As my final year project at IIT Kanpur, I studied the emergence of compositional language in social communities by extending an inductive learning model of language
learning to large populations, heteregeneous interactions, and realistic social communitites. The model induces grammar rules on the basis of phonetic resemblances
between lexical entities, and similarities in semantic meanings they correspond to.
We developed a framework where multiple agents can interact in an iterated learning setting, where each agent can receive its primary linguistic input from a set of speakers in the previous generation according to probability distributions specified by the existing social topology. We also explore a probabilistic production model as compared to a deterministic one, and investigate possible biases which can expedite the emergence of compositional syntax.
This perspective views language as a continually evolving complex system (similar to an economy or an ecosystem) where local interactions lead to global organization, and the onus of adaptation is on language itself to be transmissible; within the constraints of available cognitive architectures, limited communication channels, and noisy signal perception. In this sense, fundamental features of language such as compositionality are seen as emergent solutions in response to the system constraints and initial conditions.