Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Modeling effects-based operations in support of war games

Lee W. Wagenhals, Alexander H. Levis*


C3I Center, George Mason University

ABSTRACT

The problem of planning, executing and assessing Effects-Based Operations (EBO) requires the synthesis of a number of
modeling approaches. A prototype system to assist in developing Courses of Action (COAs) for Effects-Based operations and
evaluating them in terms of the probability of achieving the desired effects has been developed and is called CAESAR II/EB.
Two of the key components of the system are: (a) an Influence net modeler such as the Campaign Assessment Tool (CAT)
developed at AFRL/IF, and (b) an executable model generator and simulator based on the software implementation of
Colored Petri nets called Design/CPN. The executable model, named COA/EB, is used to simulate the COAs and collect data
on Measures of Performance (MOPs). One particular output is the probability of achieving the desired affect as a function of
time. Probability profiles can be compared to determine the more effective COAs. This version of CAESAR II/EB was used
successfully in August 2000 at the Naval War College in the war game Global 2000. Experiences with building and using the
models both prior to the war game and during the war game to answer topical questions as they arose are described.

Keywords: Effects-Based Operations, Influence nets, Colored Petri nets

1. INTRODUCTION

Given the potential complexity of future situations, as evidenced by the operations in Bosnia and Kosovo, and the many
consequences of the responses, an approach was needed that (a) related conventional and information operations to events
and events to effects; (b) allowed for the critical time phasing of actions for maximum effect, and (c) provided in a timely
manner the ability to carry out in near real time trade-off analyses of alternative Courses of Action (COAs). A prototype
system to assist in developing Courses of Action for Effects-Based Operations (EBO) and evaluating them with respect to the
effects they are expected to achieve has been developed and is called CAESAR II/EB1.
The process embodied in CAESAR II/EB consists of four steps. There is an additional step that must occur before the use of
CAESAR II/EB. This is the determination of the desired effects: the goals are set by the National Command Authority at the
strategic level and by the Commander through the Commanders Intent for the operational level. To reach the goals, certain
effects must be achieved. The step is shown
schematically in Figure 1.
COMMANDERS In the first step of CAESAR II/EB, intelligence
INTENT analysts, often in collaboration with planners
and operational staff, carry out Situation
Analysis. This step produces an Influence net
Dire ction from
na tiona l le ve l EFFECTS model that is key to planning for effects based
S OURCE &
BACKGROUND
operations. An application such as CAT** or
MATERIAL SIAM*** allows the intelligence analyst to build
complex models of probabilistic influences
ANALYS T
between causes and effects and effects and
KNOWLEDGE actionable events, as shown in Fig. 2. the next
& JUDGMENTS
Figure, 3, shows that the actionable events can
INTEL consist of both conventional, special, and
Figure 1: Establishment of desired effects
information operations. This graphic which also
implies the existence of a library of models that
can be used as modules to create new influence models that are appropriate for the specific situation.

*
alevis@gmu.edu; phone 1 703 993 1619; fax 1 703 993 1708; http://viking.gmu.edu; George Mason University, C3I Center, MSN 4B5, Fairfax, VA
22030, USA.
**
CAT is the Effects Based Campaign Planning and Assessment Tool under development at AFRL/IF and is used as a module in the CAESAR II/EB.
***
SIAM is a COTS product developed by SAIC to support the intelligence community. It is used as an alternative module in CAESAR II/EB.
In the next step, the Influence net model is used to carry out sensitivity analyses to determine which actionable events, alone
and in combination, appear to produce the desired effects. It should be noted that Influence nets are static probabilistic
models; they do not take into account temporal aspects in relating causes and effects. However, they serve an effective role in
relating actions to events and in winnowing out the large number of possible combinations. The result of this step is the
determination of a number of actionable events that appear to produce the desired effects and an estimate of the extent to
which the goal can be achieved.
INFLUENCE NET COMPUTATIONAL APPROACHES
INTERFACE

CAT Effects
Engine OPS Based
EFFECTS
Model

SIAM
Engine

IOPS
INFLUENCE
INTEL NET MODEL

Figure 2: An intelligence analyst developing an Influence net Figure 3: The Effects Based model includes both
model relating actionable events to effects. The network implies conventional and information operations as actionable events
access to diverse data sources through any network-centric
architecture

