In the everyday world, one must constantly be internalizing, reasoning about, and muddling through on the basis of convention. Convention helps to coordinate behavior between actors. There are few (if any) actions that are not steeped in convention. Convention is central to the formulation of a model of everyday reasoning and behavior.
One example of a coordinated behavior that has its basis in convention is crossing an intersection in the road either by car or by foot. Participants in this situation are informed by a large set of conventions concerning diverse things such as crosswalk markings, conventions for multiple cars approaching a stop sign, ordering preferences between turning cars, differing configurations of merging traffic (T-intersections and Y-merging thoroughfares) and types of roads (highways and residential streets, or one-way and two-way roads), ages and capabilities of pedestrians and drivers, et cetera. Deciding which of these policies are in force requires cooperation between individuals, and conventions also exist for making that decision. Any sort of stereotypical situation for which a script could be generated to represent the coordination of behavior between participants is also a case where convention is at work, e.g., subway stations, restaurants, buses, and the supermarket checkout. The usage of devices, tools, and artifacts is primarily reasoning about convention. Communication, in any of its forms, is tied to conventions of use.
The early parts of the paper frame the discussion of what is convention. Matters concerning joint cognition and behavior are discussed in the terms of Clark's [13] model of joint activity. Lewis' [52] definition of convention is presented. Scripts [65] is discussed as an example of a cognitive theory about conventional behavior. Research on situated activity [70] and distributed cognition [41] is used to critique the script model of reasoning about conventions of behavior. The first pass story on convention runs as follows: During the give-and-take of activity, participants reason with a bias that at one point or another their expectations about points of coordination will be met. There is also the understanding that because of uncertainty, interruptions, and numerous other opportunities to get off-track and out-of-synch, the participants must work continuously and jointly to achieve conventional coordination. The expected points of coordination collectively form a design for the activity -- the convention -- but not a complete specification of it. Conventional behaviors develop from practice within a community of actors and are partially emergent. Different actors bring to bear different knowledge and experience. Gaps in behavior are filled by the historically conditioned constraints, affordances, and patterns of coordination available to the participants as they proceed with their joint activities.
The second part of the paper explores the emergence of convention in circumstances where there is not a ruling body of knowledge developed by prior generations of actors within the community to draw on as potential designs for cooperative and coordinated behaviors. Our example domain (MOVERS-WORLD) is a group of actors who are part of a moving company. Their job is to move boxes and furniture from a house into a truck. With practice, individuals within the community begin to converge on a set of conventionalized behaviors that best match the regularly occurring problems of coordination in the domain of activity. One feature of the model is that the community improves its performance despite the fact that individual actors reason independently about their experiences. Because the actors use prior experiences to guide future behaviors, they can, however, indirectly adjust their expectations about the best way to coordinate behavior by making adjustments for the success and failure of this or that act. Another important feature of our model is that the mechanisms for improving behavior are tied to the memory function of individual actors. The middle of the paper works through some of the details of our architectural assumptions. Section 4 details how memory for coordinated behavior functions for the individual actors.
The third part of the paper presents a set of experiments that explore some features of convention within the context of MOVERS-WORLD. The first set of experiments examine the performance changes that occur as conventions of behavior develop. The second set of experiments analyze the possible hypothesis that an ideal script for a conventional behavior is being determined in the individual minds of participating actors. These latter experiments also pinpoint the function of memory of prior coordinated activities, especially in establishing expected points of coordination in the design for the activity.
The discussion section begins by connecting everyday reasoning about convention to a prior model of cognition as pragmatic action [7,5]. The `where' and `when' of convention are next explored; a discussion of shared plans [34,32] is featured prominently in this argument. The last discussion topic fits the research on convention under the umbrella of a research program that combines work in Cognitive Science with the research and methods of work on situated activity that have come out of the social sciences [39,30].