Three Generations of Work Analysis Methods (Rasmussen 1997) • Normative models : how a system should behave • Tayloristic work methods analysis, traditional task analysis, GOMS analysis • Descriptive models : how a system actually behaves in practice • Critics from anthropological, activity theory and naturalistic decision making communities • The assumptions they make about human work are not realistic and not very useful • There are problems in deriving implications for design from descriptive work analysis techniques
Three Generations of Work Analysis Methods (Rasmussen 1997) • Formative models : Requirements that must be satisfied so that the system could behave in a new, desired way. • To help us specify the design attributes that computer-based information systems should have to satisfy
Normative Approach : Task Analysis • Definitions • To examine “ the tasks that must be performed by users when they interact with systems” • The study of what an operator is required to do in terms of actions and/or cognitive processes to achieve a system goal • One best way • For identifying the ideal ways in which the job should be performed
Three levels of task analysis techniques • Level 1 : input-output • Identify inputs, outputs, and constraints • Constraints that must be taken into account in selecting the actions • Ex) Rate of gasoline consumption • Two relevant constraints : the number of kilometers in a mile and the number of liters in a gallon • Several different set of actions • Using a calculator : Reading → Typing into the calculator • Using metal arithmetic :
INPUTS OUTPUTS Task Miles traveled since last fill-up Automobile gas consumption rate in Km/L Gallons at this fill-up CONSTRAINTS : 1 Mile = 1.609 km 1 Gallon = 3.785 Liters Rate of gasoline consumption in an automobile
Level 1 : input-output • A very high level product description of the task • 행위에 대한 constraint는 다른 형태일 수 있다. 연료소비의 예에서는 변수들 사이의 관계였다. Constraint는 어떤 순차적 절차일 수도 있다. • 이들 constraint를 고려하지 않고 정확히 작업을 수행하는 것은 불가능하다. 그러나 작업이 실제 어떻게 수행되는 것과는 독립적이다. • Constraint 들은 문제상태공간을 줄이는 기능을 한다. 그러나 단일 행동절차를 지적하지는 않는다. 단지 가능한 행위 절차의 선택폭(자유도)을 제한한다.
Level 2 : Sequential Flow • To identify ordered sequence of actions • Flowchart of the process that workers should follow to perform the task. • Ex) Rate of gasoline consumption • Read current odometer value • Read odometer value at last fill-up • Calculate the difference to obtain miles traveled since last fill-up • EX) Fig 3.2 • This level of task analysis is usually dependent to the device workers currently have available to perform the task • If you had a trip odometer, --- just read
Level 3 : Timeline • To identify ordered sequence of actions with duration estimates for each action • The most detailed of all • Ex) Rate of gasoline consumption • 0-1 s: Read current odometer value • 1-2 s: Read odometer value at last fill-up • 2-3 s : Calculate the difference …. • Ex) Fig. 3.3 • Only one right way to perform this task • All of the discretion has been eliminated.
Goal Goal Goal Constraints VS Instructions(Three levels of task analysis) Level 1 : Input-output Level 2 : Flow Sequence Level 3 : Timeline
Implications • Different forms of work analysis make different assumptions about the nature of work • So they lead to different designs, which in turn, lead to different types of guidance to workers.
Conclusions • Several advantages of constraint-based approach • More discretion • Greater variability in action • Fewer assumptions about the properties of the device • It is more likely that the new design will result in new functional possibilities, • rather than being constrained by designer’s current assumptions about functionality • Task-artifact cycle : Instruction-based approach의 단점 • 두 가지 접근방법의 장점을 활용하자
+ Error Action Output Goal Worker Plant - The view from control theory • To understand goal-oriented behavior, using control theory • In a conceptual level A simple model of goal-directed behavior Goal (g), Output (o), Error (e), Action (a) Control strategy by the worker (W), dynamics of the plant (P)
A simple model of goal-directed behavior • 관계식 • action(a) = strategy(W) * error(e) • error (e) = goal state(g) – output(o) • a= w(g-o) • 만약 (w, g, o)를 안다면, action sequence를 예상할 수 있다. (초기상태 t=0, g=o) • 초기상태가 알려지지 않으면, action을 예측하지 못함. • 만약 W가 불확실하면, action을 예측하지 못함.(STS 에서 여러 가지 전략 가능)
Disturbance + Error Output State Action Goal Worker Plant - A complex model of goal-directed behavior A complex model of goal-directed behavior • Disturbance (d) : • Factors that affect the state of the plant in ways that have not been or can not be, anticipated by system designer
A complex model of goal-directed behavior • 관계식 • a=(g-d) / P • d의 예측이 불가능하므로, a 의 예측은 불가능 • 동일 목표를 위해 다른 행위 가능, 같은 행위가 다른 시간에 다른 영향을 줌 • Context-conditioned variability (motor control) • Unanticipated variability (cognitive engineering) • Situated action (cognitive science)
Resolution(Closed/Open systems) • The more closed a system is, the more amenable it is to instruction-based forms of task analysis. • Open system gives rise to context-conditioned variability • Workers must adapt online in real time to disturbance that cannot possible be foreseen by analysts (Hirschhorn 1984). • 참고문헌 : 사전에 예측불가: Ujita, Kawano & Yoshimura, 1995)
Constraint-based approaches • Negative feedback models : constraint-based structure, not instruction-based • Discretion is not the same as complete freedom • Not advocating that workers be allowed to do whatever they want • The discretion and flexibility that we advocating is bounded by constraints.
An unresolved problem • Constraint-based task analysis can lead to new design functionality, making it more likely that productivity will be improved. • A particular goal to be achieved may not be identifiable beforehand : an unanticipated emergency in a nuclear power plant. • CBTA의 한계 → Work Domain Analysis로 극복
An unresolved problem • How can we know what the goal should be? • To require workers to “Note any problems” by Shepherd(1992) • It merely provides a place holder for what workers are supposed to do • Little to identify the information or knowledge that workers require to cope with • Even constraint-based task analysis do not provide a very satisfactory basis for dealing with unanticipated events.
Dealing with unanticipated events • Ex) Spatial navigation (Thorndyke and Goldin 1983) • How people learned to find their way • Two types of spatial knowledge • Procedural knowledge : sequence of actions • Survey knowledge ; spatial relationships between locations and routes in an environment • Navigation aids • Procedural knowledge can be embedded in directions • Survey knowledge can be embedded in a map
Dealing with unanticipated events • Direction 과 Map의 장단점
Relevance to work analysis • Two forms of work analysis • Task representations (like directions) • What goals they should achieve or how they should be achieving them • Work domain representations (like maps) • To describe the structure of the controlled system • Task VS Work domain • Task : what workers do • Work domain : what workers do it on (i.e., the object of action) ★ Those familiar with computer programming - Control structure : task - Data structure : work domain
Summary • Most existing task analysis techniques are not very useful • Instruction based TA • Not suited for complex STS • To underestimate context-conditioned variability • Constrained based task analysis are better suited for complex STS • Contribute in flexibility, productivity, worker health and on-the-job training • But not capable of dealing with unanticipated events • Work Domain Analysis for unanticipated events • Work domain analysis + Constrained based task analysis