Scaled Worlds: Tractable, Realistic, and Engaging? Wayne D. Gray, Ph.D. George Mason University...

Scaled Worlds: Tractable, Realistic, and Engaging?

Wayne D. Gray, Ph.D.

George Mason University

Human Factors & Applied Cognition

gray@gmu.edu

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 2

Motivation (1)

• Outstanding in My Field– Ft. Hunter-Ligget, NTC, Ft. Knox, Ft. Benning, Ft.

Hood, Schofield Barracks, W. Germany

• The Unfinished Revolution– Problem Finding (Newell & Simon, 1972)

• Protocol Analyses

• Real-World Experts

• Real-World Novices

– Unified Theories of Cognition (Newell, 1973)

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 3

Motivation (2)

• Resistance from Academic Psychology?– Traditional Experimental Psychology slowing

yielding (work well-accepted by Cognitive Science Society types)

• Resistance from Human Factors Psychology?– Traditional HF Psychology -- Suspicious of

Theory– Ethnographic HF Psychology -- Suspicious of

Abstractions of the Task Environment

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 4

Topic

• Simulations of the Task Environment

• How do various Task Environment Simulations differ?

• What inferences do different types of Task Environment Simulations permit us to draw?

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 5

Outline

• The Problem

• Distinctions, Dimensions, Definitions

• Examples

• Scaled Worlds

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 6

The Problem

In field research there is too much complexity to allow for definite conclusions; whereas, in laboratory research there is too little complexity to allow for interesting conclusions.

Brehmer, B. & Dörner, D. (1993). Experiments with computer-simulated microworlds: Escaping both the narrow straits of the laboratory and the deep blue sea of the field study. Computers in Human Behavior, 9(2-3), 171-184.

Three Dimensions of Simulated Task Environmentsdifficulttractablerealisticunrealisticengagingboring

Researcher

Real-World Task

Participant

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 8

Definitions: Scaled World

• Beginning with a Complex, Real-World Task– A scaled world abstracts certain functional

relationships, while paring away others– Can be multiple scaled worlds of the same

complex, real-world task that differ on which functional relationships are abstracted and which are pared away

– The nature of the research question determines what is abstracted and what is pared away

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 9

Definitions: Tractable

• The “complexity” issue– Can the researcher pursue the question of interest?

• Includes issues such as– Data collection (the right data, at the right grain size,

with the right timestamp)– Learnability issues– Usability issues– Computational cognitive modeling issues -- can the

model interact with the system?

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 10

Definitions: Realism

• The simulated task environment is realistic to the situation to the degree to which experiences encountered in the environment will occur in the real-world

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 11

Definitions: Engagement

• The simulated task environment is engaging to the degree to which it involves and affects the participants; that is, the degree to which they take it seriously

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 12

Dimension difficulttractable

difficulttractableReal-WorldHi-Fidelity SimMicroworldSyntheticEnvironmentsScaled Worlds??????

Dimensions

Human OperatorTask in the Real-WorldSimulated Task EnvironmentEngagementextrinsic to the simulationknowledge possessed by the participantRealismrelevantirrelevantintrinsic to the task environment

unrealisticrealisticDimensionrealisticunrealisticReal-WorldHi-Fidelity SimMicroworldSyntheticEnvironmentsScaled Worlds??????

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 15

Dimension boringengaging

engagingboringMicroworldSyntheticEnvironmentsReal-WorldScaled Worlds??????Hi-Fidelity Sim

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 16

How Simulated Task Environments vary on Tractability, Realism and Engagement

• Four examples– Flight Simulator– VCR– Argus– Ned

• Each is followed by a discussion of whether it is or is not a scaled world

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 17

Example: Flight Simulator

• As High Fidelity as $$$ and current technology can provide

• Is it a scaled world? NO

lowmedhigh¿¿engagement??realismtractability

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 18

Example: VCR -- A High Fidelity Simulation of a Very Simple Task

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 19

Gray, W. D. (in press). The nature and processing of errors in interactive behavior. Cognitive Science.

http://hfac.gmu.edu/Publications/pubs.html

lowmedhighengagement realismtractability

Is the VCR a Scaled World?

• No -- more like a high fidelity simulation of a low complexity task

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 20

Example: Argus

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 21

Is Argus a Scaled World?

• No -- Lacks the relation to a complex, real-world task that defines a scaled world

• It is a Synthetic Environment– Focus is on the study of interactive behavior under

conditions of high workload

lowmedhigh engagementrealismtractability

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 22

Example: Ned

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 23

Is Ned a Scaled World?

• Yes

lowmedhigh engagementrealismtractability

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 24

Design Goals for Ned

• Has been carefully designed to abstract certain functional relationships of a complex, real-world task

• The functional relationships abstracted were defined by the goals of the study

• Most other functional relationships have been pared away

• Specifications for Ned arose from– Studying and listening to submarine Approach Officers

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 25

Research Goals for Ned (1)

• Functional Relationships– Project goals entail the description and simulation of the

cognitive processes and memory structures used by Approach Officers in locating enemy submarines

• Impact on Cognitive Theory– To be able to say something new and different about the

expertise required for performing a complex, event-driven task

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 26

Research Goals for Ned (2)

• Generalizability– Our goal is to be able to generalize the data and models of

Ned to a new, unknown submarine, with reduced manning, and new instrumentation

Caveat: “generalizability may be hindered by excessive attention

to mundane realism” (DiFonzo, Hantula, and Bordia 1998)

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 27

Ned on the Tractability Dimension

• Ten displays with no overlapping information• Each object clicked is timestamped to nearest

tick (60hz) and saved to a log file• Researchers can “play back” Approach

Officer scenarios• Computational cognitive models written in

ACT-R can use Ned in same manner that our Approach Officers do.

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 28

Ned on the Realism Dimension

• We decreased realism – by paring away the functional relationships provided

by the Approach Officer’s interactions with his crew and boat

• But, we maintained an information environment in which the functional relationships essential to our goals were preserved

• Used scenarios that represented situations familiar to the Approach Officer

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 29

Ned on the Engagement Dimension

• Ned maintains a reasonable level of engagement– but only for those with much prior submarine

experience -- i.e., not suitable for college sophomores (or for this audience!!)

• The engagement arose from presenting Approach Officers with scenarios, information types, and displays that were recognizable as similar to those encountered in real missions

Dimensions Revisiteddifficulttractablerealisticunrealisticengagingboring

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 31

Summary: Tractability

• Tractability– Microworld & Synthetic Environments ≥

Scaled World > High Fidelity Simulation– Tractability is defined by the needs of the

research question

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 32

Summary: Realism

• Realism for scaled worlds– Hi-Fi simulation > Scaled World > Microworlds or

Synthetic Environments– The realism of scaled worlds is a focused subset of a

more complex, real-world task. – Focus is on a limited set of functional relationships

with a limited set contextual elements (e.g., scenarios, information types, and displays)

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 33

Summary: Engagement

• Engagement for scaled worlds– Scaled worlds present schematics of the real world

situations– Participants supply the engagement via their

motivation and knowledge of the task and domain– Will not be more engaging than the real task

gray@gmu.edu Tractable, Realistic, & Engaging? Slide 34

The End

http://hfac.gmu.edu