The Effect of Efficient Messaging and Input Variability on Neural-Agent Iterated Language Learning
This work addresses a gap in understanding language evolution simulations for computational linguistics and AI researchers, but it is incremental as it builds on prior findings without introducing new methods.
The study investigated why neural network agents in iterated language learning fail to develop a trade-off between syntactic strategies like word order and inflection, unlike natural languages, by examining factors such as efficient messaging bias, input variability, and learning bottlenecks. The simulations revealed that neural agents primarily preserve the utterance type distribution from training rather than evolving more efficient or systematic languages.
Natural languages display a trade-off among different strategies to convey syntactic structure, such as word order or inflection. This trade-off, however, has not appeared in recent simulations of iterated language learning with neural network agents (Chaabouni et al., 2019b). We re-evaluate this result in light of three factors that play an important role in comparable experiments from the Language Evolution field: (i) speaker bias towards efficient messaging, (ii) non systematic input languages, and (iii) learning bottleneck. Our simulations show that neural agents mainly strive to maintain the utterance type distribution observed during learning, instead of developing a more efficient or systematic language.