Alignment is the solution, not the problem: A developmental mechanism for collective intelligence
Abstract
As efforts to build machines with human-level intelligence accelerate, most continue to approach intelligence as a property of individual agents. Human cognition, in contrast, is widely understood to arise through interactions with sociocultural environments. Yet the mechanisms by which such interactions generate our species’ powerful and open-ended cognitive flexibility remain poorly specified. This talk offers a new theoretical framework, proposing that human intelligence develops through a process that instills, from the outset, a drive to align representations across interacting minds, transforming individual learning into a population-level system of collective intelligence. We argue that infants’ precise social tuning leads them to infer minds as latent causes of communicative behavior. The alignment dynamics that emerge from mind inference create a selection pressure on representational form: representations that are compressible, interpretable, and shareable are preferentially stabilized across heterogeneous populations. Amplified by cultural technologies, this process drives abstraction, continual learning, and cumulative knowledge. Communicative Drive Theory (CDT) reframes human intelligence as an emergent property of populations and suggests that alignment, far more than being a constraint for artificial systems, is instead a developmental mechanism underlying collective intelligence.