An AgentBased Electronic Military Labor Market

1
An Agent-Based Electronic Military Labor Market

• Bill Gates
• Mark Nissen
• Graduate School of Business and Public Policy
• Naval Postgraduate School

2
Designing Agent-Based Electronic Employment
Markets

• Objective
• Analyze the technological and operational
  feasibility of establishing a web-based market,
  using intelligent agents, to match naval enlisted
  personnel to specific navy billets.

3
Market-Based Labor Markets
4
Hierarchical Labor Market
Command Allocation
Sailor Placement Assignment
ID rqmts
ID eligibles
Allocate labor
Prioritize requirements
Advertise/list jobs
ID preferences
Match sailors
Negotiate
Agree/consent
Issue orders
Report for duty
Use sailor
5
Personnel Mall

• Multi-Agent System-matching people jobs
• Adapted from supply chain domain
• Shopping mall metaphor
• Intelligent agents represents people/jobs
• Market dis/re-intermediation
• Speed, efficiency, preferences, info overload
• Lacks key properties (2-sides match, clearing)

6
Personnel Mall Screenshot
7
2-Sided Matching

• Game Theory
• Medical residency sororities
• Match explicit ranked preferences
• Match stability - 2-sided

Proposed match Person A Person B Job X
Job Y
8
Experimental Design

• Internal labor market
• 10 sailors, 12 billets (first-come-first-served,
  batch)
• Randomly drawn from pool of 2000
• Subjects
• Students, professionals
• 5 characteristics of job seekers
• 5 characteristics of jobs
• Compare performance
• Humans/algorithm/optimization

9
Results Sailors Rank
10
Matching Results - Rank
Significant at 99
11
Simulation Design

• Internal labor market
• 10 sailors, 12 billets (first-come-first-served,
  batch)
• Randomly drawn from pool of 2000
• Analyze matching performance
• Batch size
• Base case 45 sailors/60 billets/2 weeks
• Preference List Length
• Base case 5

12
Simulation Results

• Quality of matches increases with batch size
• Percent of sailors matched decreases with batch
  size (given preference list length)
• Percent of sailors matched increases with
  preference list length (given batch size)

13
Matching/Optimization (Sailor)
14
Matching/Optimization
15
Sailor Command Preferences

• What are the top sailor and command preferences
  influencing the enlisted distribution process in
  the Aviation Support Equipment Technician (AS)
  community?
• Interview AS community manager and AS detailer
• Conduct focus groups with AS Sailors
• Conduct preference questionnaire with AS sailors
  and command manpower officers

16
AS Sailor Preferences
80
80
64
60
60
43
39
40
31
30
24
20
20
10
0
Family Life
Location
Job
Training and
Incentive
Attributes
Attributes
Attributes
Education
Attributes
Attributes
Important-Chiefs
Important-E6 and Below
17
AS Sailor Preferences
18
AS Command Preferences
19
AS Command Communication
20
Redesign Methodology

• NERISSA - Navy Enlisted Resource Integrated
  System for Smart Assignments
• Targets key processes and support systems,
  keeping many current structures

21
Six Step Distribution Process
3) Sailors view scores and state preferences
through CCC
2) Commands screen sailors for eligibility
score for job-fit
1) Allocation
4) Assign sailors to billets using 2-sided
matching
5) Manage exceptions
6) Audit and write orders
22
Future Research

• Chiefs Detailing Demo-Sept. 02
• Further mall/algorithm integration
• Credits quasi-pricing
• Job priorities market clearing
• Further experimentation/simulation
• Live people jobs
• Full-scale experiments/simulations
• Examine gaming behaviors
• SCM industrial strength implementation

