Some of us sometimes make predictions with probabilities attached; does anybody here actually try to keep up a legit belief web and do Bayesian updating as the results of predictions come to pass?
Some of us sometimes make predictions with probabilities attached; does anybody here actually try to keep up a legit belief web and do Bayesian updating as the results of predictions come to pass?
No, and having a self-consistent belief net might decrease the quality of the beliefs a lot. Having multiple distinct perspectives on an issue was suggested by Tetlock to be very useful.
A Bayesian network is explicitly intended to accommodate conflicting perspectives and update the weights of two or more hypotheses based on the result of an observation. There’s absolutely no contradiction between “holding multiple distinct perspectives” and “mapping belief dependencies and using Bayesian updating”.
Some of us sometimes make predictions with probabilities attached; does anybody here actually try to keep up a legit belief web and do Bayesian updating as the results of predictions come to pass?
If so, how do you do it?
No, and having a self-consistent belief net might decrease the quality of the beliefs a lot. Having multiple distinct perspectives on an issue was suggested by Tetlock to be very useful.
A Bayesian network is explicitly intended to accommodate conflicting perspectives and update the weights of two or more hypotheses based on the result of an observation. There’s absolutely no contradiction between “holding multiple distinct perspectives” and “mapping belief dependencies and using Bayesian updating”.