[FOM] Mathematical Certainty?

David Corfield david.corfield at philosophy.oxford.ac.uk
Mon Jun 23 12:22:05 EDT 2003

Dan Goodman wrote:

> Ideally we want to make our proofs such that isolated small 
> mistakes don't ruin the entire proof. What's interesting is 
> that informal proofs are often more stable than purely formal 
> proofs because they are driven by ideas rather than purely 
> symbolic calculations, and if the ideas are right then a 
> mistake in the formalism should be easily corrected. 

Which is part of the reason why, when asked what provides 
the main evidence for the Riemann Hypothesis, mathematicians
point to a not-quite-worked out analogy to another field
rather than to the 1.5 billion verifications. Even a gappy network
of ideas can be quite convincing.

> slightly counterintuitive consequence of this is that maybe 
> we should actually prefer a slightly informal proof driven by 
> strong ideas to a formal proof checked by a computer.

An excellent way of putting it. 

Taking things a stage further, one could take proofs in a field to be 
checks that we've got the strong ideas right and linked properly.

Just one further step and:

"The quest for ultimate triviality is characteristic of 
the mathematical enterprise." (Rota, Indiscrete Thoughts, 93)

David Corfield
Faculty of Philosophy
10 Merton St.
Oxford OX1 4JJ

More information about the FOM mailing list