Once the Influence net of the situation has been developed, the situation analyst converts it into an executable model in the
form of a Colored Petri net that allows the introduction of temporal aspects (Fig. 4). An automatic algorithm that performs
this conversion has been developed, tested, and demonstrated.2 The Influence nets used for situation assessment contain a
great deal of information in the form of beliefs about the relationships between events and the ultimate outcome or effect.
While they have an underlying rigorous mathematical model that supports analysis, they provide only a single probability
value for a given set of actionable events. They do not capture the effect of the sequence or timing of the actionable events.
Additional information needs to be inserted to account for temporal and logical sequencing of actionable events. A particular
sequence of actionable events represents an alternative Course of Action. Note that in a threat environment proper sequencing
is critical; reversal of two operations can endanger lives and or negate the impact of an information operation. Such reversals
are not easily observed in a complex scenario with many concurrent tasks. The executable model brings these issues to the
fore.

Actionable
Events via COA EXECUTABLE MODEL
Sensitivity Development
Analysis
Effects
Based
Model

Automatic
conversion to
executable model

INFLUENCE
NET MODEL
EXISTING
OPS MODELS
Figure 4: Development of the Executable Colored Petri Net model
The executable model, when properly initialized with a scenario, is then used in simulation mode to determine their
effectiveness of the candidate COAs by generating the timed probability profile for each one. A Commander can then make
an informed choice and direct the planning staff to prepare the detailed plan for the chosen COA.
Carrying out simulations using the executable
Probability model is not the only way in which COA
[0 0 0 0 0] ; 0.0899

a1^c
b1 a1 b1^c
a1^b1 c
a1^b1^c

b2
[0 0 0 1 0] ; 0.0158

a1
b2^c
c
a1^c
[0 1 0 0 0] ; 0.1437
b1
a2 b1^c
a2^b1^c a2^c
c
[1 0 0 0 0] ; 0.4406

b1
a1^b1
a1
1 analysis and evaluation can be conducted. State
0.8
a1^b2 a2^b1

[0 0 0 1 1] ; 0.0158
c
[0 1 0 1 0] ; 0.1021
b2
a2 b2^c
c
[0 1 1 0 0] ; 0.1864
b1
b1^c
c
[1 0 0 1 0] ; 0.3630
b2
a1
[1 1 0 0 0] ; 0.4712
b1 a2 COA 1 Space Analysis of the Colored Petri net model
a1 a1^c a2^b1
0.6
derived from the Influence net can be conducted
a2^b2 a2^c

[0 1 0 1 1] ; 0.1021 [0 1 1 1 0] ; 0.1970 [1 0 0 1 1] ; 0.3981 [1 1 0 1 0] ; 0.1970 [1 1 1 0 0] ; 0.7524

0.4
c b2
a2 c a1^b2 b2 a2
a2^c b2^c a2^b2^c a1 a2^b2
COA 1
to reveal all of the probability sequences that can
b1

[0 1 1 1 1] ; 0.1970

0.2
[1 1 0 1 1] ; 0.2066 [1 1 1 1 0] ; 0.8227
c a2 b2

[1 1 1 1 1] ; 0.8860
0 1 3 5
be generated by any timed sequence of
0 2 4
Step actionable events. The result of the state space
State Transition Diagram Untimed Probability Profiles analysis is a State Transition Diagram that is
mathematically a lattice, as shown in the upper
left hand corner of Figure 5. This state transition
Analysts 1 COA 1 diagram can be easily converted to a plot
0.8 showing the range of probability values that can
0.6
exist at each step in any probability profile. This
COA 2 technique allows the analysts to see, at a glance,
0.4
C B COA 1 all of the potential effects that timing of the
A
t
0.2 actionable events can have. The analyst can then
C A B COA 2 0 select the profiles that gives the best results and
t 0 10 20 30
Candidate COAs Time presents them to the Commander for selection.
Timed Probability Profile

