In the literature on human computer interaction, design is a method for making a artifact more usable [75] (p. xiii-xxv): The design process creates artifacts that are well-suited to their environments of use. Design activities require the management of trade-offs. ``The ongoing process of design is iterative at two levels: iteration by the designer as a piece of current work develops, and iteration by the community as successive generations reveal new possibilities for the medium." Design entails an ongoing dialog between designers and users. Design is a social activity. Users avail themselves of the design as a support for using the artifact; it informs the users activity [58]. Participatory design occurs when the people destined to use the artifact play a critical role in designing it (see [66,29]).
All of these features of the design process for artifacts apply to the design process that co-participants realize as they produce, with practice, a design for regularly occurring joint activities. The design of a conventional behavior is tied to the emergence of an improving set of expectations about the best points of coordination in a given situation. The creation of a convention is a participatory design process. Each time co-participants iterate through a recurring joint activity, improvements in the design can be created. Over time, for regularly occurring problems of coordination, a convention of behavior (the expected points of coordination) becomes a part of the joint activity. Iteration by successive generations of the community reveal new possibilities for joint activities. The emergent design for the activity frames the expectations of the co-participants, is part of the initial common ground, and is coded in the recall of prior joint activities.
Suppose an actor is helping another actor in a wheelchair pass through a doorway. A husband Dick recently broke his leg in a skiing accident and he is in a wheelchair; his wife Jane is helping him to get about. From Jane's perspective, a design that coordinates her pushing Dick through a doorway will emerge as the the activity proceeds. How to position the wheelchair, when to open the door, and what are reasonable expectations about what participants can and will do, are features of the situation that must be discovered. At the outset, the first time, one does not know a good design for the activity; however, the physical constraints of the situation and constraints that emerge from the actions of each of the participants will shape the joint activity [58]; the participants jointly attune their behavior to the constraints and affordances of the situation [30] (p. 8-9). If another couple were in a similar situation, one would expect them to work out a similar design because many of the constraints and affordances that are reasoned about are present in the design and construction of the doors, doorways, and wheelchairs that are used. Through continued practice, Dick and Jane develop a good design for the activity (a convention of behavior). Certain points of coordination become regular features of how they negotiate this kind of joint activity; there are points of coordination in the design that Dick and Jane can reasonably expect will be achieved during the activity.
Whatever design for the activity emerges, it is historically conditioned. Each of the artifacts (i.e., doorways, doors, wheelchairs) has a history of designed activities in which it is used. Each time new actors learn to use a wheelchair, they are part of a continuing story about wheelchair design, construction, and usage. The performance of an activity is tied to the performance of this activity by prior generations of individuals within the community, (re)emerging from the movement forward of prior joint activities [14].
For a given activity, there may be multiple designs (conventions) available to participants against which they can coordinate their behavior. Consider the general case of two people coordinating their efforts to pass through a doorway. There is no unique pre-determined design for coordinating behavior. Some conventions that apply to that situation are: men open doors for women, adults for the elderly, anybody for a person in a wheelchair or a parent pushing a baby carriage, adults for children, and teenage boys never open the door for each other. Participants in an activity must agree on their assessment of the situation in order for things to proceed smoothly. Suppose Susan and Steve approach the door and Steve is older than Susan. Whether it is a case of men opening doors for women or young adults for the elderly is a matter of degree. For a joint action to run smoothly, participants must be signaling to one another how to interpret the situation; either Susan or Steve may begin to make a movement toward the door handle and begin to position their body so as to open the door for the other person.
Activities that have pre-existing designs are not completely predetermined. There are portions of the activity that are not designed and, to operate effectively, one needs to reason about those too. Removing an aspirin from a bottle is an example of emergent structure discussed by [1] (p. 164). Clearly, parts of the activity are not designed. Agre's point is that there is no way to anticipate how the aspirins are arranged in the bottle before you open it. On the other hand, other parts of the activity are designed. Opening the bottle is designed; it involves pushing the cap and rotating it counter clockwise. Instructions for the design exist on the face of the cap. Even if a design for a portion of the activity does not pre-exist, the functioning of the individual's memory in support of activity may effectively impose one. If the bottle is full, an aspirin may be accessed using two fingers. If the bottle is almost empty, slowly pour one out; if more than one lands in the palm of you hand, slowly pour all but one back. In fact, these behaviors reflect the design of aspirin bottles, which in turn are subject to the constraints that emerged from prior aspirin-bottle use.
A novice cannot obtain the design of an activity purely by analysis; she must reason pragmatically [7]. The design can only be obtained by action, through continued practice, relying on communication between participants (even if one of the participants is in absentia). If I visit a stereo components store, each of the CD players exhibited in the showroom are likely to have a different design. In one case, switching to the next song in sequence involves pushing a button with >> signs on it, but for a different device the same effect is achieved by rotating a knob. For one CD player the power button is a push button and is located in the upper left hand corner of the face of the device, for another CD player it is a switch and located elsewhere. As a member of a community of actors who use such devices, I am already familiar with a large number of conventions for the interfaces for devices of this sort. Nevertheless, for a given device there is no a priori analysis that will allow me to determine exactly which conventions and in what sequence they are to be followed. There are no rules of inference (be it induction, deduction, or abduction) that will allow one to infer the exact state of affairs at the scene of the activity in using a device independent of the activity of using the device.