Human Performance Modeling

advertisement
Human Performance Modeling
○ Model – ‘a simplified representation of a system
or phenomenon, as in the sciences or
economics, with any hypotheses required to
describe the system or explain the phenomenon,
often mathematically’
○ Perception and attention, action or motor control,
and cognition
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
General Issues
○ Misconception -- construction of “intelligent
system”  AI
○ predictions of human performance on human
factors problems not necessarily from basic
psychological processes
○ All models are abstractions and by necessity
omit certain details
○ Accuracy and generality
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
General Issues
○
○
○
○
Simplicity and understandability
Free parameters – how to set and interpret
Validation – correlation, mean deviation
Gains
○
○
○
○
Specificity vs. qualitative and vague
Modeler independent
Quantitative predictions
Explanation for observed differences
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Perception and Attention
Signal Detention Theory (SDT)
○
○
○
○
○
Make a binary judgment btn signal and noise
Hit, False Alarm, Correct Rejection, Miss
p(H) + p(FA) =1; p(CR) + p(M) =1
Type I error (FA) & Type II error (Miss)
Decompose performance into detection
efficiency (d’) and criterion parameter (β)
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Perception and Attention
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Perception and Attention
Visual Search Models
○ Feature integration theory (Treisman and
Gelade, 1980)  Salience map (Itti and Koch,
2000)
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Perception and Attention
Visual Sampling Models
○ Senders (1964, 1983) – a signal at W Hz can be
reconstructed by sampling every 1/W s
○ Wickens (2008) – Salience, Effort, Expectancy,
Value (SEEV) Model
○ p(A) = sS – efEF + (exEX)(vV)
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Perception and Attention
Workload Modeling
○ Neither commonly accepted definition nor how
to measure it
○ Psychological refractory period (PRP) paradigm
– response selection bottleneck model (Pashler,
1994): perception, response selection & action
○ Multiple resource theory (Wickens, 2002 and
2008) – the stages, the codes and modalities
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Action & Motor Performance
Hick-Hyman Law
○ Information entropy H = log2(n+1)
○ RT = a + bH
○ 𝐻=
1
𝑛
𝑖=1 𝑝𝑖 log(𝑝
𝑖
+ 1)
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Action & Motor Performance
Fitts’s Law
○ MT = a + b*ID
○ ID = log2(2A/W) – Fitts (1954)
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Action & Motor Performance
Manual Control Theory
○ Continuous tracking task
○ Between the desired and their actual behavior
○ Transfer function
○ As system frequency increases, the gain
decreases and the lag increases
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Action & Motor Performance
Manual Control Theory
○ Crossover model (McRuer & Jex, 1967)
○ Two crossover points: the frequency at which the
gain is zero and the frequency at which the lag
reaches 180°
○ Optimal control model (Pew & Baron, 1978)
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Action & Motor Performance
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Memory & Cognition
Historical Perspective
○ GPS (Newell & Simon, 1963)
○ computational models could effectively capture key
elements of human cognitive behavior
○ “modal” model of memory (Atkins & Shiffrin,
1968)
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Action & Motor Performance
Routine Cognitive Skill and GOMS
○ KLM-GOMS
○ CPM-GOMS
○ NGOMSL
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Action & Motor Performance
Models of Judgment and Decision Making
○ Optimal behavior – A baseline of comparison for
human performance  SEUT, Prospective
theory, EBA
○ Lens model (policy capturing)
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Integrated Models
Task Network Modeling
○ Network model – a modeling procedure
involving Monte Carlo simulation
○ Decomposition of the Task into discrete subtasks;
PERT chart
○ Nodes represented by a statistically specified
completion time and a probability of completion
○ SAINT, Micro Saint Sharp, IMPRINT
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Integrated Models
Cognitive Architecture
○ an embodiment of “a scientific hypothesis about
those aspects of human cognition that are
relatively constant over time and relatively
independent of task”
○ The mid-1990’s when including mechanisms for
perception and action as well
○ EPIC (1995 & 1997), ACT-R (1998) & QN-MHP
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Integrated Models
Cognitive Architecture
○ Several modules in ACT-R
두정엽
전두엽
후두엽
측두엽
소뇌
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Integrated Models
Cognitive Architecture
○ Drawbacks
○ Knowledge in ACT-R code
○ S/W integration problem with a rich simulation
environment
○ Setting free parameters
○ Exposition of ACT-R not always straightforward
INDUSTRIAL ENGINEERING
COGNITIVE ERGONOMICS LAB
Download