Collective Commitment

4.1.1 Gradations of Teamwork

The commonsense meaning of teamwork covers different gradations of being a team. Take, as a first example, teamwork in a group of researchers who jointly plan their research and divide roles, who reciprocally keep a check on how the others are doing and help their colleagues when needed in furtherance of their goal to prove a theorem. All aspects of teamwork are openly discussed in the team, and members keep one another informed about changes in the plan. Therefore, this is a paradigmatic example of teamwork.

Contrast this kind of non-hierarchical teamwork with a second example, of a group of spies who all work on the same goal, say to locate agent X. In their case a plan is designed by one mastermind, who divides the roles and divulges to each participant only the necessary information. Thus, members may not even know the main goal, nor who else is included in the group. Even though the connection between group members is rather loose, we would still like to speak about Cooperative Problem Solving (henceforth CPS), albeit a non-typical case, and not about proper teamwork.

In the examples, individual and group awareness about such ingredients of CPS like the main goal and the plan to achieve it, range from very high (as in the first example above) to very low (as in the second example). Therefore, we claim that these two cases cannot be reasonably covered by one generic logical model of teamwork. Thus far in the MAS literature, authors restricted themselves to a typical idealized understanding of teamwork, usually abstracting from organizational structures and communication possibilities

Teamwork in Multi-Agent Systems: A Formal Approach Barbara Dunin-K ¸eplicz and Rineke Verbrugge  2010 John Wiley & Sons, Ltd

56 Teamwork in Multi-Agent Systems

(Dunin-K ¸eplicz and Verbrugge, 1996; Grosz and Kraus, 1999; Rao et al., 1992; Wooldridge and Jennings, 1999). In contrast, in the sequel we will provide a full characterization of group attitudes, covering the range from proper teams to more loosely connected groups involved in CPS. We will also highlight the importance of choosing an appropriate gradation of teamwork needed for a specific goal in given circumstances. Therefore, a mechanism will be provided to create an adequate type of commitment. The proposed model of collective commitments is minimal, in order to support the system developer’s quest for efficiency at design time.

4.1.2 Collective Commitment Triggers Team Action

Suppose we have a team with a collective intention to achieve a goalϕ. Does this suffice

for the team to start its cooperative action towards the goal? Clearly not: a bridge from a still rather abstract collective intention to precise team action is needed.

What is obviously missing is a detailed plan including individual actions to realize the goal. This would enable the agents to make bilateral ‘promises’, called social commit- ments, to perform their parts. Why bilateral? Because by their simplicity, they are easy to implement and revise. The collective motivational attitude that covers the outcome of this planning and committing is a team’s collective commitment or a weaker attitude that plays a similar cohesive role. Ultimately, collective commitment in a group of agents is aimed to trigger team action, that is, a coordinated execution of agents’ individual actions according to the adopted social plan. A formal model of a group’s motivational stance towards teamwork is the focus of this chapter.

Again, agents’ awareness about the overall situation is vital. As a reminder, the notion of awareness applied in modeling agency may be viewed as a reduction of a general sense of ‘consciousness’ to an agent’s beliefs about itself, about other agents and finally about the state of an environment (as discussed in Section 2.7), naturally expressed by different degrees of beliefs. These range from the rather strong common beliefs through weaker forms, like possibly iterated general belief, to even weaker individual beliefs, depending on the circumstances.

4.1.3 A Tuning Mechanism

When asking what it means for a group of agents to be collectively committed to achieve a common goal, both the circumstances in which the group is acting and the structure of the organization it is a part of, have to be taken into account. This implies the impor- tance of differentiating the scope and strength of the group commitment. The resulting characteristics may differ significantly and even become logically incomparable.

The idea of dials to tune the nature of the commitment to the particular purpose seems to be both technically interesting and intuitively appealing. We intend to provide a sort of

tuning mechanism which enables the system developer to calibrate a type of collective

commitment fitting the circumstances, analogously to adjusting dials on an audio system. The appropriate dials, characterized in the sequel, belong to a device representing a general schema of collective commitment.

In order to illustrate the expressive power of such a tuning machine, several types of commitments corresponding to various teamwork schemes occurring in practice will

Tuning Machine for Collective Commitments 57

be discussed. Apparently, the entire spectrum of possibilities is much wider, due to the number of possibly independent choices to be made. The resulting types of collective or group commitments, described in multimodal logic, may then be naturally implemented in a specific multi-agent system. In this way the tuning mechanism may be viewed as a bridge between theory and practice.

In this chapter we concentrate on a static theory of teamwork defining complex social and collective motivational attitudes in terms of simpler individual ones. The next three chapters, in contrast, focus on the dynamics of individual intention and social commitment in the context of cognitive and social processes involved (see also Castelfranchi (1999), Dignum and Conte (1997) and Dignum et al. (2001b)).

The rest of the chapter is structured in the following way. In Sections 4.2 and 4.3, a short presentation is given of the logical framework extending that of the previous two chapters in order to construct the building blocks of collective commitments. The central Sections 4.4 and 4.5 explore different dimensions along which collective commit- ments may be tuned to fit both the organization and the environment. A general scheme is presented in a multi-modal language, as well as five different notions of collective commitment fitting to concrete organizational structures. Section 4.6 explores how sev- eral interesting organizational topologies, such as stars, rings and trees, can be explicitly represented in definitions of collective commitment. Finally, Section 4.7 focuses on dis- cussion and provides a bridge to the subsequent chapters about the dynamic part of the story of teamwork. The reader may skip Sections 4.2 and 4.3 at first reading, and instead start reading from Section 4.4, only jumping back when needing more background about the building blocks of collective commitment.