LLMs give us a really useful tool for political conversations where “mind killing” is a real risk. We can now post the entire context to Claude and ask a neutral question like “Does it look like <username> is engaging with arguments that might change his mind, or defending against them?”.
What do you predict Claude will say, and are these predictions validated by experiment?
Or rather, before running that check, what would the appropriate response be if one were to run such a check and not receive ego-syntonic feedback? If I run that check and Claude says “Jimmy’s comments contain the following signs of cognitive dissonance:”, what should I do with that?
Please critique the discussion styles of the participants in this discussion
This is an interesting discussion to analyze. Here’s my critique of each participant’s style:
Nathan Young
Strengths: Opens with genuine intellectual uncertainty and acknowledges tensions in his own position (“Which leaves me sort of confused”). He’s willing to admit aesthetic discomfort with positions he’s trying to defend on game-theoretic grounds. This shows real intellectual honesty.
Weaknesses: His final response (“I think your response would admit basically no evidence”) is a meta-dismissal rather than engagement. Jimmy offered specific criteria for what would constitute a true parallel—Nathan could have either met those criteria with better examples or challenged the criteria directly. Instead, he essentially accused Jimmy of being unfalsifiable without demonstrating it.
interstice
Strengths: Concise and makes a specific, relevant distinction (assassination vs. other attacks).
Weaknesses: Too terse to be fully productive. The brevity leaves Nathan to guess at the underlying reasoning, which creates friction.
jimmy
Strengths: The most analytically careful. He draws specific distinctions (severity of harm, nature of mockery, what a true parallel would require), acknowledges being a Destiny fan while still critiquing him, and tries to articulate where Destiny’s symmetry-breaking logic fails on its own terms.
Weaknesses: His distinctions could read as motivated reasoning—each example Nathan raises gets explained away with a different caveat. The closing move (suggesting Claude as arbiter) is clever but slightly passive-aggressive; it’s a way of saying “you’re being defensive” while maintaining plausible deniability. He also doesn’t quite grapple with whether his own framework might be similarly self-serving.
Overall pattern: The discussion shows a common failure mode—it becomes a battle over whether specific examples are “close enough” parallels rather than directly addressing the underlying question about when reciprocal norm-violation is justified. Both Nathan and Jimmy are partially right that the other is being somewhat selective.
End of claude.
I’m not really interested in continuing to discuss with you if you are going to dismiss the Paul Pelosi attack out of hand. I just don’t think we’ll get anywhere. Merry Christmas!
Well of course. Can you honestly not see the difference?
LLMs give us a really useful tool for political conversations where “mind killing” is a real risk. We can now post the entire context to Claude and ask a neutral question like “Does it look like <username> is engaging with arguments that might change his mind, or defending against them?”.
What do you predict Claude will say, and are these predictions validated by experiment?
Or rather, before running that check, what would the appropriate response be if one were to run such a check and not receive ego-syntonic feedback? If I run that check and Claude says “Jimmy’s comments contain the following signs of cognitive dissonance:”, what should I do with that?
End of claude.
I’m not really interested in continuing to discuss with you if you are going to dismiss the Paul Pelosi attack out of hand. I just don’t think we’ll get anywhere. Merry Christmas!