1. The MURI taxonomy and training for military tasks
MURI Annual Review
September 7, 2007
Bill Raymond
I’ll talk about:
Current status of the MURI training grant taxonomy
How the taxonomy is being used to understand laboratory tasks.
How the taxonomy may be used to predict the effects of training on military tasks.

2. The training MURI project
Goal
Provide principles that can be used to predict effects of training on military tasks.
Tool: 4-dimensional taxonomy
Task type
Training method
Performance (context & measures)
Training principles
Mechanism: Empirical data and task modeling
How can results be applied to military tasks?
• GOAL: provide principles that can ultimatel be used to predict the effect of training on military tasks;
• TOOL: 4-dimensional taxonomy
The MURI taxonomy will allow us to understand generic training scenarios through taxonomic categorizations in 4 dimensions:
task type,
the method by which a task is trained,
the context in which a task is performed and the measures by which performance is assessed,
and, orthogonal to these dimensions, a fourth dimension of training principles, which will allow qualitative predictions of task performance for any given taxonomic categorization.
• The methodologies by which the goal will be accomplished include empirical data from laboratory experiments and modeling of laboratory tasks using the IMPRINT and ACT-R platforms.
• The major question I’d like to address is: How can our results be used to understand the effects of training on military tasks?

3. Applying results to military tasks
Probe taxonomic space
Select simple tasks that can cover a portion of the space
Decompose tasks into taxons.
Manipulations isolate training and performance effects.
Different experiments allow generalization to training principles within taxonomic space.
Analyze military tasks
Military tasks are complex but composed of simple tasks (like laboratory tasks) or task taxons.
IMPRINT taxons widely used for describing complex, militarily relevant tasks.
The MURI taxons map onto IMPRINT taxons.
To answer this question, we start by noting that applying the results of the taxonomic analysis to military tasks can be broken into two steps.
The first is our ongoing research [go through steps];
The second is understanding military tasks and their relation to our results.

4. Overview MURI/IMPRINT task taxon mapping
Training and performance context dimensions
Planning matrix: MURI coverage of taxonomic space
Taxonomic analysis of RADAR task
Examples of simple tasks in real military tasks
[Overview of my talk]
I will show the links involved in accomplishing the application of our results to military tasks.
First, because of the appropriateness of the IMPRINT task taxonomy for understanding complex military tasks, I will show the relation between our MURI task taxonomy and the IMPRINT taxonomy.
Second, I’ll review the training and performance context dimensions that we introduced last year, with some changes and additions that reflect some insights we’ve gained this year.
Third, I’ll show the extent to which the MURI research covers our taxonomic space, as summarized in the so-call “planning matrix”.
Fourth, I’ll give a specific example of a task analysis of a laboratory task, the militarily relevant RADAR task. The analysis includes breakdown of the task into taxons and also a breakdown of the training and performance context dimensions about which our expeerimental manipulations will provide information.
Finally, I’ll show you some examples of military tasks and how they may be thought of as composed of simple tasks.

5. MURI task taxons and IMPRINT taxons
MURI taxonomy (white on left):
hierarchical
reflects cognitive processes
focuses on mental tasks (with little emphasis on physical tasks)
IMPRINT taxonomy (colors, to right)
7 relatively coarse-grained taxons
mapping from MURI to IMPRINT not one-to-one

6. Analyzing training and performance
Training variables - during skill learning:
How was the skill taught?
What kind of practice did learners get?
How did practice relate to the way the skill will be used?
Performance context variables - at skill use:
How does expected performance relate to training?
How long has it been since training?
Did learners get refresher training?
Pedagogy }
Practice }
Review of the training and performance context dimensions of the taxonomic space.
Performance

7. Pedagogy parameters Method Instruction (= default) Demonstration
Discovery
Simulation
Immersion
Modeling/mimicking
Learning location (local = default, remote/distance)
Discussion/Q&A (no = default, yes)
Individualization (no = default, yes)
The pedagogy parameters are not being extensively investigated in the MURI.
The taxons we have chosed are consistent with training systems we reviewed at the Simulation Systems Research Unit (SSRU) in Orlando in February.

