I collaborated with Shahil Patel, under the mentorship of Dr. Simon Todd, to develop computational simulations that examine how word frequency influences stability and change in exemplar-based models of language evolution. Together, we designed two contrasting architectures from the literature—one in which memory traces of words are overwritten when new experiences occur, and another in which they decay gradually over time—and derived the mathematical formulations needed to track exemplar means and variances over time. Our goal was to explain why different studies report conflicting frequency effects and to identify the specific modeling conditions that produce these discrepancies.

After co-developing the model design, Shahil implemented the overwriting model and carried out its initial analysis. I implemented the full decay model and built the Python pipeline for 10,000+ iteration simulations to analyze how decay interacts with frequency and exemplar strength. I optimized the runtime from roughly 120 minutes to ~30 seconds by restructuring parallelization and file-writing, which made large parameter sweeps feasible. I conducted the simulation-based analyses using Python and R, comparing how low- and high-frequency words diverged across different total exemplar strengths and documenting when frequency effects strengthen, weaken, or disappear.

I created the decay model section of the poster so that it reflected the decay results, and I produced the final suite of figures for the decay models and wrote the comparative explanations showing that overwriting erases most frequency differences, while decay generates strong frequency effects only when the total exemplar strength is sufficiently small. In collaboration with Professor Todd, we wrote the introduction and conclusion together.

These results formed our CAMP[7] poster, which demonstrates that frequency effects in exemplar models are highly architecture-dependent and parameter-sensitive—a finding that helps reconcile conflicting claims in the literature about how frequency shapes linguistic change. We are currently working on turning our findings into a paper for conference and journal submissions.