An information-theoretic account of semantic interference in word production

TL;DR

This paper introduces a computational model based on information theory to explain semantic interference in word production, successfully replicating key experimental patterns and offering a high-level formalization of underlying computational principles.

Contribution

It presents a novel rate-distortion framework for modeling semantic interference, integrating perceptual input and behavioral goals in a unified computational account.

Findings

Model captures Stroop and Picture-Word Interference effects

Similarity-based interference emerges naturally in the model

Provides a high-level formalization of computational principles

Abstract

I present a computational-level model of semantic interference effects in word production. Word production is cast as a rate-distortion problem where an agent selects words to minimize a measure of cost while also minimizing the resources used to compute the output word based on perceptual input and behavioral goals. I show that similarity-based interference among words arises naturally in this setup, and I present a series of simulations showing that the model captures some of the key empirical patterns observed in Stroop and Picture-Word Interference paradigms. I argue that the rate-distortion account of interference provides a high-level formalization of computational principles that are instantiated more mechanistically in existing models.