learning about the real world

The capability of embedding intentions to arbitrary depth subsumes the capability for modeling the world, thence for avoiding some peril – the capability of having ‘agents die in one’s stead’, as Dennett puts it.   One cannot precompute the world in its entirety, howeve, which leads to the inevitable conclusion that all learning that is not merely discovery of what was implicit within one’s model begins with a negative outcome for the agent and therefore entails risk.  The self-reflective agent can, through foresight, to some degree order situations so that learning by experience is possible for while minimizing risk – this is indeed one of the key exercises of foresight – but the threat to the agent cannot wholly be eliminated, for just the reason that has already been stated: the agent cannot precompute the world, and engaging with the part of the world the agent has not precomputed means engaging with the unforeseeable.

Beyond the fact that it begins with a negative outcome for the agent, learning about the real world entails a preparedness to adapt in response to that negative outcome (assuming it does not destroy the agent).  The Nietzchean aphorism “what does not kill us makes us stronger” is of course demonstrably false if ‘stronger’ is taken to mean mere physical strength, and even if it is taken to mean the power that comes from having knowledge, acquiring such power is contingent upon the agent’s being disposed to change his, her, or its models in response so as to anticipate an extension of ‘like cases’ that enhances the agent’s real prospects for survival.   Fortunately, this disposition can in some degree be cultivated from within the stream of narrative:  it is a function of and consequence of the modeling processes and representations which constitute the agent’s identity or ‘self’, and is to this extent a matter of self-determined praxis.