I am curious about the dark room problem. Even if we accept surprise minimization, the brain developed in an environment where a dark room was generally not an option: the sun came up, night fell, the weather changed. Aside from the external environment, there remains the problem of thirst and hunger.
I also note that we did go to a lot of trouble to build as many rooms as possible with constant light, constant temperature, and constant humidity. Locking oneself alone in a dark room and not coming out is the archetype of depression, and depressed people are less prone to optimistic biases—this seems like it would correlate to ‘more successful in resolving their surprise’. Depression: the state where no stimulus is a superstimulus.
What about surprise being based on a pattern of change rather than a state? I imagine if the sun failed to rise one morning the freakout rate would approach 100%. I also note that people get into habits, and become attached to them even if they are bad. This amounts to a predictable change in stimuli we impose through action. If we peg to the expected change in stimuli, then a constant stimuli is the case where the expected change is 0.
In line with these thoughts, it initially appears to me that the dark room problem is not an actual problem; the distinction between accuracy-maximization and surprise-minimization is unclear to me.
I am curious about the dark room problem. Even if we accept surprise minimization, the brain developed in an environment where a dark room was generally not an option: the sun came up, night fell, the weather changed. Aside from the external environment, there remains the problem of thirst and hunger.
I also note that we did go to a lot of trouble to build as many rooms as possible with constant light, constant temperature, and constant humidity. Locking oneself alone in a dark room and not coming out is the archetype of depression, and depressed people are less prone to optimistic biases—this seems like it would correlate to ‘more successful in resolving their surprise’. Depression: the state where no stimulus is a superstimulus.
What about surprise being based on a pattern of change rather than a state? I imagine if the sun failed to rise one morning the freakout rate would approach 100%. I also note that people get into habits, and become attached to them even if they are bad. This amounts to a predictable change in stimuli we impose through action. If we peg to the expected change in stimuli, then a constant stimuli is the case where the expected change is 0.
In line with these thoughts, it initially appears to me that the dark room problem is not an actual problem; the distinction between accuracy-maximization and surprise-minimization is unclear to me.