Trust evaluation: an AHP and multi-objective
programming approach
Keywords
Trust, Evaluation,
Analytical hierarchy process,
Programming, Goals,
Project management
Abstract
Establishing trust within groups
or teams of scientists and
technicians working on large-scale
projects is considered an essential
critical success factor when
working in a virtual environment.
To build trust among team
members for any project requires a
selection decision on which trust
factors will best achieve the goals
of the project. This paper presents
application results of using two
analytic methodologies used to
evaluate trust survey information
from the Spallation Neutron
Source (SNS), the USA's largest
science project. The two decision
methodologies produced results
that helped support SNS
management's judgment that
uniquely selected trust categories
should be emphasized to build
trust on this project.
Management Decision
41/6 [2003] 587-595
# MCB UP Limited
[ISSN 0025-1747]
[DOI 10.1108/00251740310484867]
Larry L. Radcliffe
Spallation Neutron Source Project Office (LM-14), Oak Ridge Operations Office,
US Department of Energy, Oak Ridge, Tennessee, USA
Marc J. Schniederjans
Department of Management, College of Business Administration, University of
Nebraska-Lincoln, Lincoln, Nebraska, USA
Introduction
Funded by the US Department of Energy, the
Spallation Neutron Source (SNS) is a $1.4
billion construction project that will be
operational in June 2006. SNS is the country's
largest science project and next generation,
world-class neutron-scattering facility. The
SNS construction project is using a ``virtual
project team'' comprising six of the US
Department of Energy's national
laboratories. A virtual team is defined as
``a self-managed knowledge worker team,
with distributed expertise, that forms and
disbands to address a specific organizational
goal'' (Kristof et al., 1995).
The issue of building trust in a scientific
virtual project team is important because of
the unique opportunity the SNS virtual team
presents and because the next generation of
``big science'' projects will require global
virtual project teams. Glanz (2001) reported,
``The directors of major physics laboratories
in Europe, the USA and Japan gathered this
week to make plans for a new particle
accelerator they all agreed would be so large,
powerful and expensive that it could be built
only if they all cooperated on a scale without
parallel in scientific history.''
Trust in a co-located team
Trust in a co-located team is usually built on
impersonal and interpersonal factors.
Impersonal factors are linked to institutional
norms and beliefs (e.g. organizational
policies). Interpersonal factors relate to
expectations of the working group (e.g.
management-team relationships and team
communications). Both the impersonal and
interpersonal factors of trust assume that the
relationship between co-located team
members has some longevity. Consequently,
trust in co-located working groups is based
The Emerald Research Register for this journal is available at
http://www.emeraldinsight.com/researchregister
on the expectations of members' behavior
and also the collective norms of the working
group's institution (Jarvenpaa and Shaw,
1998).
Trust in a virtual team
As described above, the traditional model of
trust for a co-located team is built on long-
term relationships that a virtual team does
not have the luxury of time to establish.
Building trust in a virtual team has been
described as more a set of team leader (TL)
and team member (TM) actions that fall into
three categories (Duarte and Tennant-
Snyder, 1999):
1 performance and competency;
2 integrity; and
3 concern for others.
Performance and competency pertain to the
TLs' and TMs' track record of results. For
example, does the TL bring projects in on
time? Do the TLs or the TMs have the proper
experience and technical expertise to make
the project succeed? Do they honor their
commitments? Commitment follow-through
may be more important for virtual teams
than for co-located teams because virtual
teams have fewer events to judge whether a
TL or a TM are really committed to the
project's overall success (Duarte and
Tennant-Snyder, 1999).
Integrity is described as a set of principles
that makes a TL or TM dependable and
reliable (Jarvenpaa and Shaw, 1998).
Examples might include standing up for the
team, speaking positively about the team's
performance and maintaining consistent and
balanced communications with the TMs
(Duarte and Tennant-Snyder, 1999).
Once performance, competency and
integrity are in place, the care and concern
shown towards others by the TLs and TMs
become important (Hardt and Brynteson,
The current issue and full text archive of this journal is available at
http://www.emeraldinsight.com/0025-1747.htm
[ 587 ]
Larry L. Radcliffe and
Marc J. Schniederjans
Trust evaluation: an AHP and
multi-objective programming
approach
Management Decision
41/6 [2003] 587-595
[ 588 ]
1998). Concern for others relates to the fact
that we trust people who are consistently
responsive to our needs and to the needs of
others in the organization (Duarte and
Tennant-Snyder, 1999). Measures of this third
category might include the TLs' actions
related to effectively transitioning people on
or off the project team or their understanding
of the impact of team actions on people inside
and outside the team (Duarte and Tennant-
Snyder, 1999).
