SAYGIN, GRICE, TURING: The Implicature Game

Speranza

Jones, elsewhere: "Can you say more about the relevance of the cooperative principle?"

I think Saygin is noting that computer-generated 'conversational moves' can be tricky.

It's interesting that Turing speaks of 'games' (imitation game). So does Grice. I once counted all the references to these phrases by Grice in "Logic and Conversation"

-- 'conversational game'
-- 'conversational move'

and so on.

Now, SAYGIN seems to base his thing on the best known account of the "implicature" theory as per Essay 2 in Grice, WoW.

But, as we know, this is a specific format of the implicature theory that Grice formulated, in a 'popular' style, as it were, for his William James Lectures.

We have evidence that he was tutoring on implicature at Oxford in a perhaps more technical context, as part of a seminar for his students.

The differences seem to be that the William James version is centred around this idea that Grice is 'echoing Kant', and thus using

'conversational category'

informally. The categories being FOUR: quality, quantity, mode, and relation.

He does see that he needs like an overarching principle to organize the 'maxims' falling under these four categories. Hence his idea of the Cooperative Principle.

In the earlier lectures, he does speak of 'helpfulness', and co-ordinated activities. But, to my mind, in a freer spirit, he does not need to 'echo Kant', and thus we do not have any reference to these four rather 'fictional' categories of quality, quantity, relation, and mode.

Recall that for Kant, these four categories pertain to a different field: the analysis of propositions. Grice is playing with Kant, and using the well-known 'ontological' or 'cognitive' categories into what the calls 'conversational categories.

In the earlier seminar notes, he speaks of principles and desiderata, such as candour, and clarity.

The main idea is there, that there is helpfulness (the later technical 'cooperative principle') and desiderata ('maxims'?).

The earlier notes are not too specific about 'implicature', but he does use the expression.

The later William James lectures are specifically addressed to the idea of 'implicature', and in particular its application to 'logic'. Within logic, Grice is especially interested in refuting Strawson who had said a couple of simplistic things about the 'horseshoe'.

That's why in "Prolegomena", in WoW, Grice makes the effort to quote verbatim from Strawson's book, "Introduction to logical theory" -- the segment about how the horseshoe of the logician does NOT correspond to 'if'.

The second lecture is a general account of implicature.

The third lecture introduces the 'modified Ockham razor': there should NOT be two 'senses' of "if" (or any other item). Do not multiply sense of 'if' beyond necessity.

It's the fourth lecture that picks up Strawson's point. Grice later reprinted this lecture (1988 for the first time, when WoW went to press) as 'Indicative Conditionals'.

After that the remaining William James lectures are a 'conceptual analysis' of what it means to 'imply' as oppose to 'say', and how 'imply' can be seen as a variety of 'mean', on which Grice had proposed a rather specific thesis with his "Meaning" of 1948.

Saygin is interested in the mechanism that Grice calls 'exploitation'.

Also, Saygin is considered the more basic 'violation'.

How is a maxim violated?

In cooperative conversation between rational agents, a maxim is violated, when 'exploited'. The justification is the triggering of an 'implicature'.

Thus, if we use 'if', we may still stick to a truth-functional account of 'if' as the horse-shoe of the logician and argue that the 'implicature' of inferrability (that Strawson was interested in, and that he saw in the natural counterpart of the horseshoe) can be understood as an INTENTIONAL violation (or exploitation) of these desiderata, principles, and maxims of conversation.

Thus Grice can defend a truth-functional account of the connectives, and state that Strawson is making in mistake in ignoring the conversational factors that play into this alleged divergence between natural-language 'if' and the horseshoe.

Saygin generalizes the idea of a 'maxim violation', and seems to place the phenomenon as a mark of rationality: when the violation is perceived as intentional by the co-conversationalist, for the purpose of generating an implicature, it is justified.

Grice is tricky here. He seems to be saying that a maxim may be 'violated' (exploited) at the level of what is EXPLICITLY communicated (or said), but NOT at the level of what is IMPLICATED.

He may also say that while the maxim is violated intentionally (exploited), the overarching cooperative principle IS NOT.

The idea of having a general principle and more specific maxims is of course Kantian in spirit. But Kant is never too clear as to what his maxims are, and Grice will pick up the issue in his more technical seminars on Kant's ethical theory at Berkeley.

I don't think THIS Kantian side to Grice interests Saygin -- but the underlying point that Saygin, Grice, and Kant -- and why not Turing? -- may share pertains to to how this 'illustrates' how 'rationality' works in conversation.