Figure 5: Decision support for COA selection

2. USING CAESAR II/EB IN WAR GAMES

Until recently, the concept for model-based development and evaluation of Courses of Action (COA) and the CAESAR II/EB
tool suite had been demonstrated using realistic scenarios and data, but had not been tested in a dynamic operational
environment. While realistic models have been created to test the concepts, the use of the tool suite within a working
command and control structure had only been postulated.
In early 2000, the Naval War College invited the Office of Naval Research to use the CAESAR II/EB tool suite in its
capstone Title 10 war game, Global 2000, to gain insight into its potential utility for supporting COA development and
evaluation. This provided an important opportunity to test these concepts and tools in a realistic environment. At one level,
this participation could help determine if this approach to Course of Action development and selection for effects based
operations can be used to support war games. Even more importantly, the participation in the war game would provide
insight into how these concepts could be incorporated in real-world operational environments. Participation in the game
would provide insights about the sources of the information needed to create the models, the type of expertise that is needed
to build and analyze the models, and the types of collaboration and dialog that need to occur between modelers, intelligence
analysts, operational planners, and the commander and command staff. This section describes how the tool was used and the
lessons learned from the experience.
To present these results, the scenario used in Global has been significantly morphed to remove sensitivities associated with
the actual war game scenario. We begin by briefly describing the modified scenario, followed by the process used to analyze
the situation to determine the desired effects. The next step is the development of the Influence nets. The sensitivity analysis
using the Influence net is then illustrated and the conversion of the Influence net to the executable model is described. The
generation and use of the probability profiles for COA selection is discussed. Several valuable insights and lessons were
noted from the use of CAESAR II/EB in the game.
This war game concerns an area of operations (AOR) composed of several nations or entities. One of the nations (the Nation
of Borg) is technologically advanced and has become increasingly willing to use its military instrument of national power,
primarily through posturing, to resolve its national security concerns. Borg shares a border with a smaller nation (Alpha)
over which there have been historic border disputes. The Borg has adopted a strategy of conducting military exercises near
the disputed border with Alpha. While these exercises have always terminated peacefully, they provide the potential for the
build up of military forces that could be used for a quick-strike invasion of Alpha. The Blue nation is technologically
advanced and is a member of a coalition of partners called the Federation. Blue is geographically separated from the AOR.
An ally of Alpha, Blue has a diplomatic and military presence in Alpha. Blue has active trade relationships with all nations in
the AOR.
The first step in using CAESAR II/EB for effects-based operations is to evaluate the situation. One of the outcomes of this
evaluation is the determination of the key effects or decisions that will be the focus of the Influence net models that will
relate causes (actions) to those key effects. As was illustrated in Figure 1, several inputs are needed for this analysis. These
inputs include Blues estimate of the Borg intent, the guidance from the Federation National Command Authority, and the
Guidance and Intent of the Federation Commander who will be responsible for conducting operations in the AOR.
In view of the Borg exercises, the situation analysts postulate that the Borg intent is to disestablish Blue and Alphas treaty
obligations, causing Alpha to ask Blue forces to leave Alpha. The Borg may launch an invasion of Alpha from their exercise.
Once established inside Alpha territory, the Borg hope to convince Alpha to sever alliances with Blue. The Borg hope their
actions are not severe enough to hurt trade and diplomatic relationship with Blue and the Federation in the long term. The
Borg possesses considerable military capability including dangerous Weapons of Mass Destruction (WMD). These consist
of the Death Star (DS) Weapon and several Assimilation Weapons (AW). Borg use of these weapons has been held in check
in part because Blue also possesses WMDs.
In light of these developments, the Federation establishes a Federation Task Force composed of Federation forces.
Unfortunately, while the Task Force is being established, the Borg begin their invasion of Alpha territory. The National
Command Authority of Blue issues guidance to the Blue Commander to (1) Get the Borg out of Alpha territory (2) Re-
establish existing boundaries (3) Reduce Borgs inclination for military coercion (4) Maintain Federation access to the AOR
and (5) Ensure that actions are consistent with maintaining stability in Federation relationships with Borg. The commander
of the Federation Task Force issues the following Commanders Intent:
Purpose: Prevent Borg from continuing the invasion of Alpha and assimilating inhabitants.
Take action to get the Borg out of Alpha territory

Methods: Deploy a grid of sensors to locate all forces


Prevent Borg from using WMD
Conduct Counteroffensive operations to cause Borg to retreat from Alpha

