The theory of actions underpinning our methodology is somewhat unconventional, drawing inspiration from the philosophical thinking of Johansson (1989) and Searle (1983). In the field of information systems, through the concept of a transaction, actions are traditionally conceptualised as causing changes in state, with the desired outcome being achievement of the goal state. We, however, eschew this idea that change can be understood as like a film strip of states of the world with action joining each frame. Instead we view the world through the dynamic lens of action itself and regard every action as simultaneously having both purposive (goal-like) and performative aspects.
The following expands on this understanding of actions. Of particular importance for the methodology are the following three properties of actions: that they are always situated in the environment, that they are multi-scale in nature and that they are dependent on the execution of other actions for instantiation as part of the action system. These properties are illustrated using material from the second case study of the project, conducted in a chemotherapy ward in a large hospital.
Structured information systems analysis and design methodologies tend to ignore the environment in which actions occur, representing actions as if they occur in a virtual space. In contrast, it is crucial to our analysis of action to take account of the environment in which actions occur. Consistent with our focus on action, we conceptualise the environment as all the actions available to an actor, that is, as the action possibility space. This means that if two different environments allow the same set of actions, they are equivalent for our purposes.
We analyse the environment in terms of structures in the environment that constrain or enable action. Gibson (1979) gave the name ‘affordance’ to the opportunities for action that structures in the environment provide to a particular class of actor. For example a chair affords sitting to a person; a hollow log may afford shelter to a small animal. Whereas Gibson focused on physical structures, the idea of affordance can be extended to temporal and organisational structures too. Physical structures include space, things and the arrangement of things. Organisational structures include roles and norms.
Every action is conducted by an actor in time and space, making use of resources. We define the actor, location, time and resources associated with an action as the action context. In other words, the action context is the particular dimensions of the environment in which a particular action occurs. The specifics of the action context mean that particular actions are constrained and enabled, depending on the affordances of the environment in which the particular time, place, actor and resources are located. The actors in whom we are most interested are human actors, although an organisation itself can be conceived as a macro actor performing actions.
Table 1 shows some of the actions undertaken as part of the administration of chemotherapy and the particular action context in which they occur. From the table it can be seen that a blood test can be taken in an external pathology lab, the hospital pathology lab, or the chemotherapy ward. It can be taken by pathology staff or a nurse using particular equipment to draw a sample of blood from a patient. Implicit in this is the idea that there are locations which are not suitable for taking blood, resources which can not be used to take blood, and actors who can not take blood. In other words, the action context indicates constraints on where, how and with what, blood can be taken. The action context also can provide clues about what action is to occur.
|
Action |
ACTION CONTEXT |
|||
|---|---|---|---|---|
|
Location |
Actor |
Resources |
Time |
|
|
a. Sign chemotherapy order |
Hospital |
Doctor |
Chemotherapy order pen |
All of these are to be done before patient appointment. Actions a. and b. need to be done sufficiently before patient appointment to allow treatment preparation |
|
b. Approve chemo orders |
Hospital |
Pharmacist |
Chemotherapy order |
|
|
c. Prepare treatment |
Pharmacy |
Pharmacy staff |
Compounds, lab, etc |
|
|
d. Assess patient as well enough for treatment |
Hospital |
Nurse |
Blood results Patient observations |
|
|
e. Give blood test |
External pathology lab Hospital pathology lab Chemo Ward |
Pathology staff Nurse |
Syringe, etc Patient |
|
|
f. Check blood test results have arrived |
Hospital |
Hospital staff |
Information system |
|
|
g. Analyse blood results |
Hospital |
Nurse |
Blood results |
|
|
h. Take observations |
Hospital |
Enrolled nurse |
Thermometer, blood pressure meter, etc |
|
|
i. Couch made available |
Treatment ward |
Previous Patient |
Couch |
|
|
j. Patient is in ward |
Chemo Ward |
Patient |
Transport to hospital |
|
Actions are multi-scale in nature. This means that both actions and the action context can be specified at different levels of detail or grain size. Another way of saying this is that an action can be expanded into a set of lower level actions that occur in valid particularisations of the context of the higher action. For example, compare the following descriptions of the same action from Table 1:
‘Gave a blood test’
Nurse applied the tourniquet to the patient’s arm, spoke some words of reassurance to the patient, and inserted a syringe into a vein in the patient’s arm. He raised the plunger of the syringe, withdrew it from the patient’s arm and transferred the blood into two vials which he then labelled.
