The NIFTi AAAI Fall Symposium on Human-Robot Teaming
NIFTi organized a AAAI fall symposium in November this year, on the theme of human-robot teaming. The symposium included invited talks by Ron Arkin, Jeffrey Bradshaw, and Satoshi Tadokoro. One of the main lessons: Think of making robot intelligence as acceptable intelligence, to make a robot a real team player.
Robot-Human Team-Work in Dynamic Adverse Environments
Report of the AAAI 2011 Fall Symposium
Geert-Jan M. Kruijff, Panagiotis Papadakis, and Fiora Pirri
Robots are gradually making their way into different aspects of our lives. We find them at home. On the factory floor. And, more and more, they are also performing missions in complex outdoor environments.
At the symposium we discussed issues in human-robot teamwork, set in environments which are dynamic and adverse. Typical examples here include urban search and rescue (USAR), or security missions. This is a timely topic: This is already happening. Humans are taking robots on such missions. Using them in situations which are too dangerous for people to (immediately) enter.
And where this is not happening yet, it is very likely to be happening soon.
Thing is, human-robot teaming in such environments is much more than "just" a technical issue. This is "people + robots." This is a complex socio-technical system in which humans and robots are trying to work together under very difficult circumstances. They are performing under heavy physical and mental stress. And we know that in such extreme situations, under such extreme conditions, human characteristics, behaviors, emotions are driven to extremes. People change. In how they behave, act, what they pay attention to, how they interact.
If we use robots in such situations, will we actually make things better? Or will we make things worse? Can we make robots to assist humans to do better, to handle such situations better?
Looking at these problems from a the viewpoint of the "user" (first responder, policeman, soldier, etc.) the symposium raised from anywhere between practical and fundamental questions. What is a team? What is a human-robot team? What is teamwork in such a context - when we consider mixed-initiative settings, do we actually want robots to take "the" lead? What makes for good teamwork, and what would be bad teamwork? What does it mean to "share" situation awareness, when humans and robots are typically geographically distributed, and each and everyone has his or her own ways of looking at a situation, experiencing it subjectively?
And, amidst all the "how's and could's" - should we?
The take-home message from the various presentations, and invited talks by Ron Arkin, Jeffrey Bradshaw, and Satoshi Tadokoro, were essentially as follows. It's about working together: There as an inherent interdependency between humans and robots. And this requires much more than "autonomy" (even when we consider autonomy to be multi-dimensional). It's about the social dynamics of actors, roles, performing particular tasks. It's about situation awareness specific to interactions between roles, in specific task contexts.
And above all, it is about making robots acceptable. As Satoshi Tadokoro aptly put it, it's perhaps better to turn AI from "artificial intelligence" into "acceptable intelligence:" robot intelligence which is clear, predictable, and acceptable in a given team context. Only then we may be able to succeed turning robots into real team-players.
Geert-Jan M. Kruijff, Panos Papadakis and Fiora Pirri served as cochairs of this symposium. The symposium was partly supported by the European project "NIFTI: Natural Human-Robot Collaboration in Dynamic Environments" [http://www.nifti.eu]. The papers of the symposium were published as AAAI Press Technical Report FS-11-07.
See also the symposium website for more: [ WWW ]