End State: End armed conflict in Alpha AOR


Establish boundaries as they were before Borg invasion
Ensure Alpha sovereignty
Alpha to continue to be open to Federation Presence
Enable conditions for Borg and Federation to cooperate
Ensure Freedom of navigation and movement in the entire AOR

Given this Commanders Intent and the perceived intent of the adversary, two key decisions that may be made by the
adversary are postulated as the focus of the Influence net. The first is a decision by the Borg to stop the invasion and the
second is a decision by the Borg to enter into negotiations. Because of concern about the Borgs possible use of the WMD, a
third key decision by the Borg to use the Death Star is also considered for the Influence net model.
Having settled on the key decisions to focus on with the IInfluence net, the analyst turns to the operational planners who are
establishing the actions that will comprise Courses of Action (COA) that are commensurate with the NCA guidance and the
Commanders intent. The planners are contemplating an Interceptive Disarmament operation to halt the Borg offensive and
protect Federation Forces in the AOR. At the same time, information operations, including potential political announcements
by the President of the Blue nation are being considered.
To build the Influence net, the situation analysts use a combination of top down and bottom up approaches. With the top
down approach, the analysts estimate what events or beliefs the Borg would consider in making each of the three decisions.
For example, in deciding whether to agree to negotiations, the Borg will attempt consider if its military operations can be
sustained and whether its national interests will be compromised. The Borg may consider the use of the Death Star WMD to
coerce the Federation to negotiate on Borg terms, however the Borg also considers that the Federation could use its own
WMD if the Borg uses its WMD. The analyst reviews each of these concerns and considers the major factors or concerns
that the Borg will consider in their deliberations. This analysis continues until concerns in the form of events, beliefs, or
decisions are devised that can be affected by actions taken by the Federation. Notice that this analysis is done from the point
of view of the adversary. A model is being created that represents Blues understanding of how the adversary thinks and
what he considers to be important.
To employ the bottom up approach, the analyst considers the actions that are being contemplated by the Federation
operational planners and the effect those actions could have on the Borgs thinking. For example, the interceptive disarming
operation could affect the Borgs belief that the Borgs invasion will have taught Alpha a lesson. It can also effect the
perception that the Borg regime will survive. The objective of the analyst is to ultimately connect the actions that are either
determined from the bottom up approach or the actions suggested from the top down approach with the key decision. In
doing this, the analyst uses knowledge of the adversary to analyze and specify interconnections between target system/centers
of gravity to determine indirect effects of potential actions. When so constructed, the Influence net can indicate the impact
that actions that potentially comprise a COA, can have on each of the three key decisions through chains of cascading effects.
An Influence net model for this situation is shown in Figure 6. The model was built using the Campaign Assessment Tool
(CAT). Each node represents an action, event, belief, or decision. A declarative sentence in the form of a proposition is used
to express the meaning of each node. The directed arcs between two nodes mean that there is an influencing or causal
relation between those nodes. The truth or falsity of the parent node can affect the truth or falsity of the child node. The
Influence net has been arranged with actions on the left and the key decisions on the right. This is to indicate visually that the
effects of the actions are expected to propagate to intermediate effects over time until their impact reaches the key decisions.
The visual construct is that there is a time scale associated with the propagation of effects between nodes of the Influence net
that moves from left to right. There are six actionable events on the left side of the Influence net as shown in Table 1. These
are candidate actions (or results of actions) that can comprise a COA that can impact the three Borg decisions of interest.

Lead To
Borg Decisions
Blue
Actions

Figure 6: Influence Net in CAT


Table 1. Actionable Events in the Influence Net
Type of Actionable Event Actionable Event
Force Operation Federation Interceptive Disarming (operation) has succeeded
Show-of-Force Operation Federation Forces are in position to strike Borg Death Star
Federation NCA offers terms for negotiation
Information Operation Federation NCA speech offers Federation Cooperation if Borg withdraws
Federation pledges not to use WMD first
Potential target for Federation convinces Borg that it has not taken steps to full war (despite
Information Operation Interceptive Disarming Operation)

