Generating referring expressions with a cognitive model

The low computational cost of the incremental algorithm for generating referring expressions makes it an interesting starting point for a cognitive model. But the algorithm has a very limited cognitive adequacy; first of all because of the fact that the preference list (the order in which features are considered for inclusion in the referring expression) is fixed. This means the algorithm does not adapt to feedback of whether a referring expression was used successfully ‐ the algorithm does not learn from past experience. This paper presents a cognitive model in the ACT-R cognitive architecture that simulates the generation of referring expressions in the iMAP task ‐ a task-oriented dialogue. The model takes the feedback it receives from the simulated dialogue partner and adjusts its estimation of the utility of the features it can use in the referring expressions. A property of the task environment in iMAP is that the colour feature is unreliable for identifying referents while other features are reliable. The model correctly simulates the observed human behaviour of decreasing use of colour terms over the course of the dialogues.