Rational Communication

by paulfchristiano 8y11th Apr 20113 comments


Hopefully LW trains its readers to be aware of personal failures of rationality and to correct them over time.

Whenever a small group (all the way down to 2) of rationalists communicate, they should make an equal effort to be aware of failures of their collective rationality (I definitely have failed at this so far in life). In particular: time is valuable, and you can hope to actually learn things from other people. Here are some remarks about communication intended to help you optimize (as opposed to communication intended to help you be happy).

1. Communicate information precisely. I have strong beliefs that contradict the beliefs of other rationalists. I need a better way to express precisely what the disagreement is and my confidence in my beliefs, particularly with people I don't know well. Right now I feel like I have no hope of correctly updating on the beliefs of others or sharing information in a productive way. To this end, I think making precise testable claims about the future and backing them up with probability estimates (as a regular feature of discussion) would be a good idea, even when your probability estimates are horribly miscalibrated. This would help me much more quickly learn about my own miscalibration, evaluate the miscalibration of others, and at least have a definite scale to make precise communication possible.

2. Communicate preferences precisely. Groupmembers often have strong preferences which they fail to coordinate well. I think groups I have been in would have done better if they consistently used any precise semantics for expressing preferences; for example, assigning monetary values or using some other fixed reference commodity. A bigger problem is that statements of preference carry way to much additional meaning in normal communication. I'm not sure how to divorce statements of preference from this other meaning, but its worth thinking about.

3. Explicitly think about the point of communication. There are important things to talk about, and the marginal benefits of spending 10 minutes on one topic versus another can be quite different. There are a lot of free moments in a day, and I think this calculus is worth spending some of them on. I would like to start directing my conversations towards high value topics much more reliably, especially in conversations with rationalists. This may involve doing things like making explicit notes to structure my conversation. 

4. Talking about procedure is OK. Improving your rationality is worth spending time on, and I think improving the quality of your communication and a community's ability to communicate is worth spending time. Entertain some discussion of process without feeling irresponsible for not moving forward on the meat of whatever issue you are discussing. Perhaps make rational communication the meat of discussion more often.

5. Acquire a lot of data. The effectiveness of strategies for communication or group rationality is a really easy thing to gather data on, at least in principle. Share data (doing this effectively requires improvements in communication).