Speech acts, which were first noted by Austin and which involve utterances that “do something,” are often cited by interaction designers as a way of structuring a software system’s conversational interactions. There’s a glaring problem with this approach, however. Speech act theory is a way of analyzing the particular kinds of statements and exchanges in which people do things with words. The most oft-cited example being the “I do” pronounced during the marriage ceremony. Speech acts are types of utterances that involve persons, and persons in the act of making commitments to one another, in particular. The speech act is a type of action that transforms (wherein the action occurs in) each interlocutor. That’s what makes it so interesting. The event, action, is of a psychic and relational nature.
To transfer the analytic of speech act theory onto sofware design, as a means of building software wizards, surveys, dialog boxes, action sequences and steps, etc, is to miss the point of speech act theory. Software events are verifiable. They occur within a “virtual” (though nonetheless real) system of organized information whose relations and events involve transactions that may or may not involve “real world” activity (banking, messaging, retailing, what have you). But the software with which we interact is not a person. The system does not “promise” to do what is says, or what we ask of it. The “OK,” “Submit” and many other buttons we “utter” as we click our way through an application are only analogically conversational. The software itself makes no personal commitment to us and establishes no special personal relationship to us.
The proper use of talk or conversational analysis in software/application design is with standardized speech, not with speech acts in particular. Designers fail to see that it’s not about bringing natural speech to software. What’s happening is the reverse: software brings standardized, sequenced, stepped interactions involving clear binaries (yes/no commitments) to its users. We get standardized, not the other way round!