Johansson (1989) shows that actions have the peculiar trait of being ‘temporally inclusive’. Roughly speaking, this means that each action can only be instantiated in time if those other actions that make up the action at a smaller grain size are instantiated within the duration of this action. For example, the duration of the action in which a blood test was given includes the duration of applying the tourniquet and the duration of inserting the syringe into the vein and withdrawing it full of blood. We suggest that this is a particular case of the more general case of the contextual inclusiveness of actions. For example, each action can only be instantiated in space if other actions (those that make up the action at a smaller grain size) are instantiated within the space of this action. This situational inclusiveness means that an actor is always simultaneously engaged in actions at all grain sizes. While inserting a syringe into a vein in the patient’s arm, the Nurse is simultaneously taking a blood sample and giving a blood test.
We can understand this contextually inclusive relationship between actions of different grain size (in this case, inserting a syringe into a vein, taking a blood sample and giving a blood test) using Johansson’s (1989) notion of an action abstraction hierarchy (here the term ‘abstract’ refers to abstraction away from particularities rather than meaning less ‘real’).
The triangle in Figure 1 represents an action abstraction hierarchy in an action system, with the ellipses representing actions. Each action in the hierarchy refers to both the purpose and performance of action. Each action that is above other actions connected to it is the purpose of the action(s) below it; that is, it is why these subordinate actions are conducted. The collection of actions connected below the action is an elaboration of how the action is performed. This means that for any particular action, asking why that action is conducted moves one up the action abstraction hierarchy to more abstract levels. Asking how that action is conducted moves one down the action abstraction hierarchy to more concrete levels. The lower you go down the action hierarchy, the more is specified about the action context (that is, details of actor, location, time and resources). In other words, the action becomes more situated in a specific practice. By the same process, the implementation details of the higher-level action become specified more precisely as a set of more detailed actions.
As the hierarchy is ascended, the goal aspects of actions are emphasised; the performative aspects are emphasised in descent. In principle, each high level action could be specified with more and more precision as we move down the action hierarchy while the very apex of the triangle can be thought of as the ultimate purpose of the actions contained in the triangle. The triangle represents the domain space of all combinations of actions, actors, locations, resources and timing that will achieve the goal at the apex of the triangle. Note that the action abstraction hierarchy is an analytic device and the representation of it in two-dimensional space is indicative only rather than a precise rendering.
We can illustrate this abstraction/concretisation relationship between the actions listed in Table 1. For example, the actions, e. give blood test, f. check blood tests have arrived, g. analyse blood results and h. take patient observations, are all aspects of how the patient is assessed as well enough for treatment (action d). In turn, action d is an expression of why actions e, f, g and h are performed. This is depicted in Figure 2.
An action dependency is another type of relation between two actions. It describes the case when one action is dependent on completion of another action for instantiation as part of an action system. Whereas the vertical relationship in Figure 1 was conceptualised as ‘why-how’, this horizontal dependency relationship can be conceptualised as ‘depends on/condition for’.
Figure 3 shows in bold the action dependencies of some of the actions shown in Figure 2. Actions on the right are dependent on the completion of those actions on the left, to which they are joined, for instantiation in the action system. This means that the actions on the right occur later in time than those actions on their left, to which they are joined. Note that what we call an action dependency relates to the rules of practice and is not necessarily the same as logical dependency or feasibility. It is quite feasible that hospital staff could check whether blood results have arrived before a blood test has been given; these hospital staff may not know the timing of the blood test. However, for this action to be part of the action system being described, checking that blood results have arrived depends on the blood test being given. Similarly, there is no a priori logical reason why analysing the blood results is dependent on someone checking that blood results have arrived. It may seem more logical to bypass this step altogether, notifying the nurse who analyses the blood results of their arrival. The point is that, in the action system this diagram describes, analysis of blood results was dependent on hospital staff checking that blood test results had arrived.
