Modes of Cognition: Implications for Large Language Models

In “Modes of Cognition: Implications for Large Language Models”, N. Katherine Hayles proposes a broadened framework for understanding cognition, extending it beyond conscious human thought to include implicit, nonconscious, and even non-neural processes. Using the SIRAL criteria—sensing, interpreting, responding, anticipating, and learning—she explores cognition across organisms from bacteria to plants, arguing that many life forms meet the cognitive threshold despite lacking brains. Hayles then applies these insights to large language models (LLMs), contending that their behavior satisfies the modified SIRAL criteria. Although LLMs lack physical embodiment, they navigate and interpret vast conceptual environments, produce flexible and anticipatory responses, and learn from data. Their cognitive capacities, she argues, stem not from consciousness but from their place within a broader evolutionary lineage of technics. By reframing cognition and decentering anthropocentric models, Hayles opens pathways to new ethical frameworks for engaging with artificial and biological intelligences alike.