Dynamics of Spontaneous Topic Changes in Next Token Prediction with Self-Attention
This work addresses a fundamental disparity in AI behavior for cognitive modeling and neuroscience, but it is incremental as it builds on existing self-attention theory.
The paper tackles the problem of spontaneous topic changes in self-attention models, showing that they differ from human cognition by being less likely with longer contexts or ambiguous inputs, and validates these findings in modern LLMs.
Human cognition is punctuated by abrupt, spontaneous shifts between topics-driven by emotional, contextual, or associative cues-a phenomenon known as spontaneous thought in neuroscience. In contrast, self-attention based models depend on structured patterns over their inputs to predict each next token, lacking spontaneity. Motivated by this distinction, we characterize spontaneous topic changes in self-attention architectures, revealing both their similarities and their divergences from spontaneous human thought. First, we establish theoretical results under a simplified, single-layer self-attention model with suitable conditions by defining the topic as a set of Token Priority Graphs (TPGs). Specifically, we demonstrate that (1) the model maintains the priority order of tokens related to the input topic, (2) a spontaneous topic change can occur only if lower-priority tokens outnumber all higher-priority tokens of the input topic, and (3) unlike human cognition, the longer context length or the more ambiguous input topic reduces the likelihood of spontaneous change. Second, we empirically validate that these dynamics persist in modern, state-of-the-art LLMs, underscoring a fundamental disparity between human cognition and AI behaviour in the context of spontaneous topic changes. To the best of our knowledge, no prior work has explored these questions with a focus as closely aligned to human thought.