Have you tried talking to people about what you believe?

It’s been widely agreed among psychologists and exasperated members of the general public that conspiracy theorists are beyond salvation by reason alone. Into the mix comes this study suggesting that LLMs are extremely good at persuading people to abandon or weaken their beliefs in conspiracies:

https://osf.io/preprints/psyarxiv/xcwdn

These are big results if they stand up to replication and scrutiny. A 20% on average reduction in belief with many respondents giving far higher change scores is quite an effect for a one-off intervention. Far more shocking is that these changes persist. At scale, this is potentially world-changing. Terrifyingly, it probably will be, as large language models are rolled out as infinitely patient, highly knowledgeable propagandists to influence us to believe… whatever it is their owners want us to believe.

[Awkardly the conspiracies tackled included some that seem to me to have merit, e.g. “Jeffrey Epstein didn’t kill himself” and “Malevolent corporations”, although the latter mostly seems to be about clearly nuts stuff like corporations have a secret cure for AIDS and cancer. Interestingly ChatGPT had almost no effect on the belief that JE didn’t kill himself, unlike most of the other conspiracies.]

But what I find most interesting about the result is not that it has been achieved by LLMs, but that it has been achieved at all. The myth that homo sapiens cannot be persuaded by reason may be just that- a myth created by a coalition of:

1. Lazy people, and people without patience who don’t want to sit through long discussions with those who they think are wrong. These people will spend 15 minutes attacking another person’s thoughts as if with a sledgehammer before throwing their hands up and claiming it’s all futile.

2. People who want to feel smugly superior to the mob who (unlike them) are not susceptible to reason.

3. Related to (2)- people who wanted to feel smarter than those sophomoric rubes who foolishly think that mere reason can persuade. Don’t they know everything is about power (in the most vulgar sense imaginable).

4. Marketers and sophists who want to sell alternatives to persuading by reasonable conversation. I’m not denying that their methods work, or even advising against using them, but their claims to sole efficacy are hyperbolic. Reason works to.

I do not think these experimental results reflect a special power of large language models beyond the reach of all or most humans. I think many people can persuade perfectly well if they put their mind to it. Even me and you, dear reader.

But we just don’t try to persuade people about big questions. People think they put a lot of effort into persuading others, but they don’t. They try to make themselves look cool and vent their frustrations.

“I’ve tried to persuade my relatives of leftwing beliefs, I just couldn’t do it!” Bullshit, nine times out of ten this is just someone who got angry at their parents at Thanksgiving dinner. Also, parents are the hardest people to persuade, this is like trying the game on Xtreme megapermadeath mode, failing, and then deciding the game is completely impossible. As with politics, so with conspiracy theories. “Conspiracists are immune to reason”, well I’m sure some of them are, but as for the rest, I suspect a lot of people simply don’t try to persuade them.

I never found persuasion especially difficult. I can’t persuade someone of X through a reasoned conversation on demand, but I have set out to persuade people of leftwing ideas as I go through life, and I’ve had reasonable success over the years.

Don’t get me wrong, I’m not saying anyone could instantaneously replicate these results. Contemporary LLMs are quite persuasive, especially for, let me be blunt, literate, but not especially widely read people. But the results are, I suspect, not unobtainable. This is especially true given our great advantage: we can talk to people in real life, not just via text, giving us significant leverage if we use it.

It is so strange hardly anyone on the left tries to be persuasive. At a first approximation, there are two things you can do to advance what you believe in:

A) Try to organize people who agree with you and

B) Try to convince more people to agree with you

People aren’t doing a lot of B, and it’s certainly not because they’re too busy doing a lot of A, so I reckon you should take inspiration from our text-golem friends and persuade people.

Let me give some tips. Your task is not to use reason alone, like a weapon to force your interlocutor into agreement, nor is it to emotionally manipulate them till they say yes. Your task is to blend (if you’ll indulge): Logos, pathos and ethos. Reason, shared feeling, and ethical standing. It is not to monologue but to dialogue. It is not to lie, or prevaricate, nor is it to raise every single disagreement you have with what is said all at once, but rather to focus on the core issues and potential points of consensus.

Here is my advice:

1. Persuasion isn’t going to work on everyone. There’s such a thing as a lost cause (although sometimes the world can surprise you).

2. Treat your interlocutor like an equal and always take them seriously.

3. Build rapport. It’s easiest to persuade friends.

4. Be and be seen to be someone worthy of respect

5. Listen closely to what they say and demonstrate that you understand what they’ve said (e.g. through mirroring and paraphrasing).

6. Affirm points of common ground, and especially points of common ground that move them closer to what you think. Build on that common ground, and expand it. Try to show that in some sense they already agree with you on the issues.

7. A lot of the time, what you are trying to do is to get them to persuade themselves that you are right.

8. Know your facts.

9. Don’t pick unnecessary quarrels, focus on the core issues. Pick your battles.

10. Seek existing tensions in what they believe

11. Clarify issues, don’t muddy them

12. Try to have a clear sense of what aspects of what you believe actually matter and will actually help people or prevent harm.

13. You’re not trying to win. You’re trying to help the world.

14. Although I prefer discussion intermingled with only a little bit of debating, arguing is more effective than is often recognized, especially if you really know your stuff. People may not change their minds on the spot because they don’t want to lose face, but their thoughts are still shaped by losing an argument. Often, even when you can’t persuade the person you’re arguing with, you can persuade onlookers. Still, try to avoid big arguments where you can and focus on discussion with a bit of argument interspersed as appropriate.

In terms of things to study- it’s worth going through a critical reasoning course. Make sure you understand:

1. Rogerian persuasion

2. Basic counseling method (mirroring statements, reflective listening)

3. Aristotlean (syllogistic) logic, soundness and validity

4. Informal fallacies (FOR GODS SAKE DON’T GO AROUND MENTIONING THEM BY NAME LIKE YOU’RE CASTING SPELLS FROM HARRY POTTER. NO ONE LIKES THE GUY WHO SAYS AD HOMINEM).

5. The basics of probability theory, including Bayes theorem and its application to everyday life.

6. Be very clear on whether a debate is substantive or merely verbal or mixed and on whether a debate is positive, normative or mixed.

All of us have to decide whether we are interested in signaling our commitment to our political beliefs as we would a subculture, marking people on their conformity to our club, and venting our frustrations with the world, or whether we are interested in HELPING. Trying to persuade is a good step for those aiming at the latter.