Command Dialogue

Abstract

We propose a representation of imperatives in computational systems, and a multi-agent dialogue protocol to argue over these. Our representation treats a command as a presumptive argument for an action to be executed by a designated agent, together with a set of associated critical questions whose answers may defeat the presumption. The critical questions enable the identification of attacks on the uttered command, and so can be used to specify a dialogue game protocol for participants to argue over the command. We present a formal syntax for part of the protocol, called CDP, and outline denotational semantics for both commands and for the protocol.