Disimplicature

Speranza

Sorensen says that Grice's assertion that p is a knowledge-lie if it is meant to undermine your justification for believing truly that ∼p, not to make you believe that p and that, therefore, knowledge-lies are not intended to deceive. 

It has been objected that they are meant to deceive because they are intended to make you more confident in a falsehood. 

But a Griceian might propose a novel account according to which an assertion that p is a knowledge-lie if it is intended not to provide evidence that p but to make you stop trusting all testimonies concerning whether p, which is how they undermine your testimonial knowledge. 

Because they are not intended to provide evidence that bears on the truth of p, they are not intended to make you more confident in a falsehood; therefore, knowledge-lies are not intended to deceive. 

This makes them a problem for the traditional account, which takes the intention to deceive as necessary for lying, and an interesting example of Kant's idea that allowing lies whenever one feels like it would bring it about that statements in general are not believed.

But Grice thought, rightly, that Kant was too severe -- he preferred Kantotle!