Emergent Linguistic Phenomena in Multi-Agent Communication Games

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 3700–3710, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 3700 Emergent Linguistic Phenomena in Multi-Agent Communication Games Laura Graesser†∗, Kyunghyun Cho†‡⋆, Douwe Kiela‡ † NYU; ∗ Robotics at Google; ⋆ CIFAR Azrieli Global Scholar; ‡ Facebook AI Research lauragraesser@google.com, kyunghyun.cho@nyu.edu, dkiela@fb.com Abstract We describe a multi-agent communication framework for examining high-level linguis- tic phenomena at the community-level. We demonstrate that complex linguistic behavior observed in natural language can be repro- duced in this simple setting: i) the outcome of contact between communities is a func- tion of inter- and intra-group connectivity; ii) linguistic contact either converges to the ma- jority protocol, or in balanced cases leads to novel creole languages of lower complexity; and iii) a linguistic continuum emerges where neighboring languages are more mutually in- telligible than farther removed languages. We conclude that at least some of the intricate properties of language evolution need not de- pend on complex evolved linguistic capabili- ties, but can emerge from simple social ex- changes between perceptually-enabled agents playing communication games. 1 Introduction Contact linguistics (Myers-Scotton, 2002) studies what happens when two or more languages or lan- guage varieties interact. It poses several pertinent open questions that are difficult to answer: how does symmetric (“mutually intelligible”) commu- nication emerge; how do languages behave under contact; how does one language come to dominate another; how and why does extensive language contact tend to lead to simplification (e.g. in cre- oles); and how does a linguistic continuum come about, where neighboring languages are more in- telligible than farther removed ones? In this work, we show that such linguistic phenomena emerge naturally given a few general assumptions about the organizational structure of networks of artifi- cial agents equipped with a minimalistic form of learned communication. We introduce a multi-agent framework for studying the emergence and evolution of language, where agents are neural networks endowed with the ability to exchange messages about their per- ceptual input. The advantage of this approach is that one can precisely control linguistic, environ- mental and algorithmic variables. First, we investigate linguistic behavior at the agent-level, and examine when symmetric com- munication emerges within a linguistic commu- nity, as well as how the topological organization of communities—i.e., which other agents an agent comes into contact with and how frequently that happens—impacts convergence and learning. We then examine the behavior of communities of such agents when they come into contact, as well as how community-level topology impacts conver- gence, success rate and mutual intelligibility. We demonstrate that the following linguistic be- haviors emerge, which correspond to known lin- guistic phenomena in natural languages: 1) the outcome of contact is a function of inter- and intra-group connectivity, i.e. that languages be- come mutually intelligible through contact, even for agents that have not themselves been exposed to the other language, provided there is sufficient connectivity between communities; 2) linguistic …

Content truncated to 500 words