This paper investigates the kind of sign making that goes on in text-based human–computer interaction, between human users and chatbots, from the point of view of integrational linguistics. A chatbot serves as a “conversational” user interface, allowing users to control computer programs in “natural language”. From the user’s perspective, the interaction is a case of semiologically integrated activity, but even if the textual traces of a chat may look like a written conversation between two humans the correspondence is not one-to-one. It is argued that chatbots cannot engage in communication processes, although they may display communicative behaviour. They presuppose a (second-order) language model, they can only communicate at the level of sentences, not utterances, and they implement communicational sequels by selecting from an inventory of executable skills. Instead of seeing them as interlocutors in silico, chatbots should be seen as powerful devices for humans to make signs with.
Improving Human Computer Interaction through Spoken Natural Language
