Abstract
Communication is one of the most impressive human abilities. The question of how communication arises has been studied for many decades, if not centuries. However, due to computational and representational limitations, past work was restricted to low dimensional, simple observation spaces. With the rise of deep reinforcement learning methods, this question can now be studied in complex multi-agent settings, which has led to flourishing activity in the area over the last two years. In these settings agents can learn to communicate in grounded multi-modal environments and rich communication protocols emerge.

Last year at NIPS 2017 we successfully organized the inaugural workshop on emergent communication (https://sites.google.com/site/emecom2017/). We had a number of interesting submissions looking into the question of how language can emerge using evolution (see this Nature paper that was also presented at the workshop last year, https://www.nature.com/articles/srep34615) and under what conditions emerged language exhibits compositional properties, while others explored specific applications of agents that can communicate (e.g., answering questions about textual inputs, a paper presented by Google that was subsequently accepted as an oral presentation at ICLR this year, etc.).

While last year’s workshop was a great success, there are a lot of open questions. In particular, the more challenging and realistic use cases come from situations where agents do not have fully aligned interests and goals, i.e., how can we have credible communication amongst self-interested agents where each agent maximizes its own individual rewards rather than a joint team reward? This is a new computational modeling challenge for the community and recent preliminary results (e.g. “Emergent Communication through Negotiation”, Cao et al., ICLR 2018.) reinforce the fact that it is no easy feat.

Since machine learning has exploded in popularity recently, there is a tendency for researchers to only engage with recent machine learning literature, therefore at best reinventing the wheel and at worst recycling the same ideas over and over, increasing the probability of being stuck in local optima. For these reasons, just like last year, we want to take an interdisciplinary approach on the topic of emergent communication, inviting researchers from different fields (machine learning, game theory, evolutionary biology, linguistics, cognitive science, and programming languages) interested in the question of communication and emergent language to exchange ideas.

This is particularly important for this year’s focus, since the question of communication in general-sum settings has been an active topic of research in game theory and evolutionary biology for a number of years, while it’s a nascent topic in the area of machine learning.