AI and Pragmatics: Do Chatbots Follow Speech Acts Maxims? Evaluating AI-Generated Conversations Using Pragmatics

The increasing use of AI chatbots in human communication raises critical questions about their ability to adhere to pragmatic principles, particularly Speech Act Theory (Searle, 1969) and Grice’s Cooperative Principle (Grice, 1975). This study investigates whether AI-generated conversations successfully produce appropriate speech acts and adhere to Grice’s maxims. Furthermore, it examines how these AI-generated dialogues compare to human conversations in terms of pragmatic competence. A dataset consisting of 120 AI-generated responses (from ChatGPT, Google Bard, Alexa, Siri, and Cohere) and 120 human conversations was analyzed to assess speech act distribution, adherence to conversational maxims, and pragmatic inconsistencies. The conclusions have shown that AI chatbots overuse Representative speech acts (factual statements) and Directives (commands, requests, refusals) unlike Expressivesunlike Expressives (apologies, gratitude, humor) and Commissives (promises, commitments). Thus, AI-generated conversations seem mechanical as well as they lacks both emotional intelligence and relational depth. Due to Grice’s maxims, frequently AI responses violate Quality (false or unverified information), Relation (irrelevance), Manner (ambiguity), and Quantity (either over-explaining or omitting key details) maxims. Contrastively, human speakers balance speech acts naturally. They adhere to conversational maxims to ensure clarity, coherence, and engagement. Although AI has advanced in syntactic fluency, the findings have shown, there is a lack in pragmaticin pragmatic adaptability and contextual awareness. In other words, this case led has to dialogic breakdowns and user’s reduction engagement. Accordingly, enhancing the AI’s pragmatic competence, focus should be on the expansion of speech act diversity, minimizing maxim violations, as well as on improving conversational memory for contextual adaptation.