Airports have traditionally used a largely manual system for landing planes (Mackay et al., 1998). The system is still respected and used in many places, and in this sense is resilient. The system is routine and has an air traffic controller seated in front of a radar screen at an angled table of flight strips. The flight strips can be placed on the table in various configurations in relation to each other. Each airport has several air traffic controllers controlling different parts of the air space around the airport.

The activity of landing a flight begins with a printed-paper flight strip containing basic flight plan information. This strip is generated by computer or can be handwritten in the absence of a working computer system. Figure 14.2, “A flight strip describing Air France Flight 540 (Mackay, 1998, p. 322).” shows a typical flight strip.

As an aircraft approaches an airport, a flight controller takes over control of its landing. When a new strip is generated, the controller’s first task is to remove it from the printer and insert it into a strip holder. Strips are continually picked up and put down, reordered, grouped, offset, moved into columns, arranged and rearranged on the controller’s table to denote different traffic conditions. The placement of the strips provides the controllers with information regarding action additional to that written on the strips. As the landing progresses, the flight passes from one controller to another by physical handover of the flight strip that, by its nature, is palpable for both controllers. Often controllers are side by side thus facilitating handover to another sector by structuring the area to help the activity.

Once a controller takes control of a flight strip, he gradually adds information to the typed strip (as seen in Figure 14.3, “The strips being manipulated by an operator.”). The markings allow controllers to look at a group of flight strips and quickly select the ones coming under their control and other information about how the activity is progressing. The layout of strips also gives a controller an immediate appreciation of the control environment (involving many flights) thus helping the controller to select the next action.

Each strip represents the activity of landing a specific flight and contains information important to landing the plane, not directly relating to the plane itself. In this way, the information on a strip is not tied to any specific object. Neither does each strip contain all of the information required to land the flights that are under control. It is the structuring of the controller’s environment using the strips that shows information beyond that which is on the strips themselves. For example, the way the strips are stacked in Figure 14.3, “The strips being manipulated by an operator.” means something specific for the controllers and helps them to remember and reason about the flights they are landing. Handing over control of flights from one controller to another is achieved by passing strips and is facilitated by how the room is laid out, and such handovers seldom include verbal exchanges. The limited space where strips can be placed alerts the controller to busy situations because room for new strips is then hard to find. In these situations controllers will hold new strips in their hands. Controllers sometimes write their own strips when unusual things occur, relying on the convenience and flexibility of paper.

Analysing this system using the deliberative approach is not straightforward. No representation is tied to a specific object. Each strip is about landing a flight rather than the flight itself or the aeroplane. Together, the strips are about the activities of landing that are under the control of the controller. None of them describes a flight enough to say that they represent an object in the sense of the deliberative approach to modelling.

In contrast, this system is easily related to the situational theory of agency. The strips represent the activity of landing a plane and, together with other strips and their relative position, these are sufficient to enable a controller to appreciate the current situation and to select actions. This, in our view, is a more plausible explanation than one based on the deliberative theory of agency.

A small factory (Cash Engineering Research) manufactures air compressors and has built up a system for doing so over several years. The workers in the factory have played an active role in designing the system. Known as the Cash Compressor System, the system is for production control in a small factory of four staff manufacturing about 200 air compressors a year. The system has a whiteboard that represents non-routine aspects of the compressors being made. There are no computers in the factory. What is interesting is how little information is represented on the whiteboard without compromising control or efficiency.

The factory is designed so that the person taking orders on the telephone in the middle of the factory has full view of all available stock hanging on shelves lining the walls. The main components of the system include a whiteboard of open customer orders and the physical parts of the air compressors that, by their construction, implicitly contain information about their own method of manufacture. The information on the whiteboard is job-specific including name of client, and options such as colour, and compressor motor size. The system has been designed deliberately in this way to reduce the need to represent things.

Manufacturing commences when the order is received by phone and a line order is added to the whiteboard. The parts for making the customer’s compressor are checked for availability visually, and if need be, ordered on a one-off basis. The machine assembler then takes a machine base and begins construction, referring to the whiteboard only for order-specific information that is not part of the standard assembly routine.

What is interesting in the Cash System is what is not represented. There is no information about how to construct the machine: the machine acts as its own ‘jig’ through devices for guiding a tool or part to a specific place. Employees have learned the limited number of techniques used with the ‘jig’. There is no parts-list or inventory system: the availability and quantity of parts holdings are clearly seen on the shelf. The only recorded requirements-related information is in the reference to non-standard choices on the whiteboard.

If this system were to be explained by the deliberative theory of agency, representation would include detailed information about each compressor being manufactured. This is not the case in the Cash System where very minimal information is kept explicitly on the whiteboard. No rules can be found to enable a worker to take the individual parts and assemble a compressor. Instead we see the next action being selected by the partly manufactured machine being presented to the worker. Only a limited range of choice is available to them. The worker knows what happens next because there is very little (often no) choice confronting them. When there is a choice, the whiteboard tells them the option to be selected based on the customer’s desires.