In developing the Influence net, the modelers must incorporate two types of knowledge about the Borg. The first involves
the actions, events, beliefs, and decisions and the relationships between them. This knowledge is captured in the structure of
the influence. The second type of knowledge is estimates of the strength of the influences represented by the arcs. There
are several ways that this strength factor can be incorporated. One of the more analyst-friendly ways is based on an
algorithm implementing the Casual Strength (CAST) logic developed by George Mason University. This algorithm has been
implemented both in SIAM and CAT. The algorithm requires the specification of two parameter values for each arc in the
Influence net plus a single baseline probability for each node in the Influence net. The algorithm uses these values to
generate a complete conditional probability specification for each node in the net. The result is an approximation of
conditional probability values needed to perform the probability propagation to evaluate the effects of the actions in the
Influence net.
Once the Influence net has been completed, it can be used to evaluate the impact of actions on the effects (decisions) of
interest. This can be accomplished in several ways. The first, and simplest method, is by setting the probabilities of the
actionable events to either zero or one, depending on whether the action is planned or not, and evaluating the Influence net.
Algorithmically, this means that the tool propagates these probabilities until all effects are accounted for at the nodes with no
parents. These nodes are the key decision nodes. In the variant of CAT used to create the example model, the results of this
evaluation are visually shown by the color of each node and by providing marginal probability values of each node in a small
circle on the lower left corner of the node. In the color scheme, dark red means that the declaration in the node is false, dark
blue means the declaration is true, gray means the probability of the declaration is true is near 0.5. Lighter shades of red
mean that the probability of the declaration is between zero and 0.5 and lighter shades of blue mean that the same probability
is between 0.5 and 1.0. This colorization helps with the visualization of the complete impact of a COA that includes the
actions that have been given a probability of one. An analyst can experiment with the Influence net by changing the
probabilities of one or more of the actionable events and seeing what the effect is on the key decision nodes.
To assist in analysis of the effect of the actionable events, both CAT and SIAM have built in Sensitivity Analysis routines.
After selecting a node, usually the key decision nodes, the routine will indicate the effect each actionable event will have on
the selected node, if the probability of the actionable event is set to either zero or one. This analysis can indicate which
actionable events increase or promote the likelihood of the key decisions and which actionable events decrease or inhibit that
likelihood. It also indicates the magnitude of the affect each actionable event has by itself on the situation. The Sensitivity
Analysis output for the first key decision node of the Influence Net of Figure 6 is shown in Table 2.
The Sensitivity Analysis Table indicates that the current probability of the decision node Borg decides to negotiate end of
conflict is 0.48. This is based on all of the actionable events not occurring (Marginal Probability = 0.0). The second
numerical column of the table indicates the probability of the decision node if the probability of each actionable event is zero.
In this case, it is the same as the current marginal probability node of 0.48. The third numerical column shows the marginal
probability if each actionable event occurred by itself (and all the others remained unchanged). This column shows that four
of the six actionable events increase the likelihood of decision and two of the actionable events decrease its likelihood.
Furthermore, the first and last actionable events can cause the greatest increases while the Interceptive Disarming (ID)
Operation causes the greatest decrease. The same conclusions can be drawn from the fourth numerical column of the table
that contains the sensitivity coefficients. Positive coefficients mean that the action promotes the decision and negative
coefficients means that the action inhibits the decision. Comparing the magnitude of the coefficients can indicate which
actions have the greatest promoting or inhibiting effect.
Table 2. Sensitivity Analysis Output from CAT
Sensitivity Analysis for Node: Borg decides to negotiate end of conflict
---------------------------------------------------------------------------------------------------------------------
Probability: 0.482921
---------------------------------------------------------------------------------------------------------------------
NAME Marg. Prob. Prob. = 0 Prob. = 1 Sens. Coef.
---------------------------------------------------------------------------------------------------------------------
Fed convinces Borg that it has not 0.0000 0.4829 0.6537 0.0016
Fed ID has succeeded 0.0000 0.4829 0.3538 -0.0011
Fed NCA offers terms for negotiation 0.0000 0.4829 0.5283 0.0005
FED NCA speech offers cooperation 0.0000 0.4829 0.5305 0.0005
Fed NCA pledges not to use DS first 0.0000 0.4829 0.4573 -0.0003
Fed forces are in position to strike 0.0000 0.4829 0.5745 0.0010

