Ambiguity can compensate for semantic differences in human-AI communication

Abstract

Ambiguity and semantic differences are each known to be independent sources of communication difficulty. However, we show using computational models that ambiguity can compensate for semantic differences across communicators. Given the heterogeneity of humans with which artificial systems interact, semantic differences will be the norm. Therefore each time a machine starts to communicate with a new user, our results suggest it will do well to start with a moderately ambiguous code in order to more effectively bridge semantic differences. We dub this the “adaptive ambiguity” hypothesis.

Keywords

• Collaboration theory
• Communication
• Communication codes
• Computational modelling