Swift trust
There is also a new theory on building trust
in high performing, co-located, and
temporary teams. Trust building on these
high performing temporary teams is focused
on achieving mission success, on having
clear goals and purpose and on meeting tight
deadlines. There is no time to develop trust
through more traditional means of
impersonal or interpersonal trust factors.
High levels of action rather than personal
relationships are what establishes trust in
these temporary teams. The concept is
referred to as swift trust (i.e. the ability to
quickly get down to business, focus on the
task, keep distractions to a minimum,
complete the job and then dissolve the team).
Meyerson et al. (1996) developed the concept
of swift trust to explain how temporary teams
can enjoy high levels of trust, even though
members do not share any past affiliation
and cannot necessarily expect to have any
future association (Jarvenpaa and Shaw,
1998). A major premise of swift trust is that
the time pressure of a team's project will
hinder the ability of TMs to socialize and
learn about one another's behavior, abilities
and goals.
A trust survey
One of the ways to measure trust in teams is
with trust surveys. Results from a trust
survey can indicate policies and practices
that should be maintained and identify areas
for improvement. Trust surveys can also be
used to identify which of the trust factors are
important on a project.
A survey of trust factors was developed
based on the principles in Duarte and Snyder
(1999) and Meyerson et al. (1996) to give SNS
management a gauge of the trust climate
within the project. The survey questions
were selected to capture measures on three
types of personnel, in three categories of
trust (i.e. co-located, virtual, swift), which are
further broken down into seven different
types of trust factors (representing the
decision variables in this study), as
presented in Table I.
Survey results were obtained from SNS
managers, technical and support staffs from
the six national laboratories involved in the
project. A total of 105 completed survey forms
were returned out of approximately 600
e-mail requests to SNS employees. The scale
used in the survey that measured trust on
each question ranged from low trust score
(a score of 1) to high trust score (a score of 5).
Average scores for all participants by type of
personnel and by trust factor are presented
in Table II.
The final survey results indicated that SNS
employees gave the project an overall
average score of 3.89 for all trust categories,
with approximately 3.9 for co-locate trust, 4.3
for virtual trust, and 3.5 for swift trust. The
overall survey results clearly indicated that
trust factors associated with virtual teams
were the most important.
Overview of the managerial
decision-making problem
Despite the strength of the survey scores for
the virtual trust factors, SNS management
knew the project was in transition. The
partner labs were completing their project
assignments and remaining project work
would soon be performed only at the SNS
home laboratory. SNS management wanted
to use the existing survey results to support
their belief that the trust factors related to
co-located trust should be emphasized over
virtual and swift trust factors at the project's
home site. However, management personnel
did not feel their overall preference for trust
factors were more important than the overall
preferences of the technical and support
personnel preferences.
Based on the above requirements, SNS
management wanted to use one or more
management decision-making methodologies
to determine on which trust factors they
should focus. Once the trust factors (which
are the decision variables in this study) were
identified, management planned to review
associated survey questions and employee
responses to identify to what specific
emphasis and action areas they could best
devote effort to improve trust for the project
as a whole. SNS management felt that by
using more than one selection methodology
they might better reveal solution patterns
that would help to validate their final
selection of trust factors.
Methodology
One of the more popular methods that allows
for the incorporation of judgmentally scaled
opinions into a ranking for selection decision
Larry L. Radcliffe and Table I
Marc J. Schniederjans Trust factors, categories and descriptions
Trust evaluation: an AHP and
multi-objective programming Trust factors
approach
(decision variables/
Management Decision
41/6 [2003] 587-595 alternatives) Trust categories Description
x1 Co-located team trust Organizational policies
x2 Co-located team trust Management and team relationships
x3 Co-located team trust Communications
x4 Virtual team trust Performance and competency
x5 Virtual team trust Integrity
x6 Virtual team trust Concern for others
x7 Swift trust Ability to get down to business and form trusts quickly

Table II
Average survey preference scores by type of personnel and trust factors
Co-located team trust Virtual team trust
Management Performance
Type of Organization and team and Concern Swift
personnel policies relationships Communications competency integrity for others trust
Management 3.833 4.038 3.906 4.042 3.821 3.670 3.439
Technical 3.801 3.953 3.687 4.238 3.990 3.453 3.545
Support 3.828 4.120 3.900 4.344 4.104 3.524 3.583
Total (column) 11.462 12.111 11.493 12.624 11.915 10.647 10.567

purposes is the analytical hierarchy process technicians). According to Bryson (1996),

(AHP), as described by Saaty and Vargas
(2001). AHP can be used to determine the
ranking of trust factors, and thereby provide
prioritized focus of effort of managers to aid in
trust building. AHP has also been combined
with the multi-objective programming
methodology of goal programming (GP) by
Schniederjans and Garvin (1997), Badri (1999),
and Kim et al. (1999) to permit individual
preferences to be explored. Combining GP and
AHP to determine which of the trust factors
were most important in light of personal
preferences represents a unique possibility in
this case study. To that end, AHP is presented
here first as a stand-alone methodology and
then a combined AHP and GP model is
presented as an extension to consider
additional criteria in the decision-making
process and to compare both solutions'
results.
