Gears (hypotheses, paradigms, theorems, narratives) have a lot of internal certainty and robustness in the face of confusion about priors, even absence of their instances in reality is only a problem of usefulness/relevance, not validity. When talking about the world, summaries of observations such as likelihood ratios are still more robust than beliefs or deeper theoretical interpretations of observations (for example bug reports written by non-programmers make less sense when they don’t stick to communicating direct observations).
There is no point to accepting beliefs into one’s identity, instead they can be more like cached results of frequently issued queries, something occasionally recomputed from the gears and observations, not meaningfully changed in their own right. It’s the changes in the available gears, in understanding, that make this happen, the changes in beliefs are merely symptoms of this process.
Gears (hypotheses, paradigms, theorems, narratives) have a lot of internal certainty and robustness in the face of confusion about priors, even absence of their instances in reality is only a problem of usefulness/relevance, not validity. When talking about the world, summaries of observations such as likelihood ratios are still more robust than beliefs or deeper theoretical interpretations of observations (for example bug reports written by non-programmers make less sense when they don’t stick to communicating direct observations).
There is no point to accepting beliefs into one’s identity, instead they can be more like cached results of frequently issued queries, something occasionally recomputed from the gears and observations, not meaningfully changed in their own right. It’s the changes in the available gears, in understanding, that make this happen, the changes in beliefs are merely symptoms of this process.