8. Practice parameters Scheduling of items in trials and sessions
Number and difficulty of items
Spacing (massed = default, spaced, expanding/contracting)
Distribution (mixed = default, blocked)
Scope of practiced task (partial, whole = default, whole + supplemental)
Depth of processing (no = default, yes)
Processing mediation (no = default, yes)
Stimulus–response compatibility (yes = default, no)
Time pressure (no = default, yes)
Feedback - presence and type (no = default, all trials, periodic)
Context of practice
Distractor/distractor task (no = default, yes)
Secondary activity (no = default, yes)
(Note: We’ve added number and difficulty of items to the scheduling category)

9. Performance context parameters
Transfer
New items, item order, or distribution of items
New context (relative to training)
New task (relative to training)
Delay interval to task performance
(default = none, time period)
Refresher training
(default = no, schedule)

10. (Planning matrix)

11. A cognitive model of RADAR (exp 2)
Reasoning/
Problem solving  
Motor planning, Motor response
Auditory processing
perceive tone
deviance decision
count deviant tones  
Memory:
Symbolic representation, Visual representation  
tone count #  
fire
response
post-response
decision
Visual processing
scan display
monitor targets      
target
response

12. access representation
A cognitive model of digit data entry 
1395
Language/reading
Symbolic representation
Motor planning
Fine motor manipulation
read digit
represent digit
access representation
plan response
(Cognitive model of Digit Data Entry, a simpler tasks, for comparison.)
access plan
execute plan 
1395 

13. RADAR task decomposition
components
Training features
Performance context
Pedagogy
Practice
Visual
Method:
Simulation
Scheduling:
Sessions (blocks, weeks)
Item difficulty (letters/#s, planes)
Transfer:
New task (tone counting; fire decision)
Numerical analysis
Context:
Distractor (visual detection)
2ndary task (fire decision)
Information
processing
Item difficulty (VM/CM; memory load)
Distractor (tone counting)
Fine motor -
discrete
IMPRINT task taxons

14. Laboratory tasks in military tasks

15. Summary
Collect effects (principles) on MURI taxons from experimental manipulations.
Compare and generalize different measures across simple tasks within MURI taxons.
Consolidate generalizations onto IMPRINT taxons.
Apply to military tasks:
Decompose tasks into IMPRINT taxons.
OR, Decompose into simple tasks, to the extent possible
Here is a summary of the process we envison can be used to apply the results of our MURI research to the understanding of effects of training on performance of military tasks.

16. END

17. Data entry tasks in planning matrix

18. Task, training, and performance matrix
Task components
Training features
Performance context
Pedagogy
Practice
Visual
Numerical Analysis
Information processing
Fine motor - discrete
Fine motor - continuous
Gross motor - light
Gross motor - heavy
Communication (reading & writing)
Communication (oral)
Data entry
IMPRINT task taxons

19. Task by Pedagogy parameters
Task components
Pedagogy
Method
Learning location
Discussion/Q&A?
Individualized?
Visual
Numerical Analysis
Information
processing
Fine motor - discrete
Fine motor - continuous
Gross motor - light
Gross motor - heavy
Communication (reading & writing)
Communication (oral)
Data entry
(Instruction)
Classification
Inst/Discovery
IMPRINT task taxons

20. Stimulus-response compatibility
Task by practice
Task components
Practice
Scheduling
Scope
Processing depth
Processing mediation
Stimulus-response compatibility
Time pressure
Feedback
Context
Visual
Numerical Analysis
Information processing
Item repetition,
# Sessions, Spacing
Part/
whole
Yes (presentation format)
Yes (prior knowledge)
No (Input-output
Format)
Yes (response & accuracy)
Distractor/2ndary
activity (vocal activity)
Fine motor - discrete
Fine motor - continuous
Gross motor - light
Gross motor - heavy
Communication (reading & writing)
Communication (oral)
Data entry
Data entry
IMPRINT task taxons
Data entry

21. Task by performance parameters
Task components
Performance context
New context
New task
Delay interval
Refresher training
Visual
Numerical Analysis
Information
processing
Yes (typing hand, output configuration)
Yes
Fine motor - discrete
Fine motor - continuous
Gross motor - light
Gross motor - heavy
Communication (reading & writing)
Communication (oral)
Data entry
Data entry
IMPRINT task taxons
Data entry
Data entry