Combining Different Interaction Strategies Reduces Uncertainty When Bootstrapping a Lexicon

Abstract

When bootstrapping a new language, the agents in a population need to be able to agree on the meaning of the individual words. In order to do so, they need to overcome the problem of referential uncertainty, which captures the idea that the meaning of words can not realisticly be transferred directly between agents nor through the environment. One way to reduce the amount of uncertainty, is to allow the agents, based on their current knowledge of the language system and the environment, to choose the interaction script they play based on a motivational system. We show the impact of this idea through a computational model on the time needed for a population of agents to converge on a shared language system and how the motivational system allows the agents to self-regulate this process.KeywordsConvergence TimeLanguage GameInteraction StrategyQuestion WordNaming GameThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.