Cognitive Alignment as Proto-Language: Toward a Formal Framework for Human–AI Interaction

This paper introduces and formalizes the concept of cognitive alignment as a proto-linguistic system arising from sustained iterative interaction between human agents and large-scale artificial intelligence systems. We argue that such interaction is not reducible to command-response exchange or to the established paradigms of natural language communication. Rather, it constitutes a structurally distinct phenomenon: an emergent, rule-governed, trajectory-dependent system of meaning-construction in which human intentionality and machine-generated inference are jointly productive. The paper develops this claim across four registers. First, we situate cognitive alignment within the philosophy of language, arguing that it extends Wittgensteinian use-theory and Gricean pragmatics into a novel domain while also challenging their implicit restriction to human participants. Second, we propose a formal dynamical model of cognitive alignment, specifying a state space, evolution functions, a prompt grammar, and a trajectory-based semantics. Third, we identify the criteria that distinguish a proto-language from a full language and argue that cognitive alignment satisfies the former while structurally aspiring toward the latter. Fourth, we develop the ontological implications of this framework, drawing on Heideggerian disclosure theory and extended mind theory to characterize the new epistemic and cognitive conditions that cognitive alignment introduces. We conclude by identifying six open research problems and proposing a research programme for the formal and empirical development of this framework.