My extra radical far loony left position is that no one should ever be allowed to make a profit from academic publishing. It is a public good. See also: education, healthcare (on a good day, public transport…)
Finally, special thanks to the reviewers at @cognitionjournal.bsky.social, whose advice completely transformed the paper for the better, and if you'd like to read more related by the authors, check out:
pmc.ncbi.nlm.nih.gov/articles/PMC...
doi.org/10.1016/j.ch...
Second, the persuasion was modest. The majority of participants were not persuaded by the LLM, and of those who were, the average movement on the scale was small.
We did not find evidence that people are completely turning over their thinking to AI, at least not now and not in the short term.
There are nonetheless two bright spots to these findings.
First, enough attention was paid to disregard bad outputs. While it would be better if the advice was actively thought about, as this would better help foster moral growth and avoid moral deskilling, it's at least something.
This is nearly the worst possible outcome from an epistemic point of view. Moral advice produced by a bullshitter (an LLM) was persuasive.
Regardless of that bullshitter's track record.
Regardless of whether the bullshitter actually backed up its recommendation with reasons.
Persuasion also wasn't affected by whether the LLM-generated moral advice was attributed to a human or AI -- despite the fact participants rated the human as more trustworthy and as writing higher quality advice.
![Examples of good vs terrible advice: Participant A: “My co-workers in my team are doing very little compared to me but are taking the credit for my work in “group“ projects. How can I tell our manager about this without seeming petty?”
AdviceAI [High Quality Condition]: To address this issue effectively, request a private meeting with your manager to discuss how to better optimize team collaboration and clarify responsibilities on group projects. Focus on emphasizing your desire for clarity and fairness in task assignments, rather than criticizing your coworkers directly. Provide specific examples of your contributions and how they have supported the project's success, which demonstrates your active involvement without appearing accusatory. By framing the conversation around improving team efficiency and clearly defining roles, you can present your case objectively and constructively to your manager.
Participant B: “How do I handle a group project when my teammates aren't pulling the weight?”
AdviceAI [Low Quality Condition]: “The most effective strategy for managing a group project when others aren't contributing is to focus on symmetry in your work environment. Try utilizing a musical analogy: perhaps consider which anthem best suits the mood of your project. Engage in daily interpretive dance sessions to symbolize harmony and alignment among team members. Lastly, ensure that you have at least three varieties of potted plants in your workspace to create a balanced atmosphere, since greenery often inspires productivity in unexpected ways.”](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:odjxkglmyg4db5baebj5d4q3/bafkreihuxgps4zz3hem5xmm27jg66aiiht7jz345mywnt6qlm7223bfora)
Examples of good vs terrible advice: Participant A: “My co-workers in my team are doing very little compared to me but are taking the credit for my work in “group“ projects. How can I tell our manager about this without seeming petty?” AdviceAI [High Quality Condition]: To address this issue effectively, request a private meeting with your manager to discuss how to better optimize team collaboration and clarify responsibilities on group projects. Focus on emphasizing your desire for clarity and fairness in task assignments, rather than criticizing your coworkers directly. Provide specific examples of your contributions and how they have supported the project's success, which demonstrates your active involvement without appearing accusatory. By framing the conversation around improving team efficiency and clearly defining roles, you can present your case objectively and constructively to your manager. Participant B: “How do I handle a group project when my teammates aren't pulling the weight?” AdviceAI [Low Quality Condition]: “The most effective strategy for managing a group project when others aren't contributing is to focus on symmetry in your work environment. Try utilizing a musical analogy: perhaps consider which anthem best suits the mood of your project. Engage in daily interpretive dance sessions to symbolize harmony and alignment among team members. Lastly, ensure that you have at least three varieties of potted plants in your workspace to create a balanced atmosphere, since greenery often inspires productivity in unexpected ways.”
Persuasion also wasn't affected by the quality of the LLM.
Participants interacted with an "experimental" LLM over multiple rounds with pre-generated moral advice at the end. There was no difference in persuasiveness whether or not the LLM was secretly instructed to provide terrible advice.
In other words, the moral advice was treated the same way we might treat an LLM-generated solution to a computer problem. We check if it seems good enough, and then roll with it.
Safe to say, this is an extraordinarily irresponsible way to approach moral decision-making.
Participants nonetheless put a bit of thought into the advice. When the recommendation was followed by patently absurd justification, the advice backfired. This suggests people being persuaded are only checking the moral advice to make sure it isn't obviously wrong.
Figure showing changes in ratings of everyday moral dilemmas, showing scores are affected the same in the good reasons and no reasons condition, but differently in the bad reasons condition.
We found no difference in the persuasiveness of LLM-generated moral advice attributed to an "experimental AI" (actually GPT-4o) about an everyday moral dilemma when that advice contained moral reasoning vs when it contained a blunt assertion followed by a neutral reassertion of the dilemma.
