I agree that the central challenge is no longer only who controls AI models or which system is more persuasive, but also how these systems begin to shape the very process of legal understanding.
Even with human oversight, a prior question remains: how do we preserve the autonomy of the judge's or lawyer's interpretive process so that it is not silently replaced by the logic of the system?
AI governance should therefore be complemented by the governance of the human process of understanding. Otherwise, we may preserve formal accountability while gradually losing the interpretive sovereignty that makes genuine legal judgment possible.
Now, it's a different question whether the lab experiments leading to this conclusion have any purchase in real life, and it also depends on the claim's scope ("better than most humans" is maybe a better conclusion).
I agree. And I think this raises another question: we should ask not only whether AI is more persuasive, but also what kind of cognitive process produces that persuasion. Persuasion is not the same as understanding, interpretation, or legal judgment. If we conflate these dimensions, we may end up optimizing influence while silently eroding legal judgment.
Maybe there’s no applied value at all to this “research.” Maybe it’s aspirational, but not grounded in actual transactions that measure the UX, end to end.
I agree that the central challenge is no longer only who controls AI models or which system is more persuasive, but also how these systems begin to shape the very process of legal understanding.
Even with human oversight, a prior question remains: how do we preserve the autonomy of the judge's or lawyer's interpretive process so that it is not silently replaced by the logic of the system?
AI governance should therefore be complemented by the governance of the human process of understanding. Otherwise, we may preserve formal accountability while gradually losing the interpretive sovereignty that makes genuine legal judgment possible.
“First, it is increasingly evident that LLMs are, in general, better persuaders than us humans.”
Nonsense.
That's what research papers establish, so far !
Now, it's a different question whether the lab experiments leading to this conclusion have any purchase in real life, and it also depends on the claim's scope ("better than most humans" is maybe a better conclusion).
I agree. And I think this raises another question: we should ask not only whether AI is more persuasive, but also what kind of cognitive process produces that persuasion. Persuasion is not the same as understanding, interpretation, or legal judgment. If we conflate these dimensions, we may end up optimizing influence while silently eroding legal judgment.
Maybe there’s no applied value at all to this “research.” Maybe it’s aspirational, but not grounded in actual transactions that measure the UX, end to end.
https://www.itpro.com/technology/artificial-intelligence/ai-agents-arent-cutting-it-in-customer-service
It’s similar to legal AI. A lot of “increasingly evident” adoption. And a lot of increasingly evident distrust and disdain.
Brilliant stuff
Thanks !