"LazImpa": Lazy and Impatient neural agents learn to communicate efficiently

TL;DR

LazImpa demonstrates that by making neural agents lazy and impatient, they can develop more efficient communication codes that align with natural language patterns like Zipf's Law of Abbreviation.

Contribution

This work introduces LazImpa, a novel neural communication framework where agent behaviors are modified to produce more natural and efficient language-like messages.

Findings

Emergent messages become more efficient and Zipf-like.

Lazy and impatient agents produce shorter, more natural messages.

Communication efficiency improves with agent behavior modifications.

Abstract

Previous work has shown that artificial neural agents naturally develop surprisingly non-efficient codes. This is illustrated by the fact that in a referential game involving a speaker and a listener neural networks optimizing accurate transmission over a discrete channel, the emergent messages fail to achieve an optimal length. Furthermore, frequent messages tend to be longer than infrequent ones, a pattern contrary to the Zipf Law of Abbreviation (ZLA) observed in all natural languages. Here, we show that near-optimal and ZLA-compatible messages can emerge, but only if both the speaker and the listener are modified. We hence introduce a new communication system, "LazImpa", where the speaker is made increasingly lazy, i.e. avoids long messages, and the listener impatient, i.e.,~seeks to guess the intended content as soon as possible.