23
Two-sided Matching Example
6
6
2
6
6
3
3
8
4
24
Experimental Design
Sailor/Billet Characteristics Sailors
Billets Pay grade (3) Pay grade (3) NEC
(4) NEC (4) Performance rating (4)
Promotion prospects (5) Preferred location (4)
Job location (4) Personal emphasis Employer
emphasis (promotion/location)
(performance/training)
25
Sailor Characteristics
26
Billet Characteristics
27
Sailor/Command Preferences
28
Results Commands Rank
29
Combined Sailor/Billet Rank
30
Simulation Design
Sailor/Billet Characteristics Sailors
Billets Pay grade (3) Pay grade (3) NEC
(4) NEC (4) Performance rating (4)
Promotion prospects (5) Preferred location (4)
Job location (4) Personal emphasis Employer
emphasis (promotion/location)
(performance/training)
31
Sailor Characteristics
32
Billet Characteristics
33
Sailor/Command Preferences
34
Satisfaction Vs Batch Size
35
Matches Vs Batch Size
36
Matches Vs Preference Lists
37
Sailor Optimization
38
Command Optimization
39
Optimized Sum
40
Family Life Attributes-Top 3 of 11
41
Location Attributes-Top 3 of 10
42
Job Attributes-Top 3 of 10
43
Training and Education Attributes-Top 1 of 3
44
Incentive Attributes-Top 1 of 3
45
Enabling Technologies

• Algorithms
• 2-sided matching
• Optimization
• Information Technology
• Intelligent Software Agents
• Expert Knowledge based systems
• Navy Marine Corps Intranet
• Existing legacy systems (JASS, EDPROJ, EPRES)
• Technology - Tried and Tested

46
Screen Score Sailors
Screen sailors on Must Have Attributes
Score eligible Sailors on Should Have Attributes
Rank sailors based on Scores
SaBiSS
NERISSA MODULE
Sailor and Billet Scoring System
47
Sailors List Preferences
Sailors view their billet eligibility and ranking
in JASS
Make appointment to see CCC
See CCC
Enter Preference list into JASS
Maintain Human Touch
CKBS
Career Knowledge Based System
NERISSA MODULES
JASS
Job Advertising and Selection System
CCC - Command Career Counselor
48
Manage Exceptions
AFTER 3 CYCLES
Unmatched sailors and billets
Match unmatched sailors to billets using
Optimization
Last resort/ exceptions Manual matching
NERISSA MODULE
SaBMaM
Sailor and Billet Matching Module
49
Audit and Write Orders
Orders Written and Sailors informed
Summary report complied
EPMAC Audits Assignments
NERISSA MODULE
ACOM
Assignment Control Module
50
Handling Exceptions

• Tied Movers
• Exceptional Family Member Program
• Sailors who do not state their preferences
• Sailor Priority Programs
• GUARDS III
• TWILIGHT
• SWAPS

51
Completed Theses

• Short, Melissa M., Lt., USN, Analysis of the
  Current Navy Enlisted Detailing Process, December
  2000.
• Schlegel, Richard J., LCDR, USN, An Activity
  Based Costing Analysis of the Navys Enlisted
  Detailing Process, December 2000.
• Wasmund, Todd R., Captain, U.S. Army, Analysis Of
  The U.S. Army Assignment ProcessImproving
  Effectiveness And Efficiency, June 2001.
• Hill, Kim D., LCDR, USN, An Organizational
  Analysis Of The United States Air Force Personnel
  Center (AFPC) Airman Assignment Management System
  (AMS), March 2001.
• Robards, Paul A., Captain, Australian Regular
  Army, Applying Two-Sided Matching Processes To
  The United States Navy Enlisted Assignment
  Process, March 2001.
• Tan, Suan Jow, Major, Republic of Singapore Navy
  and Major Che Meng Yeong, Republic of Singapore
  Air Force, Designing Economics Experiments To
  Demonstrate The Advantages Of An Electronic
  Employment Market In A Large Military
  Organization, March 2001.
• Ng, Hock Sing, Major, Singapore Armed Forces and
  Major Cheow Guan Soh, Singapore Armed Forces,
  Agent-Based Simulation System A Demonstration
  Of The Advantages Of An Electronic Employment
  Market In A Large Military Organization, March
  2001.

52
Theses in Progress

• LT Virginia Butler, NC, USN and LCDR Valerie
  Molina, NC, USN, Command and Sailor Preferences
  in a Two-Sided Matching Distribution Process.
• Major Gerard Koh, Army, Singapore, A Redesign of
  the Navys Enlisted Personnel Distribution
  Process.