Full Idea
Reliability is not all we want in an inductive rule. Completely reliable methods are available, such as believing nothing, or only believing logical truths. But we don't value them, but value less reliable methods with other characteristics.
Gist of Idea
Believing nothing, or only logical truths, is very reliable, but we want a lot more than that
Source
Hartry Field (Apriority as an Evaluative Notion [2000], 3)
Book Reference
'New Essays on the A Priori', ed/tr. Boghossian,P /Peacocke,C [OUP 2000], p.125
A Reaction
I would take this excellent point to be an advertisement for inference to the best explanation, which requires not only reliable inputs of information, but also a presiding rational judge to assess the mass of evidence.