8. What Counts As Refinement?

Recursive examination can change a conviction.

But we do not yet know whether these changes have enough in common to call them refinement. We also do not know whether refinement should mean improvement.

A change in conviction formation may make something more explicit without making it better.

It may make consequences easier to reconstruct while leaving a poor representation untouched. Alternatives may remain available, admissible moves may become restricted, or disagreement easier to locate. The examination may also reveal that the conditions already supplied do not determine the answer we wanted.

These possibilities do not yet give us one scale on which conviction formation can be ranked.

And asking which of them counts as improvement immediately returns us to the problem already encountered.

Which criterion would decide: coherence, empirical success, logical validity, predictive success, causal adequacy, or something else?

Here the fundamental claim applies again. We cannot simply choose which criterion will become convincing. A criterion can become sufficiently stable to guide inquiry. It can be compared with others, applied, challenged, defended, or later reopened. But calling something a criterion does not place its convincing force outside conviction formation.

This gives Part II a more specific task.

Rather than decide in advance what refinement must be, we can examine some strong methods of constraint and ask what they actually change.

What has to be supplied before their constraint can operate? What becomes no longer independently movable once enough has been fixed, and what remains open?

Where can disagreement move? What becomes available for further examination?

Logic, probability theory, and causal reasoning provide useful cases.

They are widely regarded as strong tools for constraining thought and inquiry.

They can also be examined from within conviction formation.

Their representations, assumptions, procedures, and consequences can themselves become objects of inquiry. We do not yet know whether examining them will leave their convincing force unchanged, weaken it, strengthen it, or make it more conditional.

Next essay: Sources Of Logical Constraint