When there are conflicts of observation, when experiments cannot be replicated, scientists may then retreat to a study of the various specific observations so as to explain the conflict, in the course of which they would make use of the concept of observation, or of some specification of that concept.

We need some standard that will determine how likely a belief is to be true given just that it is stored in one of us, including strangers that one can ask for directions, and with whom one might collaborate.

Any defense of common sense must allow that it is revisable.

When there are conflicts of observation, when experiments cannot be replicated, scientists may then retreat to a study of the various specific observations so as to explain the conflict, in the course of which they would make use of the concept of observation, or of some specification of that concept.

The success of an archery shot may bring food to the hunter's starving family, or may constitute a horrible murder. But these outcomes are irrelevant to the assessment of that shot as a hunter-archery shot, as an attempt to hit prey without running excessive risk of failure.

Lowered reliability obviously yields a lesser competence. But lowered breadth does so as well.

One does not avoid incompetence if one makes an attempt whose likelihood of success is too low. This seems little more than analytic: when the performance is in a domain that imposes standards of risk, attempts may or may not meet such standards. And the relevant competence of agents then includes reliably enough meeting those standards.