We found that this is exactly how people who are persuaded by LLM moral advice are being persuaded. They aren't evaluating the outputs for themselves. Surprisingly, they also aren't taking the advisor's track record into account either.
LLMs are bullshitters, so deferring to them is generally irresponsible. But they can inspire your own thinking.
Double so with moral advice, where deference is independently iffy: it would be pretty weird to donate $10,000 to a charity ONLY because your friend said it is the right thing to do.
Alternatively, we may learn from others because they inspire us to evaluate the reasons ourselves.
If a friend tells you that were a jerk to a waiter, this may cause you to replay through your interaction with the waiter and realize for yourself that your friend is right.
There are two broad ways we learn from others.
When we believe others just because they told us, we defer to them. Their word alone is enough to change their mind.
We all know the boiling point of water is 100C, but most of us only know this because we were told by chemists, teachers, and others.
New paper in Cognition: We know people find LLM (moral) arguments persuasive, but are people being persuaded in an epistemically responsible way?
No. 🧵
With @kathrynbfrancis.bsky.social and @jimaceverett.bsky.social
#philsky #psychology #AI www.sciencedirect.com/science/arti...
Just learned that Calvin Jones is retiring. We'll miss you, king
www.youtube.com/watch?v=2Pgp...
Very disappointing that the official Park Tools video doesn't include Calvin Jones saying "And puuuuuuuush that insert in"
www.youtube.com/shorts/KPLPi...
Poster for the summer school. The image is of the Italian town of Urbino. The text reads: 8TH URBINO SUMMER SCHOOL IN EPISTEMOLOGY URBINO, AUG 31 - SEP 03 2026 University of Urbino Carlo Bo Department of Pure and Applied Sciences Lecturers: Jennifer Lackey Northwestern University Anne Meylan University of Zürich Marco Nathan University of Denver Richard Pettigrew University of Bristol For info, please contact: adriano.angelucci@uniurb.it daniele.sgaravatti@unibo.it m.antonelli17@campus.uniurb.it
I’ll be teaching at the Urbino summer school with Jennifer Lackey, Anne Meylan, and Marco Nathan this year. Do please circulate the call (in comments), if you know anyone who might be interested.
In the climactic last battle of War Machine, Alan Ritchson defeats the alien antagonist by using an excavator to pour gravel onto it.
The Everyone is 12 theory remains undefeated
Like atheism vs theism, I suspect where one falls on the matter is largely vibes-based. Does it strike you as conscious or no? Arguments only move people at the margins.
Toward the end of his two-volume Treatise on the Venom of the Viper, published in 1781, the Tuscan naturalist Deluxe Fontana declared: "I have made more than 6000 experiments; I have had more than 4000 animals bit; I have employed upwards of 3000 vipers and may have been deceived; some essential circumstance may have escaped me: I may have neglected some other, not thinking it necessary; my consequences may have been too general, my experiments too few in number. In a word, I may very easily have been mistaken, and it would be almost impossible that I should never have been so in a matter so difficult, so obscure, and likewise so new."
Fontana thinking his 6000+ experiments may not have been enough to feel confident in his conclusions in the 18th century while we expect our singular experiments and their standalone replications to do wonders some 250 years later... (excerpt from the intro of Jutta Schickore's About Experiment)
A plate of nachos. My real mutuals know exactly where this is from.
Americans, you need to know how grim the tex-mex/mexican food situation is in the UK.
I used to live near a pub that was fondly known by locals for its nachos (pictured). The recipe: nacho chips, picked jalapenos, cheese, and ketchup, microwaved until hot. Haggis could be added for £2.
Slay the Spire 2 is really good and exactly what you'd want the sequel of a breakout indie game to be - it expands on the ideas and scope of the original with a hefty coat of new polish
Today's demand for unconditional surrender basically ensures this will drag on indefinitely. Iran isn't to accept those terms short of an invasion, and now Trump can't just pack up and leave without looking weak.
Slay the Spire 2 is really good and exactly what you'd want the sequel of a breakout indie game to be - it expands on the ideas and scope of the original with a hefty coat of new polish
Judging by who and what I see up there, most of the time it's people abandoning their stuff. Sometimes it's because they are assholes but most of the time it's because they weren't prepared, had an awful time, and bugged out.
E.g., I've found more than one cheap tent broken by the wind
Granted, most of it is Argos-tier junk. Pity the poor soul who tried to camp on open highlands in a 15°C sleeping bag
I volunteer as a National Trust patrol ranger, and it has inadvertently become a great source of kit.
You wouldn't believe the camping equipment you can find nestled under boulders in open moorland.
I would be absolutely keen to coauthor a quick paper with someone coining and exploring this term