AHP is a technique for considering data
about a decision in a systematic manner
(Saaty and Vargas, 2001). AHP is a highly
flexible decision methodology that is
typically used in decision situations, which
involve selecting alternatives from several
candidate decision alternatives on the basis
of multiple decision criteria of a competing
or conflicting nature. Particularly important
for the trust factor selection situation, the
decision criteria may hold a different
perceived degree of preference or level of
importance to the decision in the eyes of
the decision makers (i.e. scientists and
AHP helps to bring consistency in selection
problems whose decision criteria are
expressed in subjective measures based on
managerial experience.
The basic steps in the computational
process of AHP are described in the
Appendix. In this paper, we compute the AHP
weighting using the computer software
application Expert Choice (Decision Support
Software, 2002). The procedure for computing
the AHP preference vector used in this case
to decide on trust factors can be found in
most basic decision science textbooks
(see Render and Stair (2000)).
Schniederjans and Garvin (1997) found
that, by combining AHP with GP, decision-
makers could incorporate mathematically
adjusted weightings on select decision
criteria not already reflected in the AHP
preference weightings. The GP model
permits an added priority structure
reflecting the added mathematical weighting.
The computational aspects of the GP
approach are also indicated in the Appendix.
The GP model adds an additional dimension
of flexibility in pre-emptively restating
several levels of managerial preferences
within a decision-making model.
Application
The AHP approach
The use of AHP was investigated first as a
means to establish a trust factor ranking.
[ 589 ]
Larry L. Radcliffe and
Marc J. Schniederjans
Trust evaluation: an AHP and
multi-objective programming
approach
Management Decision
41/6 [2003] 587-595
[ 590 ]
Utilizing the Expert Choice software
(Decision Support Software, 2002),
preferences were established for criteria and
sub-criteria. Survey average scores from
Table II were used to build alternative
preferences required in the process.
Consistent with AHP, a ``decision hierarchy''
(i.e. a chart) is established to map out how the
final decision is related to the multiple
criteria in the decision environment. This
hierarchy is presented in Figure 1.
Application issues used in the Expert
Choice generated AHP solution of the trust
factor problem are described below:
. SNS management viewed each of the
criteria as being equal (i.e. management's
criteria were no higher in importance
than the technical staff's or the support
staff's criteria). A preference value of 1/3
was consequently assigned to each.
. Sub-criteria preferences were based on
management's weighting of 8, 5 and 1 for
the three trust categories; co-located team
trust, virtual team trust, and swift trust.
. The AHP preferences for decision
variables (i.e. trust factors) under each
sub-criterion were assigned normalized
values. These values were based on trust
factor average preference scores from the
SNS trust survey (Table II).
. Additional tests for AHP consistency were
undertaken to ensure the computational
accuracy of the application. This test
evaluates the validity of the basic AHP
pair-wise comparison process (Render and
Stair, 2000; Saaty and Vargas, 2001). A
consistency index to random index ratio
of less than 0.1 indicates a satisfactory
degree of consistency (Render and Stair,
2000; Saaty and Vargas, 2001). The Expert
Choice software calculated the resulting
ratio at less than 0.001, indicating near
perfect consistency for this trust factor
problem.
Figure 1
AHP criteria, sub-criteria and alternatives
. Overall scores for xjs using the Expert
Choice AHP solution and the resulting
ranking of trust factors are shown in
Table III.
