crowds of unarmed civilians at aid sites[3]. The BBC does acknowledge it, but it's buried behind a headline about the IDF starting an investigation[4]. They do this because they always prefer quotes from official sources to eyewitness stuff.
[1]: https://www.jonathan-cook.net/blog/2025-06-20/bbc-editors-tr... [2]: https://docs.google.com/document/d/1n3926pSPNwXd8j7I716CBJEz... [3]: https://archive.is/pQw3k [4]: https://www.bbc.co.uk/news/articles/cj9vveg0vp9o
Like when the BBC said all converts to Islam are "reverts"; https://www.spectator.co.uk/article/why-did-the-bbc-say-musl... ; the nasty implication being that one religion is the status quo of humanity
Or when the BBC framed an Ad as racist, because it called out a politician's sectarian & anti-LGBTQ appeals https://www.bbc.com/news/articles/c5yg0g18989o https://www.reddit.com/r/unitedkingdom/comments/1ksv0y7/refo...
The other day actually, I saw a newscaster describe the horrors Apartheid in South Africa as though it were happening today. She then closed with "... [Apartheid] is now no longer enforced" and transitioned quickly to the next topic. Not ended or abolished, but merely "not enforced".
Their bias is rather pernicious.
They didn't. They used their interviewees' own preferred terminology to refer to them in the story, which is fine. "Reverts" is the most common term among Muslims. Sort of like how capitalizing God in a story about Christians doesn't invalidate the beliefs of people who worship multiple gods.
> Or when the BBC framed an Ad as racist
The first line: "Scottish Labour has described [...]".
This is an insane characterisation of that article, which seems fair to me and not at all what I was talking about. Wanting your race to be represented in politics is a perfectly reasonable view and it doesn't need "calling out".
The way I see it: a politician was intimating to a community, known for its conservative religious practices, that they can one day enforce their ideal in their children's education. If the dogwhistle is too quiet, consider how Muslims, Christians and other non-racial communities drove the US Supreme Court ruling to allow opt-out of LGBTQ education.
IMHO, the insane idea is "a community is a race". As if South Asians getting more votes translates directly to "representing their race", not their culture or religion. It's a similar kind of crazy to "Islamophobia is basically racism" which the UK government has been trying to legislate: https://www.arabnews.com/node/2588955/amp
Outright lies are very ineffective in manipulating public opinion because they can be easily disproven.
It's much more about pretending its only a little rain when its raining alot. Ignoring the rain when it doesnt support your narrative. Pretending the rain is really important when its not important at all. Pretending it only rains where you are and much less everywhere else etc etc etc.
I sincerely wish this were true. "A lie can run around the world before the truth has got its boots on."
Just look at what happened in Iran, it's a shitfest of constantly changing narratives, going from warmongerism to peace claims. Lying about what they did and wether it was successful. Lying about wether Israel is on board with making peace.
https://www.cnn.com/2021/04/27/politics/donald-trump-joe-bid...
> Party B: wants way more slaves
> Media: record low percentage of Americans satisfied with agriculture today!
Impartiality is very important.
General Sherman Begins Demolition in Georgia For Urgently Needed Infrastructure Projects
I mean, you could ask ChatGPT or something to make some pretty braindead impartial newspaper headlines about, like, the Holocaust or slavery. I guess you're right, in some cases, impartiality will come off as a bit ridiculous to the common sense segment of society.
That being said, I'm not sure how much this actually matters, trump seems to pretty comprehensively prove that people would rather hear lies, even if they know they are lies.
I don't want to get into the gory details in this tiny text box, but any number of current fear based political campaigns use wildly "distorted" facts to bolster their arguments, and most of the time reporters just blindly repeat whatever the person says without attempting to verify it.
"Might subtly critique"?! This isn't protecting versus bias, this isn't even charitably regardable as cowardice.
It's an attack on reason & reporting. To shield the worst cry bullies & offenders. It's ghastly how basic simple reporting so offends, that the public interest is sabotaged by these foes. We face such a need to improve the media, and this is the opposite path, of craven blunt un-intelligently intermediated corporatism.
There was one piece on a case I’ve been following quite closely where I was genuinely unable to make sense of what had happened from the Law360 report without sitting down and spelling my way through it.
If you wanted unbiased information, you should have asked for facts..
and then collected as many of them as you possibly could.
"Narrative" in this context is more like "attempting to steer the current discourse by making connections among things and positioning them as part of a coherent story", which can be positive (if the result is accurate and reasonable and helps people better understand what's going on) or negative (if it's spin or manufacturing consent). It's "narrative" to say "you should be afraid of X, it's the cause of all your problems". It's also "narrative" to say "here are the five different things we're currently doing to improve Y, and how they tie together into a coherent picture".
Also, bias would be easier to deal with if it were always "lying", or if it advanced a coherent position. It's much broader than that.
Why would you assume that the full spectrum of opinions on an issue would be unchanging?
The concept of the Overton window is meant to represents how attitudes do change in response to variations in perceived extremes.
> Assuming that the midpoint is probably correct is equivalent to saying that any change from the status quo is probably bad. Is that what you really mean?
Basically, yes. I you have a mature system that's reached a stable equilibrium, then its parameters are likely to be optimized against its inputs and constraints, so deviation from that equilibrium is likely to produce less optimal results. If the inputs or constraints have changed significantly, of course, then that might not be applicable.
The spectrum of actual opinions (which is related to the Overton window) may not be unchanging, but the spectrum of all possible opinions by definition can't change. The range of opinions on nuclear weapons goes from "Let's use all of the nukes that exist right now, and also build more and use those, at any cost." to "Nuclear weapons are so evil that any amount of sacrifice is justifiable if it effects even a minuscule change in the likelihood of nuclear weapons being used, or more being built, or existing ones dismantled." This has been true since they were invented, and will continue to be true forever. But both opinions are outside the Overton window, which changes over time.
For instance, if you "just" look at abortion, trans rights, and immigration, you may assume the two parties in america present diametrically-opposed groups of people (...which is even itself quite debatable). But this is only because the two parties don't differ much (or at least, don't bother to platform enough to evaluate) on most topics politics could be about.
Let's say a media outlet, like a newspaper, commits to common journalism best practices, such as neutrality. In my opinion, this does not mean avoiding certain words; it means presenting different perspectives on a topic based on evidence.
Besides that, if someone writes an Op-Ed or commentary, the author expresses their personal opinion. How does this work with this AI thingy? Currently it looks like plain censorship, isn't it?
And why should I buy a newspaper that does not represent my political or personal opinion? I thought that was how it worked.
This is a marketing/sales piece for LexisNexis by their marketing agency as they try to sell AI services to their customers (law firms and other legal professions). It gives them a news article to share with their customers and tell them something like "see, we're taking the AI plunge. You should too. We offer a bunch of legal AI services, lets talk". Now their CEO and sales people can keep saying "I'm sure you've seen the news about Law360. We think AI is finally ready for legal tasks. You know, as an assistant."
In reality, they are probably to run the article, look at what the AI says and shrug and move on.
Mandate productivity, mandate quality; the tools will be adopted if they are sharp.
But elevating the way work is performed above the work itself? Weird.
This is LexisNexis requiring all of their workers run their output through a censorship machine to comply with the whims of the current administration. It has zero benefits to the workers so of course they have to mandate it
The article says that explicitly: customers have complained about the quality of the work. The business owners view unhappy customers as a threat to their business, and so have introduced automation to make product quality higher and more even. The workers are unhappy because they don't want to be automated away and resent the implication that their artisanally hand-crafted articles aren't good enough. This is a very old story that's been going on for hundreds of years. There's nothing special about it.