Next: 2.4 The Design of an Activity Up: 2. Joint Activity and Convention Previous: 2.2 Convention

2.3 Internalized, Emergent, and/or Situative?

Scripts [65] is an example of a cognitive theory that is meant to account for convention-based behavior. A script represents a sequence of events coupled to a particular context with actors having roles and coordinating their behavior in predictable ways in order to achieve overlapping goals. Attending a movie, a restaurant, a wedding, or a Bar Mitzvah are all occasions of script-like activity -- as are riding a train, flying in a plane, or boarding an oceangoing vessel.

The notion of a script was developed before models of goal-directed behavior were transformed from plan-based to activity-based [70,3]. In the mid-80's, models of goal-directed behavior began to account for the uncertainty of the world. Plan-based models assumed a benign world, where things went according to plan: one makes a plan and then executes it. For various reasons, researchers began to consider alternate models of behavior: the intractability of planning [11], the inadequacies of models of acting that were dependent on internal representations [9], the binding problem [2], and the general uncertainty of the world. Under these newer schemes, plans were not a complete specification of action. Rather, plans were vague, only providing an orientation [70]; a plan was a recipe which specifies what an actor is to do but not how [60]. Much of the structure in the activity emerged from the individual's interaction with the environment during activity and did not exist prior to the activity [22].

These kinds of critiques also apply to models like scripts. A script is a rigid representation of a particular sequence of events; for everyday behavior, the uncertainty of the world forces a less rigid representational form for knowledge of conventions. Pedestrians crossing in opposite directions at a busy intersection in downtown San Francisco is an example of individuals coordinating their behavior in a regular manner where their joint behavior is not characterizable by a script. Another example is the joint activity of a team of movers. In moving furniture and boxes from the house into a truck, there exist regular patterns of coordination among the crew, but none of these can be reified into a script. In both cases, there is too much uncertainty for events to unfold exactly according to some pre-determined script. Even activities that seem more script-like in nature, such as buying lunch at the drive-through at McDonald's, are filled with numerous events that are part of the `drive-through scenario' but deviate from the specifications of the script. While placing your order, your daughter Emma interrupts to change her order from a Sprite to a Coca Cola, and maybe, on second thought she'd prefer a plain cheeseburger rather than the Chicken McNugget s. Deviations in the main current of the activity are constant and commonplace. A pre-determined script that each of the actors follows only accounts for some of the richness of the phenomena.

For a joint activity of buying food at the drive-through at McDonalds, there is more than one expected point of coordination. Collectively, these points of coordination form a design for the activity. Internalization of some portions of the design is a pre-requisite for generating conventionalized behaviors, but the design is not a complete specification of what occurs when the conventionalized behaviors unfold. Whatever expectations are realized require joint work by the participants. Much of the work they do to keep themselves in-synch and on-track involves helping one another by confirming and realizing each other's expectations through verbal and non-verbal communication. For a joint activity, familiarity with the designs of conventionalized behaviors is part of the initial common ground, but the conventions themselves only emerge as part of the activity. This account is both interactive [70,48,1] and situative [39,41,30].

A key feature of an interactive account is that structure emerges from the give-and-take of activity that does not exist prior to the activity. For Suchman (1987: p. 52), a plan for canoeing down a series of whitewater rapids may orient one, but much of the apparent structure of the activity emerges from the activity itself. The work of Sachs, Schegloff, and Jefferson [64] makes the case that turn-taking in conversation is a locally and interactionally managed form of group decision-making that is administered by the participants of the conversation; it emerges from the conversation and is not pre-determinable. Lave's work [48] on the `dialectic' of a shopper doing arithmetic in the grocery store shows how context is inseparably part of the activity. For Lave, Murtaugh and de la Rocha [49], one's personal sense of an activity of shopping at the supermarket is not completely internalized; the potential of a given `setting" is only realized in the dialectic of activity. In Artificial Intelligence, the work of Agre [1] and Agre & Chapman [2,3] demonstrates how `plan-like' behavior can emerge without planning from an interaction between the machinery of the individual's reasoning processes and the dynamics of the world. In their PENGI model, the actor continuously redecides what to do; its representations are deictic, thus requiring an interaction to be realized. In each of these cases, cognition and some of the structure of behavior is emerging from activity and interaction.

Conventionalized behaviors are partially emergent. The participants in a joint activity have a predisposition to act in ways that will simplify coordination. They have knowledge of points of coordination (designs) for certain kinds of activities, but the potential of these designs are realized and emerge only during the activity. The convention of behavior is not uniquely determinable independent of the occasion and procession of a given activity. One time it emerges one way, a second time a different way; certain points of coordination tend to be realized on each occasion, but in different manners. Some of the structure of the activity, of which one would attribute the term `convention', emerges from the activity itself.

The situative view emphasizes the study of cognition in terms of units of analysis larger than an individual [40] (p. 62):

If the individual mind itself is the only locus considered for the structures that organize thinking, then everything that is required to create a sufficient account of cognitive activity has to be crammed into the individual mind. This leads the followers of this view to try to put more in the individual mind than belongs there. The properties of groups of minds in interaction with each other, or the properties of the interaction between individual minds and artifacts in the world, are frequently at the heart of intelligent human performance.

One of Hutchins' example is the airplane pilot and the cockpit. He argues [39] (p. 286) ``The cockpit system remembers its speed, and the memory process emerges from the activity of pilots. The memory of the cockpit, however, is not made primarily of pilot memory." Some significant amount of what needs to be remembered by the airplane pilot is located in representations (e.g., the `speed bugs' on the airspeed indicator) available in the task environment; thus tasks requiring complex and costly internal memory operations can be distributed. A second example is the navigation of a naval vessel that is underway [41,40]. There are two parts to navigating a ship: position fixing (which essentially answers the question where are we) and dead reckoning (where are we heading). Position fixing is distributed among the ship personnel and their tools. Bearing takers, timer-recorders, plotters, keeper of the deck log, and the fathometer operator all participate in position fixing. Several actions are coordinated to occur simultaneously at the call of ``standby to mark." Much of the computation in fixing the point is done by the nautical chart (the mercator projection chart), which is a specially constructed artifact that supports this computation. The cognition that is occurring is distributed and is part of a larger socio-technical system.

Everyday actors also rely on the larger context to manage the achievement of a conventional behavior. There are no guarantees that two actors' prior knowledge of a particular joint activity exactly match. They may have some overlapping expectations about points of coordination, but other parts of the design are contingent on the specifics of the larger context. Novices at a particular kind of conventional activity can rely on the larger context to fill in gaps in knowledge. The first time an individual uses a device, she may realize a conventionalized behavior, even though she may not be familiar with the type of device she uses. This is because instructions are generally available in the larger context. When the individual dials a number on the air telephone for the first time, information is provided in the design of the setting to aid the user in performing the needed conventions of behavior. The label on one of the buttons indicates `dial tone', the button affords pushing, other buttons are labeled with numbers. The instructions inform the user she must first push the dial tone button and wait for a dial tone before dialing the number. Other situations, with different arrangements in the design of the interface, require the user draw on her knowledge of other conventions for operating devices. In each case, the setting of device usage provides signs for the user to interpret in order to determine the situation. Emerging from the development of the activity in the larger context is a conventionalized behavior that did not pre-exist in the mind of the individual actor, but fits the design for the activity that has been developed within a community of actors.


Next: 2.4 The Design of an Activity Up: 2. Joint Activity and Convention Previous: 2.2 Convention
Last Update: March 10, 1999 by Andy Garland