Based on the AHP ranking results, SNS
management should first place their efforts
in building trust on co-located team trust
(all factors) and second on virtual team trust
(just the performance and competency
factor). If this decision based on the AHP
ranking is true, then another methodology
should reveal a similar result. To explore
the validity of this AHP ranking and for
comparison purposes, we next turn to AHP's
use in a GP model.
Goal-programming (GP) approach
The detailed development of the GP model
is also described in the Appendix. For
the SNS management problem, the GP
model formulation consisted of seven
decision variables, three priorities and
20 goal constraints (i.e. three management
constraints and nine trust factor category
constraints; one AHP weight constraint;
and seven maximum preference score
constraints).
The GP objective function had to reflect the
desire to permit the technical and support
staff's goal constraints to be viewed as being
as important as management's goal
constraint. Management also wanted to
assign weights of 8, 5 and 1 to constraints
associated with co-located team trust, virtual
team trust and swift trust ``categories''. These
weights corresponded to the sub-criteria
preferences established using the AHP
process.
The AB:QM software system (Lee, 1996)
was used to generate the GP solution. The
computational results indicated that all of
the goals as stated in the objective function
have been fully achieved. The optimal GP
weights of the trust factors and their
corresponding ranking are presented in
Table IV.
Based on the GP ranking results, SNS
management should first place their efforts
in building trust on co-located team trust
(just the management and team relationships
factor) and second on virtual team trust (just
the performance and competency factor).
While these are the same two categories as
chosen by the AHP method, the GP solution
helped narrow the co-located team trust
factors from three to just one.
Results and discussion
Both methodology rankings suggested the
same results with regard to the following:
Larry L. Radcliffe and Table III
Marc J. Schniederjans Ranking of trust factors using AHP
Trust evaluation: an AHP and
multi-objective programming Trust factors Resulting Resulting
approach
(decision variables/ AHP AHP
Management Decision
41/6 [2003] 587-595 alternatives) Trust categories weighting rankinga
x1 Co-located team trust: organizational policies 0.192 1
x2 Co-located team trust: management and team relationships 0.193 1
x3 Co-located team trust: communications 0.188 1
x4 Virtual team trust: performance and competency 0.134 2
x5 Virtual team trust: integrity 0.116 3
x6 Virtual team trust: concern for others 0.107 4
x7 Swift trust: quick trust 0.071 5
Note: a The larger the weight, the higher the rank. The equal rank of ``1'' for decision variables x1, x2, and x3 is
typical in AHP studies where there is no statistically significant difference in AHP weighting values

Table IV
Trust factor rankings based on combined AHP/GP decision methods
Trust factors Resulting Resulting
(decision variables/ GP GP
alternatives) Trust categories weighting rankinga
x1 Co-located team trust: organizational policies 5 3
x2 Co-located team trust: management and team relationships 7.484 1
x3 Co-located team trust: communications 5 3
x4 Virtual team trust: performance and competency 6.220 2
x5 Virtual team trust: integrity 5 3
x6 Virtual team trust: concern for others 5 3
x7 Swift trust: quick trust 5 3
Note: a The larger the weight, the higher the rank

. Most of management's emphasis should be respectively). This shifted the analytical

placed on building trust via the co-located
team trust and virtual team trust
categories.
. The ranking of the individual trust factors
of ``co-located team trust: management
and team relationships'' and ``virtual team
trust: performance and competency'' were
first and second, respectively.
. They both identified swift trust with the
lowest priority and should receive the
least emphasis.
On these points, the methodologies appear to
help validate each other's answers. While it
can be argued that the GP model was based
on the outcome parameters of the AHP
method (and therefore had to come up with
the same solution), the GP model introduced
additional decision-making criteria (in the
form of personnel preference weighting of
trust categories) that could have made, but
did not make, a substantial difference in the
resulting solution for the highest ranked
trust factors.
The GP solution is different in that it
adjusted the individual trust factor
preferences from the AHP method to include
SNS management's category weighting
(i.e. 8, 5 and 1 for co-locate, virtual, and swift,
process to favor better the co-located team
and virtual team trust category factors,
resulting in a different priority structure
from using AHP alone. It is interesting to note
in the bottom row (i.e. total row) of Table II,
the two trust factors (i.e. co-locate:
management and team relationships and
virtual: performance and competency) with
the largest summed average preference
scores are the same as those GP model
assigned the highest two ranks. In this
regard, the GP model appears to have
provided a solution that better reflects the
trust survey findings than the AHP
weightings. SNS management expected
co-located team trust factors to be important.
The results of the combined AHP and GP
decision methodologies have in this
application reinforced their expectation.
