I could not follow the whole second day of the Workshop on Complexity in Language (see previous post), but here is what I heard in the morning.
Salikoko Mufwene talked about the emergence of complexity, which he sees as a self-organization process : we don’t plan the way we are going to speak.
He adopts a relativistic perspective speaking of a multi-agent system and asking if the agents are really agentive or if there are triggers of particular behaviors. He likes to consider language as a technology that evolved. At the end of the talk he also tackled the notion of communal complexity and communal patterns used by speakers (also known as norms).
Luc Steels explained his understanding of language complexity and how he simulates communication with robots. He thinks there is an alternative to the evolutionary framework: according to him grammar is functional and not superficial and complexity has grown step by step in a cultural evolution rather than a biological.
His perception of self-organization bases most notably on alignment, structural coupling and linguistic selection. That’s what he builds models for by letting robots find common words to describe a situation (for example the fact that a given item is on the left of another one). After a communication sequence is complete, the robots show if they agree on what was said or not, it enables the research team to look for efficient communication strategies.
While the first talk was abstract, the second one included many simulation results. But since I don’t know the artificial intelligence background that was used, my personal feeling is that it’s hard to tell at first sight if nothing interferes with the experiment, as the robots do many things at a time (observing, moving, speaking and hearing). By all means it seems interesting.