In an Influence net with multiple nodes with no children (multiple key decision nodes), sensitivity analysis needs to be
conducted for each such node. This is done to determine if actions that affect one key decision in the desired way have an
adverse affect on other key decisions. For example, convincing the Borg that the Federation has not taken steps to full war
may decrease the likelihood of using the WMD weapons even though it decreases the likelihood of negotiations. If such
conflicts are detected, they can be examined to determine whether or not the action is a candidate for the COA.
The Influence net models can focus attention on conflicts in another way. In this example, the sensitivity analysis shows that
the Interceptive Disarming Operation has a negative affect on all three key decisions. This can bring into question the value
of this aspect of the COA. Recall, that the operational planning staff prefers the Interceptive Disarming Operation. Their
analysis is from the operators perspective and shows that the operation is necessary to protect forces and assure access to the
AOR. This is a legitimate argument for the operation. The Influence net points out that his operation could have unintended
consequences. The Influence net also indicates other actions that can be taken to mitigate these consequences.
Once the analysis of the Influence net has been completed and the actionable events for the COA have been selected, the
operational planners assess the availability of resources to carry out the tasks that will result in the occurrence of the
actionable events. The resultant plan will indicate when each actionable event will occur. The next step in the COA
evaluation process is to convert the Influence net to an executable model so that a temporal analysis of the COA can be
performed. Using the executable model, the analyst is able to generate the probability profiles that show the marginal
probability for any node in the net as a function of time. These profiles can indicate how long it will take for the effects of
the actionable events to affect various nodes in the Influence net. The analyst will most likely concentrate on the probability
profiles of the key decision nodes, the nodes with no children.
To create the executable model, the analyst has to add two more types to knowledge to the model. The analysts ha to indicate
the time delay that will exist between each node in the Influence net and that nodes children. This time delay represents
processing and communication time delays that slow the propagation of knowledge about events, actions, beliefs, or
decisions that are represented by a parent node to the child nodes. The second type of knowledge that is needed is the timing
information of the actionable events. This information will initially come from the operational planners.
With the CAESAR II/EB tool, the conversion to the executable model is an automated process. In CAESAR II/EB, the
executable model is a Colored Petri (CP) Net, which is a generalized form of a Discrete Event System Model. In CAESAR
II/EB, the CP net is hidden from the user. Instead of working with the CP net, the analyst interacts with the model through a
web browser interface such as Netscape or Microsoft Explorer. To create this executable, the CAT tool automatically exports
a special file that contains all the information needed to generate the CP net. This file is placed in a special folder on the web
site that hosts the CAESAR II/EB tool. The analyst then uses a browser to access the web site and create the CP net.
The home page of the web site is shown in Figure 7. The analyst initiates the conversion of the Influence net to the CP net by
typing the name of the net in the window of the home page and clicking on the Make Delay File button. The web site will
automatically bring up the page shown in Figure 8. To create the CP net from the Influence net, the analyst types in the time
delay information into the appropriate boxes in the browser form shown in Figure 8. The analyst also indicates nodes in the
Influence net for which probability profiles will be generated. Once the form is completed, the analyst clicks on the run
button, and in a few seconds, the executable model is generated and automatically placed in a folder in the CAESAR II/EB
web site so it can be used to perform temporal analysis of the COAs.

