We characterize social applications as those involving interaction
among multiple autonomous agents. We are interested in the essential concepts
and approaches for modeling such applications. We make the case that i* has
some limitations with respect to the modeling of social applications. The problem is in the intentional nature of i*. The deeper roots though lie in the centralized
machine-oriented approach of current requirements engineering approaches. We
recommend an interaction-oriented approach to requirements modeling, modeling in terms of social commitments rather than dependencies, and in general, accommodating a distributed perspective right from the earliest phases of software
engineering. For clarity, we also distinguish social commitments from various
similar-sounding notions in the literature.