The application and information provided
by the GP method have helped SNS
management to initially focus their
trust-building efforts more intensely on a
fewer number of trust factors. This is
expected to help reduce management's use of
resources in their efforts to build trust and
helps to enhance the resulting level of trust
necessary in what will be one of the world's
largest energy research projects.
[ 591 ]
Larry L. Radcliffe and
Marc J. Schniederjans
Trust evaluation: an AHP and
multi-objective programming
approach
Management Decision
41/6 [2003] 587-595
[ 592 ]
Summary
This paper has described an application of
AHP and GP as methodologies to establish
ranks for managerial efforts in building trust
on a large, virtual team project. Remotely
located team members have completed their
assignments and are transitioning off the
project. The AHP and GP application has
determined factors on which management
should now focus to improve trust among the
staff located at the new facility.
The benefits of using the proposed methods
as trust factor selection aids include:
. When AHP is used with GP, GP considers
all multiple, judgmentally-based criteria
and preferences required in complex
managerial problem situations.
. When used individually, they provide a
simple measure that can easily be used to
establish clear rankings for management
goals.
. There is also an implied fairness in the
combined GP approach that permits
greater flexibility in including
mathematical weighting of important
decision-making criteria than with AHP
alone, while allowing participants (in this
study the technical and support staff) to be
given the same priority in the model's
solution (as that of management).
From a managerial perspective, the efficient
allocation of resources by SNS managers
were, with the aid of these methodologies,
better able to prioritize their trust-building
efforts. That prioritization permitted a more
able focusing of those limited resources on
the best trust factors that would efficiently
build trust in the project. Also, the trust
survey used in this study creates a
methodology that is expected in the future to
be used to monitor ongoing trust issues as
they arise in the research partnerships
necessary to complete such a large project.
The values of these later surveys can then be
used as a means to identify emerging new
trust issues and where SNS management can
best reallocate their trust-building efforts on
an ongoing basis.
Fortunately the size of this application
helps to limit the possible mathematical
complications and limitations of the AHP and
GP methodologies. While Saaty and Vargas
(2001) discuss a variety of AHP limitations,
the fact that this problem only has seven
decision variables makes this application
ideal for AHP and the resulting accuracy
ratio helps to confirm this fact. Romero (1991)
likewise has substantially itemized a variety
of possible problems with the use of GP, but
again the simplistic nature of the problem
(e.g. few priorities, etc.), the small size of the
model, and the lack of conflicting goals in the
solution, all help minimize the possibility of
mathematical complications in the use of GP
in this application.
References
Badri, M.A. (1999), ``Combining the analytical
hierarchy process and goal programming for
global facility location-allocation problem'',
International Journal of Production
Economics, Vol. 62 No. 2, pp. 237-48.
Bryson, N. (1996), ``Group decision making and
the analytic hierarchy process: exploring the
consensus-relevant information content'',
Computers & Operations Research, Vol. 23,
pp. 27-35.
Decision Support Software (2002), Expert Choice,
Decision Support Software, McLean, VA.
Duarte, D.L. and Tennant-Snyder, N. (1999),
Mastering Virtual Teams: Strategies, Tools,
and Techniques That Succeed, Jossey-Bass,
San Francisco, CA.
Glanz, J. (2001), ``Particle physicists plan the next
big thing'', New York Times, 10 July, p. 2B.
Hardt, P.O. and Brynteson, R. (1998), ``Swift
virtual trust'', Ibiz Magazine, December,
available at: www.ibiz.net/dec98/table.htm
Jarvenpaa, S.L. and Shaw, T. (1998), ``Global
virtual teams: integrating models of trust'',
Proceedings of the Volnet-Workshop, Institute
of Information Systems, Department of
Information Management, University of
Bern, 27-28 April, available at: http://
virtual-organization.net/news/
workshop98_abstract.pdf
Kim, P.O., Lee, K.J. and Lee, B.W. (1999),
``Selection of an optimal nuclear fuel cycle
scenario by goal programming and the
analytical hierarchy process'', Annals of
Nuclear Energy, Vol. 26, pp. 449-60.
Kristof, A.L., Brown, K.G., Sims, H.P. Jr and
Smith, K.A. (1995), ``The virtual team:
a case study and inductive model'', in
Beyerlein, M. M., Johnson, D.A. and
Beyerlein, S.T. (Eds), Advances in
Interdisciplinary Studies of Work Teams:
Knowledge Work in Teams, Vol. 2, JAI Press,
Greenwich, CT, pp. 229-53.