The Cash System is a highly situational one where representation is almost absent. Action is selected by routinely acting on the partly manufactured machine based purely on the current status of the machine. In the Cash System, ‘the world (is) its own model’ (Brooks, 1991) in that the machine ‘jigs’ itself and parts are visible, obviating the need for stock data. Consequently, the current situation is found in the visible state of the stock on the shelf, the number of jobs on the floor, the condition of the partly manufactured compressors that are the jobs on the floor, and the markings on the whiteboard. Due to the careful design of the factory layout, all these are immediately visible to a worker. In addition, the use of a single small whiteboard allows the foreman to grasp the total production situation at a glance. The recorded information on the whiteboard is largely ephemeral (except for a small amount of recorded information for warranty purposes that is kept in a book). When a job is finished it is removed from the whiteboard and the new situation is revealed.

The Japanese Kanban system (Schonberger, 1987; Womack et al., 1990) is widely used in the automotive industry for the activity of replenishing parts for production. Kanban is the Japanese word for ‘card’ and the movement of cards in this system controls stock levels and replenishment activities. For each part there is a fixed size container. A Kanban has printed on it minimal information about the item it is used for, usually product ID, the primary supplier and the workstation where the part is used (see Figure 14.4, “A typical Kanban card.”). There are a fixed number of Kanbans in existence for each item and, except when desired manufactured capacity changes, they are neither created nor destroyed.

Imagine a container half-full of parts on a factory floor. The container has a Kanban attached to it. Goods are taken from the container, which is stored at the production workstation, until it is empty. The Kanban in the empty container is then placed on a Kanban board near the goods receiving area where it becomes a signal that the item needs replenishment. The board has hooks in supplier order. When placed on the board the Kanban becomes ‘free’. The board has the Kanban system operation rules (Kanban rules) clearly displayed. When a supplier’s truck arrives with shipments of items to deliver, the driver checks the Kanban board and takes the Kanbans on the relevant hook back to the supplier’s site to authorise replenishment of these items next time around. When the items are subsequently supplied, the Kanbans are returned to the work stations, in the full containers, where they are used.

The deliberative theory of agency cannot relate at all to the Kanban system. The cards do not consistently correspond to anything specific in reality. When they reside with the parts they could be thought of as being a representation of these parts, although they have no system purpose in this state and they will later actually refer to a different group of parts. When they are free, they represent a stock shortage. When they are on the Kanban board they are an authority to re-supply the parts. There are no records of stock levels that we would expect to see in a deliberative approach.

Examining the system using the situational theory, a Kanban card represents part of the activity of maintaining stock of a specific item. All of the Kanbans, together with the rules by which they are used, provide simple ways of reasoning about stock. If many of the same type of Kanban appear at the board then an undersupply may be occurring or there may be trouble with the supplier’s transport. An absence over a prolonged period indicates a delayed manufacturing process. In addition to simple, reactive rules for Kanban movement, the affordances of the physical nature of Kanban cards (they can neither be created nor destroyed and they cannot be in more than one place at a time) indirectly enforce all important replenishment business rules, in particular that there can only be a fixed number of parts in the system.

The Kanban is rebound over time from one full container to a different full container some time later. An interesting feature of this system that is the Kanban’s meaning changes according to where it is. When it is travelling back to the supplier it functions as a request for an order from the manufacturer. When it is on the board it shows a shortage of a specific item.

Each of the systems outlined above has features in common and that mirror the features found described in situational systems literature: activities, situations, aspects of situations, environmental structure, and environmental affordances. We now examine each of these characteristics, highlighting the approaches each system uses. We emphasise interesting features that add practical depth to our understanding of the theoretical constructs.

All systems use tokens to represent activities. Physical strips in landing aircraft represent flights being landed. Rows on the whiteboard represent activities of making a compressor at Cash. Cards in the Kanban system represent the activity of replenishing goods. None of the tokens represent objects and properties in the way advocated in existing data modelling methodologies. An interesting feature of these manual systems is the use of positions of tokens to help actors in reasoning about activities. In the landing system, the relative position of the strips helps the controller to reason about all landings. Kanban cards on the board help operators to reason about goods shortages and priorities. The importance of manipulation of concrete things in the environment for practical reasoning has been emphasised by writers on situated cognition (Lave, 1988; Clancey, 1997).

Tokens representing activities often have information about aspects of situations on them that show the actor the situation they are in. This is best illustrated in the landing system by the markings on strips. However, even in the landing system the position of strips relative to each other also shows aspects of situations. Similarly, in the Kanban system the presence or absence of tokens in various places reveals situations to human actors in the system. The Cash System partly shows aspects of situations on the whiteboard and partly in how much of the compressor has been completed. This is because the stage of manufacture of the compressor shows part of the situation to the worker. This parsimonious use of representation is quite consistent with the situated view, in which a small number of aspects are sufficient to trigger a situated action, but inconsistent with the deliberative approach.

