Intention recognition became a research topic in computer science through Searle’s work on speech acts, and the various attempts by the artificial intelligence (AI) community to formalize speech act theory. Much of the research on intention recognition is therefore connected to the area of natural language processing. Kanno, Nakata, and Furuta depart from this tradition in two respects. Their work targets the area of computer-supported collaborative work and human-computer interaction (HCI), rather than natural language. Furthermore, their intention inference method is based solely on nonverbal actions, whereas methods based on speech act theory, as the name implies, focus primarily on linguistic input. The main contribution of the paper is its procedural analysis of joint intentions under those specific constraints.

The method is illustrated by a study of intention recognition in a collaboratively operated plant simulator. Evaluation is done by checking intentions attributed by the system to the group against intentions attributed to it by a human observer. Although the method performs well in this restricted scenario, it is not obvious that comparable performance can be achieved in realistic situations.

The authors explain the method fairly clearly, with the exception of the relationship between the logical analysis of team intention presented at the beginning of the paper and the procedural inference technique described toward the end. Readers with an HCI background and an interest in decision support systems are likely to find this paper quite useful, even though its contribution to the state of the art in AI is limited.