Them incorrigible pirots

Grice wants his pirots to have privileged access to things and be incorrigible.

Grice writes: "We shall not want pirots to depend, in reaching second order
beliefs and wants about themselves on the observation of manifestational
behaviour. So the genitor makes these pirots subject to the law that
ceteris paribus,

PSI(PIROT, p) -cp-> BEL (PIROT, (PSI(PIROT, p))

("if a pirot wants/believes that p, ceteris paribus, the pirot believes
that the pirot wants/belives that p". "To build in this feature IS to build
in "PRIVILEGED ACCESS" to beliefs and wants." "Further, to minimise the
waste of effort which would be involved in trying to suppress a WANT which
a pirot mistakenly BELIEVES himself to have, the genitor may also build in
conformity to the converse law, that, ceteris paribus,:

BEL (PIROT, (PSI(PIROT, p) -cp-> PSI(PIROT, p)

"if a pirot believes that it wants/believes that p, ceteris paribus, the
pirot wants/believes that p." "These are ceteris paribus laws and there
will be room for counterexamples: in SELF-DECEPTION, either law may not
hold". Grice then abbrevaites

PSI-2 (PIROT, p) <-cp->df. PSI-1 (PIROT, p)

"Let us suppose that we make the not implausible assumption that there will
be no way of finding NON-LINGUISTIC manifestational behaviour which
distinguishes PSI-3 from PSI-2." Linguistically, there could be
distinguishing manifestational behaviour. "A Talking Pirot can express that
he PSI-2-s that p. "We can maintain as a general law that ceteris paribus,

EXPRESS (PIROT, (PSI-1 (PIROT, p)) -> BEL(PIROT, (PSI-l (PIROT, p))

"If the pirot expresses that the psi-s that p, the pirot believes that he
psi-s that p". "We get the result that PSI-(n>1) are INCORRIGIBLE, whereas
PSI-n(n=1) are only matters of privileged access".

D. F. Pears complained: "This is as beautiful as you can get. But can a pirot self-deceive?" Grice replied, "I have a train to catch".