Learning and Assessment in the Age of LLMs: A Distributed Cognition and Actor–Network Theory Approach

The emergence of large language models (LLMs), such as ChatGPT, Perplexity, or DeepSeek, has confronted educational systems with a paradigmatic challenge in the domains of learning and assessment. In contexts where artificial intelligence can continuously participate in the production, organization, analysis, and representation of knowledge, traditional assessment models based on memory and information retrieval face fundamental challenges. Drawing on distributed cognition theory and actor–network theory, this article argues that learning can no longer be understood solely as a process confined to the individual mind; rather, it increasingly takes shape within a socio-technical network of human and nonhuman actors—a network in which large language models function as active “cognitive companions,” rather than merely tools. The article first analyzes the relationship between classical learning theories—behaviorism, cognitivism, and constructivism—and different models of educational assessment, and then demonstrates how the emergence of LLMs may require a reconsideration of the ontological and epistemological assumptions underlying learning and assessment practices.