While both goals are desirable, they conflict with each
other. The emphasis on predefined team structures and
processes often reduces the flexibility of the team, while
maximizing the flexibility of the team usually requires
making shared, redundant, and hence ambiguous role
assignments. To balance these two conflicting goals, we
need a practical and flexible computational framework for
representing and reasoning about the overlapping shared
mental models amo ng teammates. Such a framework would
enable an agent to dynamically reason about the status of
the entire team and adapt its behavior accordingly.