WEIZENBAUM'S PERFORMANCE AND THEORY MODES: LESSONS FOR CRITICAL ENGAGEMENT WITH LARGE LANGUAGE MODEL CHATBOTS

Abstract

In 1976, Joseph Weizenbaum argued that, because “[t]he achievements of the artificial intelligentsia [were] mainly triumphs of technique,” AI had not “contributed” to theory or “practical problem solving.” Weizenbaum highlighted the celebration of performance without deeper understanding, and in response, he articulated a theory mode for AI that could cultivate human responsibility and judgment. We suggest that, given access to Large Language Model (LLM) chatbots, Weizenbaum’s performance and theory modes offer urgently-needed vocabulary for public discourse about AI. Working from the perspective of digital rhetoric, we explain Weizenbaum’s theorization of each mode and perform a close textual analysis of two case studies of Open AI’s ChatGPT shared on Twitter to illustrate the contemporary relevance of his modes. We conclude by forecasting how theory mode may inform public accountability of AI.