Humans in complex dynamic systems. Situation awareness, automation and the GMOC model

Transcription

1 Humans in complex dynamic systems Situation awareness, automation and the GMOC model Human-Computer Interaction Dept of Information Technology Uppsala University Bengt Sandblad

2 Agenda Human control of complex dynamic systems Control work Operators Control systems and interfaces A model of human control (GMOC) Situation awareness Automation problems

3 Some examples of complex dynamic control situations What is complex? What is dynamic? What is human control?

4 Train traffic control

5 Train traffic control

6 Train drivers

7 HSC bridge design

8 Other examples Forest harvesters

9 Other examples Stone crushers

10 More examples.

11 Are there problems? Systems are complex and dynamic. Control is indirect, via a control system and an interface. Demands are high (speed, accuracy, quality, safety,.) Control systems and operator interfaces sometimes provide inadequate support. Operator performance not always satisfying (optimality, errors ). Work environment problems. More...

12 A model of control We need a model that helps us to describe, analyse, design, control of a complex dynamic system

13 The GMOC model To control a dynamic system requires: Goal (G) Model (M) Observability (O) Controllability (C) Goal Control Process Model Observe time

14 Goals Goals are often complex Contains conflicts (e.g. safety vs speed) Are: Formal - informal Organisational - individual Operators have their own goals To relate design to the goals, we must understand (all) the goals!

15 Models Models are mental models. Models are individual and subjective. Models are (mainly) developed during work. This takes time! Different operators often have (very) different models. Organisational development of models and control strategies (more unified) can solve many problems.

16 Observability We can only observe what the interface shows. We often lack information and precision in information. Often observations require actions. We can overview much but remember little. Example?? Difficulties to identify and understand complex patterns (requires good design).

17 Controllability We can only control what the interface allows us to control. We can sometimes only control a process at certain times. Different control modes can cause confusion. Time delays make control complex. Problems with feed-back.

18 How to use the GMOC model? Structure for describing and analyzing. Understand problems and difficulties for the organisation and operator. Design so that all requirements are fulfilled.

19 Exemple Train traffic control: Goals not related to observation and control possibilities Can not observe decision relevant data How trains are moving, speed position Presentation not geographically correct Observation do not support development of mental model Control actions can not be taken when the decisions are made

20 Situation awareness Situation awareness (SA) to always be in control, in the loop. Endsley, Garland: Situation awareness analysis and measurements. Lawrence Erlbaum, 2000

21 Situation awareness Three parts: Perception (observation) Comprehension (understanding the significance of the information) Projection (prediction, evaluation of actions) Perception means observation Comprehension means model Projection means dynamic information and model

22 SA cont. Examples of SA? Examples where we do not have SA?

23

24 SA cont. Two different approaches to control: Control by exception Control by awareness An operator can not work without SA An operator will always try to obtain SA This can, if not properly supported, be extremely cognitive demanding and lead to stress, safety problems, cognitive overload etc.

25 Show dynamic information temp min

26

27

28 Automation problems Automation: Mechanical or electronic replacement of human labour (physical or mental).

29 Problems with automation Automation surprises (hinders SA) Difficulties to predict The irony of automation No help when it is needed Allocation of functions (human vs. machine) MABA-MABA analysis Authority dilemmas (e.g. in aviation)

30 Autonomy The problems with autonomous automatic systems (autopilots) The automatic systems does exactly what it is told to do (non-autonomous) The automatic system has its own plans (autonomous) this can cause severe problems! Example operators turn the automatic system off ( irony of a ).

31 OR...