Figure 7: Home Page of the CAESAR II/EB Web Site


Having created the executable model, the analyst can generate the probability profiles for any COA composed of the
actionable events and the time each is expected to occur. To do this, the analyst fills in a COA form on the web page found
at the CAESAR II/EB web site as shown in Figure 9. The form automatically lists all of the actionable events and provides
boxes where the probability of the actionable event and the time of the probability can be entered. Note that the executable
model is initialized with all actionable events set to a probability of zero to represent the condition that none of the actionable
events has (yet) occurred. The analyst also checks the boxes that indicate the nodes for which probability profiles will be
generated. After filling out the COA form, the analyst clicks the run button and in a few seconds the probability profiles are
generated and displayed.
In the particular example of the Federation versus the Borg war game, the operational planners provided initial timing
information for the actionable events. The COA specified that the Interceptive Disarming Operation would occur
successfully at the beginning of the campaign (D Day). It is set to time zero. This action is synchronized with a pledge by
the Federation NCA not to use WMD. An aggressive Information Operations plan will be executed to convince the Borg
that the Federation has not taken steps to full war. The goal is to convince the Borg by Day 2. On Day 1, the NCA will give
a follow up speech offering cooperation with the Borg if they will terminate hostilities and withdraw. One Day 2 the NCA
will issue the terms for a negotiation with the Borg. By this time, the Federation will maneuver its forces within striking
range of the WMD capability of the Borg as well as its other military forces. These times are used to fill out the COA form.
The probability profiles shown in Figure 10 were generated for the COA proposed by the planners. The annotations have
been added to indicate the three separate probability profiles.
Figure 10 is a composite plot of three probability profiles generated by CAESAR II/EB. It shows the marginal probability of
each of the three key decision notes as a function of time. The vertical axis is probability value and ranges between 0.0 and
0.8. Time is on the horizontal axis and ranges from 0 to 6. The time scale is in days. A review of the composite probability
profiles indicates that the COA is probably acceptable in terms of the three key Borg decisions. Collectively the actionable
events nudge the Borg toward negotiation while dissuading them from using their WMD. The likelihood of negotiation
begins to decrease slightly at Day 3 but begins to increase at Day 5 and reaches its highest value at Day 8. The propensity to
use WMD starts with a probability of 0.37, decreases to 0.2 at Day 3 and increases slightly before falling to close to zero by
Day 5 as the combination of influences convinces the Borg not to use WMD. The major driver is the positioning of the
Federation force within striking range of these weapons. The plot also shows the Borg are likely to continue hostilities right
up to the time that they decide to negotiate. Indeed, while COA moves the Borg in the desired direction on two of the key
decisions, these actions also increase the Borg desire to continue to fight.

Time
Delays

Figure 8: CAESAR II/EB Web Page Form to Specify Time Delay Information

3. OBSERVATIONS FROM WARGAME PARTICIPATION

Participation in Global 2000 provided an invaluable opportunity to test the theories, tools, and techniques in an operationally
realistic environment. This section provides observations and lessons that were learned from that experience. The section
begins with the objectives for the war game. An assessment of whether the objectives were achieved is provided followed by
discoveries and findings from the game.
Setting time of
Actionable
Events

Selecting
Nodes for
Probability
Profile Plots

Figure 9: CAESAR II/EB COA Generation Form


There were four objectives for the use of CAESAR in Global 2000:
Determine if CAESAR II/EB has the potential to enhance the assessment and decision making process in Military
Operations
Gain insight into how the tool is used in a game cell
Identify changes that should be made to CAESAR II/EB to enhance its usefulness
Contribute to Blue decision making by providing critical insight into the effects of actions and their timing on
desired outcomes as well as unintended and undesirable consequences.
The first three objectives were achieved. CAESAR II/EB appears to have utility in support of model-driven COA
development and evaluation. Sufficient insight was gained to create an operational concept for the future use of CAESAR
II/EB and to develop an operational architecture view. As a result of the extensive modeling, before and during the game,
several fixes and enhancements both to the tool and the concepts were identified. The fourth goal was partially achieved.
Insight into the key issues of the use of WMD as an effect of several Blue actions was highlighted through the CAESAR
modeling effort. Faster model turn around time was needed to reach full potential.

