Well, for my part, I have a couple of things that have helped me here (determined by change in outcome post-intervention which I, unfortunately, cannot share):
1. I know Aumann's Agreement Theorem and the appropriate lemmas about information transfer and prior modification (i.e. if I start failing to modify my priors with repeated conversation, I am failing or the other person is failing, either of which leads to our conversation failing at converging => we should exit)
2. I try to place myself in situations where the difference in outcome is a large multiple of the difference in knowledge of reality. This biases me to accept input well, since I want positive outcome and biology will ruthlessly optimize me toward it. Not every situation is like this. Sometimes, the part of reality you want to optimize against is specifically the people, ignoring the underlying reality. e.g. Though I had a respirator with P100 filters on all exhale and inhale valves I took it off and wore my standard N95 mask when on a plane when asked to. The truth didn't matter there as much as the part of the truth which is "don't piss off flight attendants".
3. I force skin in the game. For instance, I took all my money out of Vanguard and actively managed it in 2019 on a 5x lever. My idea was that I could beat the market, and I did. I also believed in early Feb that despite having advance warning from my friends in Asia about COVID-19 that it was a huge deal, that American bureaucracy was likely to be sophisticated and evidence-based. Since I knew the correct actions to be taken in response, I assumed they knew, and I assumed they had the wherewithal to make it happen. It appears that none of these things are true. But it was a good lesson. It helps me understand how some people I am distantly related to make all that money on defence contracts in Virginia - the bureaucracy is actually not sophisticated, it is merely large.
4. I model us all as message-passing elements that incur a cost per message. I make a good effort to pass you information that has been effective to me to encourage you to do so to me. Multiple elements doing so will lead to easier misinformation detection, less variance in information availability, etc. This is good for me since I am a very capable execution machine. I can only lose in the world in the ways I care about because I don't have information, not because I can't execute.
The consequence of the message-passing element idea is that I don't try to convince, i.e. if I know that there is a red apple at the corner of Bush and Sansome and I tell you this and you say "oh yeah? Prove it!" I will not attempt to do so. I have given you the information but the cost of ensuring good provenance is high in an untrusted information network with few repeated interactions. So, what I do is that I attempt to repeat interactions as frequently as possible with a limited set of elements so that provenance-information is cached and I am operating on the trusted subnetwork. For my part, I try to make falsifiable claims so that other people can either add or remove me from their trust network. For HN, I have an extension that removes comment-threads started by low-information low-provenance individuals. I know I will fail sometimes at not being one of these people and I'll accept that risk and let others choose to do the same to me.