Strangely reminds me of machine learning:

   1. We have some initial beliefs
   2. We use it to make decisions
   3. We receive some new data
   4. Given the new data, we confirm/change the beliefs
   5. We can make different decisions
   6. Go to 3.

The initial beliefs may or not be true, the key to get as much accurate data as fast as possible.

A better algorithm would be to ask other systems beliefs and compare it to ours too (eg: neuronal network).