Borg decides to
terminate hostilities

Borg decides to
negotiate

Borg decides to
use DS

Figure 10. CAESAR II/EB Generated Composite Probability Profiles


One of the first decisions that had to be made at the war game was where the CAESAR II/EB tool suite and the staff that
operate it should be located. Initially, it was thought that the best location was in the command cell of the Federation Joint
Task Force. However, the organizational structure of Global 2000 had a special entity called Blues Borg Cell. This cell was
separate from the standard intelligence cells and had the function of providing quick reaction analysis of the adversarys
potential actions and reactions to the Blue COA. This cell reported directly to the command cell. Because the cell was
populated with numerous subject matter experts (SMEs) from intelligence, information operations, WMD effects, and nodal
analysis, it was decided within the first hour of the game that this was the most appropriate location for CAESAR II/EB.
As the game progressed, the operational concept for the use of CAESAR II/EB evolved to the four major steps: creating the
Influence net, converting the Influence net to the discrete event model, evaluating COAs, and preparing explanations of the
results and findings. The first step, creating the Influence net requires the assistance of SMEs and is the most challenging
step in the process. Accomplishing this step provides immediate assistance to the Blue team by helping structure the
conceptualization of the adversarys potential reaction to contemplated Blue actions. While this step does not address any
timing issues, it does assist analysts in formulating their analysis and answers to what if questions presented by Blue.
Three models were created with CAESAR II/EB, one before the game and two during the game, and each contributed to the
overall output of Blues Borg Cell via this mechanism.
The probability profiles provided by CEASAR II/EB help highlight the timing aspects of Blue actions and potential Borg
responses. Even though the third model was not available until late in the game, the questions it answered were still relevant
to Blues Borg Cell and the decision makers the cell supported.
4. RESULTS AND CONCLUSIONS

CAESAR II/EB embodies a non-traditional concept for supporting Effects Based Planning and Operations. One of the key
findings from the Global experience is that its use requires the concerted effort of specialist from different disciplines:
intelligence, operational planning (current and future), information operations, and command. At Global 2000, CAESAR
II/EB was used in support of Blues Borg Cell because this cells mission was most closely aligned with CAESARs
capabilities. The composition of Blues Borg Cell provided a rich combination of SMEs that assisted the CAESAR II/EB
developer/operator in model construction and analysis. The developer/operator was able to take textual documents produced
by the SMEs and convert them to Influence nets. The SMEs then reviewed the Influence net and made corrections as
necessary to validate the model.
To be effective, SMEs from both intelligence and operations must be available and work together to build and evaluate the
right kinds of model. This was highlighted in the game. A better job could have been done in determining the actions that
should be incorporated in the Influence net models. The domain experts (SMEs) about the adversary assisted in the creation
of a major portion of the models, but the set of optional actions that can be taken to provide the influencing stimuli to the
adversary needed to be provided by military operators including the information operations specialist. This type of expertise
and information was available during Global, but the SMEs with the operational expertise were not located in the Blues
Borg Cell for direct consultation and dialog. The limited number of people on the CAESAR team prevented adequate
interaction with operational planners to extract that information.
Better methods for presenting the results and recommendations from the use of the CAESAR II/EB tool need to be
developed. Presenting probability profiles to command and staff who are not familiar with the CAESAR II/EB or the models
does not convey the information needed to make informed decisions. Appropriate visualization is needed to show decision
makers in an intuitive way the results and the reasons for them. A one-page narrative format was used at Global and was
posted on the appropriate web page.
The authors believe that the contribution of CAESAR II/EB can be significantly enhanced in war games if it is used during
the development of the campaign plan. It is during campaign planning that the basic COAs are developed. Models
developed using CAESAR II/EB can provide the rationale to the staff and the commander for the actions in the COA. But
there is a more important reason for the early use of the tool. During game execution, it is not very likely that the game
players will use recommendations derived from the analysis of the tool if they are not familiar with the tool, the models, or
the format of presentation of the results. By introducing these concepts to the command and staff early and obtaining their
comments and buy-in, they will be able to quickly assimilate recommendations derived from the tool during the game. .
An approach to Course of Action development and selection for effects based operations that integrates Information
Operations with conventional operations has been presented and CAESAR II/EB, a decision support tool prototype, has been
described and an example has been used to illustrate the operation. By participating in the Global 2000 war game, the
CAESAR II/EB team has gathered important information that will be used to develop an operational concept for the use of
CAESAR II/EB and support effects based operations.

ACKNOWLEDGEMENTS
This work was supported by the Office of Naval Research under grant no. N00014-00-1-0267.

REFERENCES
1. A. H. Levis, Course of Action development for information operations, Phalanx, The Bulletin of Military Operations
Research, Vol. 33, No. 4, December 2000.
2. L. W. Wagenhals, I. Shin, and A. H. Levis, Creating executable models of Influence nets with Coloured Petri nets, Int.
J. STTT, Springer-Verlag, Vol. 1998, No. 2, pp. 168-181.

You might also like