Grice comes from a philosophical tradition and he uses "cooperation" in a specific sense. Davies  argues that with the use in linguistics there has been what he calls the Cooperation Drift, i.e. there is a confusion between the common and the Gricean technical usage of the notion of "cooperation" and the Cooperative Principle (CP). In common language cooperation means something like "the action or process of working together to the same end" (Oxford Dictionary of English, 2nd edition 2003). However, the Gricean technical "cooperation" does not state that conversation is cooperative, but rather that in the process of calculating the meaning from the spoken, the hearer may take into account violations of the Cooperative Principle; i.e., the theory of implicatures. E.g. Davies has an example of flouts. This is not cooperative, but very well fits the model of interpretation provided by the CP.
Davies is entirely right in this, and indeed in our own work, being even further removed from philosophy than linguistics, the use of the CP is quite different from Grice's which makes confusion easier. Davies uses [Bernsen et al. 1996a] as an example of the confusion, see below. He notes rightly that we use the principle contrarily to how Grice uses it, but seemingly assumes that we then also understand the CP this way. This is not correct. We only claim that the CP can be read and used in this opposite way, within certain limits. These limits include that we deal only with task-oriented dialogues where user and system share a common goal. Flouts and exploitations cannot be said to be rational in this kind of dialogue.
A few more comments on his points concerning our use of Grice (p 6):
Bernsen et al.’s error seems to be the assumption that if failure to adhere to the maxims may lead to clarification/repair (which may or may not be true), then adhering to the maxims will necessarily avoid this outcome. This is flawed reasoning.Yes, but we only claim that adherence to the maxims are, in task-oriented human-computer spoken, unimodal dialogue, necessary to minimize miscommunication, not that is is sufficient, nor that it is at all possible to completely eliminate miscommunication.
Firstly, at a purely practical level, what may seem explicit and obviously clear to the speaker may not be so for the hearer: there seems to be too great an assumption of shared knowledge/common ground here.One of the results in the paper is precisely that we need design guidelines concerning the shared knowledge. Several such guidelines are found to be needed in addition to the maxims, e.g. P4, P7, P8, and P9 (today numbered as GG11, SG6, SG5, and SG7, respectively). These are not explicit in the maxims even though they, too, may have implications for the semantic interpretation. In the case of human-computer communication, these are highly important considerations because systems are very inept at decoding the users' background and thus one fixed dialogue design must fit every user, possibly with a few attempts at variations that fit expected, high-frequent or important user sub-groups.
Secondly, Grice makes no claim that following the CP will improve or enhance conversation: at a basic level, the CP is simply a description of what happens.Agreed. What we claim is that based on empirical studies of a concrete spoken dialogue system, we developed a set of design guidelines that serve to reduce the communication problems of that system. We further claim and prove in the paper that the maxims can be related to meta-communication and that they match (with some explained exceptions) our principles.
Finally, at a more theoretical level, it seems to have been forgotten that orienting towards the CP does not entail being explicit: flouts are exploitations of the CP, but they still come under its aegis.Concerning explicitness, it is part of some of our principles, but not all of them, and we do not claim that it is entailed by the CP. However, explicitness in the sense of "say enough" is important, and is in human-computer dialogues weighted somewhat higher when balanced against "don't say too much" (don't be wordy) than in human-human dialogue.
Let me finally cite:
When I re-read this, I don't think our and Davies' interpretations are that far apart, even if it at times may be confusing that [Bernsen et al. 1996a:224] not always speak of cooperativity in the technical sense of Grice and maybe are not fully explicit about the difference.
Grice claimed that adherence to principles, such as the CP and the maxims, is rational in the sense that anyone who cares about achieving the goals that are central to the dialogue must be expected to have an interest in conducting the talk exchanges in accordance with those principles. We accept this claim.
Violation of our principles would normally lead the interlocutor to ask questions of the speaker. Grice did not consider such cases of communication failure and it is clear that he has not developed the CP and the maxims in order to help preventing partner-initiated metacommunication. Rather, his purpose was to investigate conversational implicature. Conversational implicature is a phenomenon in human-human dialogue in which the speaker manages to imply, or suggest, unstated information under the general assumption of adhering to the CP, and manages to be understood by the interlocutor as doing so. In understanding conversational implicature, the interlocutor is guided by inferences based on the CP and the maxims, according to Grice
The assumption of perfection leads to the assumption of miscommunication avoidance. In a paper about Human-Computer dialogue, Bernsen, Dybkjer & Dybkjer (1996) describe a dialogue system which is designed in order to avoid repair and clarification sequences as much as possible, because these are notoriously difficult to deal with in the context of Natural Language Processing. They state:
“A crucial point in what follows, however, is that system-dialogue breaks down when users ask questions of the system. A key, therefore, to the successful design of system-directed dialogue is to design the dialogue in such a way that users do not need to ask questions of the system. To do this requires optimizing the dialogue cooperativity of the system.”Bernsen et al. (1996:214)
Note the use of the phrase ‘dialogue cooperativity’ here. Cooperation is being marked as equivalent to being totally explicit. (NOTE: We do not here state anything about being totally explicit, but about avoiding the need for user questions. More importantly, we do not write that this is equivalent to dialogue cooperativity, but that dialogue cooperativity is required in order to reduce the need for user questions. See also comments on explicitness above) Again, we have the confusion between technical and non-technical uses of the term. A non-technical use of the term may imply this type of meaning (although this is far from certain). This is then related to the CP:
“Although Grice’s maxims have been conceived with a different purpose in mind, they can be seen as serving the same objective as do our principles, namely that of preventing interlocutor-initiated clarification and repair metacommunication.”Bernsen et al. (1996:215)
“We conclude that the CP and the maxims, as a necessary side effect of improving understanding and enhancing communication, serve the purpose of preventing the need for clarification and repair metacommunication.”Bernsen et al. (1996:225)
Here, now, the CP is directly linked with miscommunication-avoidance (NOTE: Yes, this is the whole point of the paper, that the CP and its maxims can be utilised this way; for the following, see my discussion above). The problem here seems to be that orienting to the CP is taken to be equivalent to being explicit, when Grice’s insight is, in fact, concerned with precisely the opposite situation. Grice is concerned with the distinction between saying and meaning: how hearers recognise the speaker’s intention when the speaker uses implicit language. Bernsen et al.’s error seems to be the assumption that if failure to adhere to the maxims may lead to clarification/repair (which may or may not be true), then adhering to the maxims will necessarily avoid this outcome. This is flawed reasoning. Firstly, at a purely practical level, what may seem explicit and obviously clear to the speaker may not be so for the hearer: there seems to be too great an assumption of shared knowledge/common ground here. Secondly, Grice makes no claim that following the CP will improve or enhance conversation: at a basic level, the CP is simply a description of what happens. Finally, at a more theoretical level, it seems to have been forgotten that orienting towards the CP does not entail being explicit: flouts are exploitations of the CP, but they still come under its aegis.
However, Bernsen et al. are unashamedly putting a theoretical principle to practical use: they are quite upfront in stating that they are using those aspects of Grice’s work which they perceive as useful. They would not claim to be operationalising the CP (as we have tried to do elsewhere, e.g. Davies 1997), or testing Grice’s work in any particular way, and perhaps these factors should be borne in mind when considering this paper. However, what we are concerned with are representations of Grice’s CP in the literature, and how all these problematic interpretations can have an additive effect on the overall perception of Grice’s work. This paper, however well-intentioned, does affect this perception.