15.12.2011. FSP 7 (Prof. Dr.-Ing. Sebastian Möller)

Dr. David Reitter, Research Psychologist, Department of Psychology Carnegie Mellon University

Natural language, like all forms of communication, is an instinct constrained by cognitive function. At the same time, languages have evolved through a social process, in which adaptation plays a central role. Adaptation has been suggested to be the key to mutual understanding in the short term, that is, in dialogue between humans or in human-computer interaction.

Linguistic adaptation can be measured in large datasets of spoken language. One finding relevant for human-computer dialogue systems is that we found adaptation to be positively correlated with task success. We describe a cognitive model (ACT-R), which explains such adaptation as a result of memory decay and retrieval. Scaling up to groups of individuals, we then show the propagation of information and communication standards in groups. Here, the role of communicative strategies - promiscuous propagation vs. selective relaying - suggests an interaction of communication patterns with the nature of human memory: knowledge can be held in the individual, or externalized into the network. This, again, is shown empirically through a multi-participant experiment as well as using networked cognitive models. Scaling up to the case of thousands of communicating humans, we can derive predictions about whether human memory may be uniquely adapted to the social structures prevalent in communities, such as small world networks.