In the context of computational learning, the concept of 'probably approximately correct' (PAC) learning provides a framework for understanding the efficiency and feasibility of learning algorithms.
We are lonesome animals. We spend all our life trying to be less lonesome. One of our ancient methods is to tell a story begging the listener to say — and to feel — 'Yes, that's the way it is, or at least that's the way I feel it.'