Cognitive Engineering Flashcards
Difference between cognitive engineering and HCI
- top-down vs bottom-up
- need to know what people are doing with technology in cog eng; HCI = graphics
Cognitive engineering triad
(1) agents (2) world (3) technology artifact
Artifacts
things that are designed
Design the window, but…
don’t eat the menu
Behaviorism
context independent
Industrial engineering tradition
tasks are independent and quantifiable
Supervisory control
We just monitor
Why complete automation won’t work
Impossible to predict unknown
Issues relating to complexity
- risk
- lots of interconnected parts
- dynamic
- time pressure
Idiot-proof myth
- make systems safe from their users or operators
- myth because people design them in prospective environments / designers are people too
Deskilling myth
- automation replaces skill
- only unskilled labor left
- myth because they need new skills to deal with automated system
- automate vs. informate
- automation leads to new opportunies
Top down approach
theoretical / directed
Bottom-up approach
evidence / experiential
Rationalized work
- business process
- procedures
- work flow
- assembly line
Activity oriented view of work
opposite from rationalized work
Synchronous
telephone
asynchronous
colocated
same location
distributed
different locations
authority-responsibility double bind
can’t control comp, but responsible for outcomes
Schema
- data structures
- stereotypes
- lead to expectation about meaning
Phenomological
- conscious
- what we think schema are
Task-artifact cycle
Small changes cause huge change, new opportunities
Cognitive artifact
Tool that replaces or acts cognitively with memory and planning
- makes cues apparent to retrieve schema
Distributed cognition
- system of people and tools
- cognition emerges from the interaction of people and tools
Properties of distributed cognitive system
- representation is observable
- multiple people’s knowledge
- communication and coordination
- memory/computational aids
Overlearned
happens when the world is predictable
Why model work domain?
- to create better displays
- to create better documentation
- to make human-centered automation schemes
- communication and task coordination / social-technical organization
- training
CWA
- WDA
- CTA
- Strategies
- Social organization
- knowledge/skills
CTA in CWA
the “what”
Strategies in CWA
the “how”
Social/organizational in CWA
the “who”
Concept mapping
allows mapping of semantic network
data collection
- analytic formalisms / representations
- good systems explicitly ID system/info requirements
Design traceability
documenting the relationships between layers of information
Task analysis
- set of goals and subgoals
Normative
what should be done
Descriptive
What actually gets done (task analysis)
Problems with HTA
- will always be unanticipated events
- even for anticipated events, cannot develop complete procedures due to uncertainty and complexity
- differences in expertise lead to different strategies
Abstraction hierarchy
multiple levels of system description
Going down in AH
the “how”
Going up in AH
the “why”
Levels in AH
Functional purpose Abstract functions/priorities/abstract constraints General Processes Physical Function Physical Form
What do nodes in AH represent
Things that need to be measured for effective system control
Functional Purpose
Produce radiators/be cost effective
Abstract Functions
Conservation of materials; inventory flows; monetary flows; safety; labor constraints
General Processes
Forming; brazing; assembly; storage; testing; transporting
Physical Function
Copper; aluminum; sheets; pipestock; assembly lines
Physical form
Where is it? Quality of it; layout
Control Task Analysis
What is necessary to develop system
Inputs->outputs
EID
- develop controls/displays
- allow support/lowest level of cognitive control possible
Example of EID
Turning the game of 15 into the tic tac toe game
Information limits that lead to bias
- unpredictable
- cannot feasibly enumerate
- utilities / probabilities unknown
Heuristics
simplify problems, make them tractable, results in predictable biases
Ecological decision making
(1) world (2) (cognition) choice, processes (3) action
feedforward
- low time pressure
- errors/changes
- costly
- smaller range of acceptability
- known/simple model of the world
feedback
- high time pressure
- cheap/easy to correct errors
- many acceptable solutions
- model is complex/unknown, many interactions
Situated action / ecological psychology
decisions are very context-dependent, can’t model cognition
Brunswik’s view on ecological psychology
judgment and representative design -> vicarious functioning
Affordance
what the environment offers to the animals, what it provides or furnishes
Structure of organism/environment relation
world = true state
proximal cues = mediating
perception = judgment
(ambiguity between world and proximal cues)
Lens model
criteria->cues->judgment
environmental predictability; judgment consistency
regression environment; regression judgment
achievement = corr(total yE, yS) knowledge = corr(ind yE, yS)
Cognitive engineering
Engineering of human-technology systems in which humans and technology work together to solve problems
What becomes apparent when you study a work system in transition?
old, implicit work is revealed
Clumsy automation
increases work at times of high demand, works primarily when tasks/demands are low
Why is deskilling a myth?
new technology transforms workers’ skills
Irony of automation
operators have to pay more attention so that they can react in urgent situations
Problems with rationalization of work
fails to capture implicit elements
Informate
give information rather than automate
Propositional knowledge
world is represented in symbols connected to each other
Analogical knowledge
world is represented in images - discrete, but able to be manipulated
Procedural knowledge
knowledge of how to do things
Means-ends approach
(1) where they currently stand
(2) what they are allowed to do to bring them closer
depends on individual strategies and perceptual properties of task
Role of metaphor
People come up with interpretations of target domain based on their understanding of metaphor domain
Difference between novices and experts
different schemas - abstract vs. surface, chunking
Rasmussen’s three categories of behavior
skill-based (signals)
rule-based (signs)
knowledge-based (symbols with abstract concepts)
Point of external representation
to make problem solving direct (skill/rule based)
