Scientific report

 

Human-machine interaction is increasingly important in domains where artificial systems are expected to take up tasks, take initiatives, make decisions, and coexist and collaborate with people. This implies a need to understand the consequences of the interaction (cooperation, competition, coordination) between people and robots or agents. The aim of HART (Workshop on Human-Agent-Robot Teamwork) was to create a roadmap of the relevant research questions and topics for the coming two, five and ten years. The workshop brought together an excellent mix of senior researchers and young PhD students. The lectures provided by the senior researchers were an immense bonus for the young PhD students to help them understand more about the underlying problems and the state of the art in Human-Agent-Robot Teamwork. The talks by the PhD students gave them the opportunity to discuss their research plans and ideas with the more experienced researchers. The feedback they received will have a clear impact on the focus of their research and the approach they will follow in their projects.

 

During the workshop a number of topics attracted most attention: understanding between team members, the notion of teamwork in relation to task dependencies, and interdependencies of team members, and the purpose of HART. Based on these discussions, a number of questions and issues were raised that will help focus future research on topics related to HART. The plan is to refine our understanding of these questions and issues in the coming months and to try and publish that overview, and perhaps also additional articles summarizing aspects of the state-of-the-art, in a high profile journal that is relevant for HART.

 

The remainder of this report contains a short summary of some of the notions put forward. From Answer.com we used the running definition: “teamwork is the actions of individuals, brought together for a common purpose or goals, which subordinates the needs of the individual to the needs of the group.  … The interactions among the members and the work they complete together is the teamwork.”

The debate over the week repeatedly came to the questions: what makes HART different from pure human teamwork? What is teamwork, and what is essential for teamwork? How can we model that? How do we program agents and robots capable of flexible teamwork? What do humans expect from HART? Is there a unified theory of teamwork? Is all teamwork the same?

 

These broad questions where further refined through small and large group discussion. For example, what is the role of the agent in HART? Should it be an intermediary between human and robot? The different teamwork phases should be modeled and implemented in order to harmonize goals and team members: initiation (recognition of the need for others to get the work done), team formation and role adoption, doing the work and team maintenance, and abandoning team activity (recognition of end-resolution). What is the effect of the timing of teamwork (synchronous, asynchronous, ad-hoc)? How do we model and implement pre-organized vs. self-organized teamwork? What is the difference between coordination, cooperation, collaboration and teamwork? To what extent does the team need common ground / shared understanding / shared mental models? What are the roles of context and environment? How should we model and how should team members deal with the difference between individual goals (for the individual or for the team) and team goals (for individual team members or for the team as a whole)?

 

Catholijn Jonker, Jeff Bradshaw, Virginia Dignum