Understanding & Reducing Human Errors in Complex Systems: Cognitive Modeling & Engineering, Study notes of Psychology

Download Understanding & Reducing Human Errors in Complex Systems: Cognitive Modeling & Engineering and more Study notes Psychology in PDF only on Docsity!

Cognitive Modeling,

Cognitive Engineering, &

Human Error

Acknowledgments Michael D. Byrne Bonnie E. John

Newell’s 20 Questions Article

 (^) Cognitive psychology has reached a point (35 years ago) where continuing to amass a catalogue of phenomena ceases to be very helpful – what we need is a grand theory of cognition.

Potential Remedies

 (^) “…the ‘normal’ means of science may not suffice.”  (^) What did he mean by that?  (^) Hint: What was the “slippery eel” problem Newell identified?

One Potential Remedy:

The Cognitive Architecture

 (^) Embodies the invariant human cognitive resources and constraints  (^) Used to construct models of human task performance

ACT-R Example

 (^) Let’s say you have a large environment with many devices controlling distal equipment  (^) Like, say, a Navy ship  (^) You want to reduce crew requirements  (^) Meaning more distal control  (^) New functionality must be added, new procedures learned by the operators  (^) That is, things change  (^) How do we design human-machine interfaces and procedures to minimize the risk of error and slowdown?  (^) In even the routine procedures

ACT-R Example

 (^) To understand this, we need to know how people mentally represent routine procedures  (^) Perform experiments to identify factors relevant to human performance  (^) How do people represent the routine tasks that they preform?  (^) How do people represent the space in which they perform those tasks?

ACT-R Example

 (^) We can build models that embody the collection of human performance phenomena for a given task  (^) Architecture: what’s invariant about human cognition  (^) Model: what’s specific to the human and the task  (^) CAT Triad: running the model allows interactions to play out

Engineering Models of Human Performance

 (^) A priori quantitative predictions of human performance  (^) Learnable and usable by system designers  (^) Cover total tasks  (^) Usefully approximate

GOMS

 (^) A Framework for Cognitive Engineering  (^) Based on Model Human Processor  (^) Goals: the objective of the task and sub-tasks  (^) Methods: well-learned sequences of subgoals and low-level actions that can accomplish a goal  (^) Operators: low-level actions  (^) Selection Rules: if more than one method applies, specifies when each should be used

The Model Human Processor

 (^) Three processors  (^) Associated memories  (^) Parameters  (^) Principles of Operation  (^) Quantitative predictions could be made for simple tasks, e.g.,  (^) Speed of animation to create illusion of movement  (^) Position of function keys for most efficient performance  (^) And many more…

Another Simple Example

 (^) Given n equiprobable choices, how long will it take the user to pick one?  (^) Hick-Hyman Law  (^) T = b•log 2 (n + 1)  (^) b = empirically-determined constant  (^) + 1 because there is uncertainty about whether to respond at all, in addition to which response to make

Human Error

 (^) Powerful technologies can have catastrophic consequences  (^) “Be more careful” admonishments don’t work  (^) Systems engineering approach