Structuring the environment of systems is critical to situational activity because without it repeated actions would not reliably result in goal attainment. This is achieved in two ways. First, it constrains the possible new situations an actor experiences as a result of action and this reduces the cognitive burden of choosing alternatives. The structure of the maze is an example of this. Second, the environmental structure may help reasoning about activities. Often the palpability of tokens and their physical properties help deliver both benefits. In the landing system, a controller can position the strips relative to each other because of the slope and size of the table, thus enabling situation detection and reasoning about activities. In the Cash System, the partly manufactured machine, by being its own jig, only permits a restricted range of actions. The Kanban system limits the quantity of stock circulating by having only a limited number of cards. Workers can reason about delays in the activity of stock replenishment or in manufacturing by noticing prolonged absence or presence of Kanban cards in particular places. In addition, each of these systems requires considerable structuring of the broader environment of work to make these simple reactive systems work. For instance, the Cash factory is laid out so that the availability of all relevant part options is directly visible to the foreman when adding new records to the whiteboard. The need for work environment structuring, for instance the use of teams and production cells, for the successful implementation of Kanban is also emphasised in the Just-In-Time literature (Schonberger, 1987).

Tokens, or other parts of environments, also help actors to hand over situations to others. In the landing system, a controller can hand a strip to another controller because the controllers are often next to each other. In the Cash compressor system, a half-finished machine by its very state facilitates a worker in taking over the activity and situation from another worker.

In the flight landing system not only does the passing of a single strip pass the situation of a particular flight from one operator to another, but the visible arrangement of all strips is used to hand over the total flight situation at the change of shift (Mackay et al., 1998). Similarly, Kanban movements hand over a shortage situation between the multiple participating actors, while the arrangement on the Kanban board allows the total shortage situation to be seen by foremen.

By examining the systems described above we can draw two groups of implications for a situational methodology. First, the examples exhibit many characteristics consistent with the situational theory of agency. Second, there are characteristics that emerge from these systems that add to our understanding of the practical application of the situational theory. We discuss each of these below.

Each of the systems confirms activities, situations, actions, environmental structure and environmental affordances as important ontological categories for situational systems. Fundamentally, activities, not objects, are represented in these systems. Situations and aspects of situations are shown to actors so they can select actions or reason about their activities. Environments are structured and use affordances that increase the reliability of the goal of an activity being realised. The relationship between environmental structure and affordance is sometimes complex. This is seen in the Cash System where the machine is its own jig. The jig is designed so that the environment of the worker changes in such a way that precisely one situation is returned.

A characteristic emerging from the study is the extensive use of physical tokens (e.g. strips) to contain information about situations and activities and to facilitate reasoning. Tokens often represent different situations according to their physical relationship to other tokens, and their physicality aids reasoning about situations for human actors. The need for manipulation in situated reasoning is an important feature of these systems and cannot be ignored when designing a methodology for situational systems.

Physical tokens are also used to hand over situations to other actors involved in an activity. This is illustrated in the flight landing system where controllers routinely hand strips to other controllers. Successful handover is helped by the receiving controller having to physically handle the token.

In these systems tokens and other parts of an agent’s environment play a critical role in representation. A human agent uses tokens in the environment to reason about activity and to notice situations. Tokens contain some, but not all, information about situations and activities. Often it is the relationship between tokens that completes the picture for an agent. Contrastingly, deliberative systems require a model of the world where objects and their properties are self-contained and correspond with objects in the human agent’s environment.

The use of physical tokens requires that careful attention is given to the capabilities of computerised technology such as mobile devices when designing information systems. Poor selection of devices that do not deliver the required palpability, capacity for manipulation, or representational ability may place the success of the whole information system in danger. Further, environments of computerised parts of the information system must be carefully designed with these findings in mind.

Following our examination of manual situational systems, the methodology still consists of six steps. The details of specific steps, however, must be augmented with results from our analysis of the systems. These largely give insights into the implementation of Stage 6 in the method. First, in the manual systems explored, physical tokens are often employed by agents to represent parts of activities and contain information about situations. These are seen to be important for both situation recognition and for reasoning about action. Arguably, they do not reduce simply to the information displayed on them (Mackay et al., 1998). This gives an important insight into the unique character that the informational component of situational systems should possess. Although not all situational systems may need to employ the idea of physically manipulable tokens as the representational component of the system, it seems that it is a prudent approach to consider this possibility in conjunction with information and communication technology as a possible form that part of the information system might take. For instance, in a follow-up study of the manual air traffic control system described above, Mackay et. al. (1998) trialled a computer-enhanced physical token-based substitute for the flight strips. Clearly new ICT technologies, such as mobile devices and ubiquitous computing can play a role here. Second, where relevant, the environment of the agents must be designed to aid situation handover between agents.