Not incompatible to view belief as an optimization while also realizing our measurement of the truth function is biased and noisy (with everyone measuring getting different biases).