Lee, S.M. (1996), AB:QM Systems Software,
Prentice-Hall, Englewood Cliffs, NJ.
Meyerson, D., Weick, K.E. and Kramer, R.M.
(1996), ``Swift trust and temporary groups'',
in Kramer, R.M. and Tyler, T.R. (Eds),
Trust in Organizations: Frontiers of Theory
and Research, Sage, Thousand Oaks, CA,
pp. 166-95.
Render, B. and Stair, R.M. (2000), Quantitative
Analysis for Management, 7th ed., Prentice-
Hall, Upper Saddle River, NJ.
Romero, C. (1991), Handbook of Critical Issues in
Goal Programming, Pergamon Press, Oxford.
Larry L. Radcliffe and Saaty, T.L. and Vargas, L.G. (2001), Models,
Marc J. Schniederjans Methods, Concepts and Applications of the
Trust evaluation: an AHP and Analytical Hierarchy Process, Kluwer
multi-objective programming
approach Academic Press, Norwall, MA.
Schniederjans, M.J. and Garvin, T. (1997), ``Using
Management Decision
41/6 [2003] 587-595 the analytical hierarchy process and
multi-objective programming for the
selection of cost drivers in activity-based
costing'', European Journal of Operational
Research, Vol. 100, pp. 72-80.
Appendix
Analytical hierarchy process ±
methodology
The computational aspects of AHP involve
several steps, as outlined by Saaty and
Vargas (2001). The AHP procedure involves
six steps:
1 Establish the ``decision hierarchy''. The
decision maker must identify:
. the overall decision;
. the factors that must be weighted or
used to make the decision; and
. the alternative choices from which a
ij xj di‡ ‡ di ˆ ai ; for i ˆ 1; . . . ; m …A2†
decision is to be made.
2 Establish pair-wise comparisons of all
alternatives. The decision maker must
compare each alternative with all other
ij xj dic ‡ dic ˆ aic ; for i ˆ 1; . . . ; m
alternatives, one factor at a time. This is
accomplished using a weighting process
that ranges in values from 1 to 9
(i.e. 1: equally preferred; 2: equally to
moderately preferred; 3: moderately
preferred; 4: moderately to strongly
preferred; 5: strongly preferred; 6: strongly
to very strongly preferred; 7: very strongly
preferred; 8: very strongly to extremely
strongly preferred; 9: extremely strongly
preferred).
3 Compute factor priorities. The decision
maker uses the previous comparison
ratings to compute a set of priorities for
the individual factors by averaging them,
in this case, by trust factors.
4 Compute factor weight priorities. The
ij in
decision maker uses the same approach as
presented in steps 2 and 3 above, but this
time it is applied to a comparison of the
trust factors with each individual trust
factor.
5 Compute the overall decision priorities or
preference weightings. The decision maker
uses the factor weights from step 4, and
the values from step 3 to compute expected
values for the overall decision priority
weighting for each of the alternatives.
6 Determine consistency ratios. AHP does
more than just generate a solution, it
includes additional analysis, which
permits decision makers to investigate if
the subjective ratings by managers are
consistent enough to justify using the
resulting overall AHP decision priorities.
In other words, AHP checks itself to make
sure the ratings consistently make sense
for the purpose of using the AHP analysis
on which to base a decision.
The implementation of these steps results
in a preference vector or weighting, wAHP j
= (wAHP
1 , . . . , wAHP
n ), where wAHP
j is the
assigned preference in the selection out of
some n number of possible trust factors. The
greater the weight, the greater the preference
for the specific jth trust factor.
Goal programming ± methodology
Extending the use of the AHP methodology,
the pre-emptive, weighted GP model for the
trust factor selection problem can be
generally expressed as below:
K X
X
min Z ˆ Pk …wi di † …A1†
kˆ1 i 2 m
subject to:
Xn
jˆ1
C X
X n
‡
cˆ1 jˆ1
…A3†
X
n
wAHP xj d‡ ‡ d ˆ 1 …A4†
j
jˆ1
xj di‡ ‡ di ˆ ;
…A5†
for j ˆ 1; . . . ; n; i ˆ 1; . . . ; m
and
xj ; di‡ ; di ; dic
‡
; dic ; d‡ ; d  0; for all
…A6†
j ˆ 1; . . . ; n; i ˆ 1; . . . ; m; c ˆ 1; . . . ; C
where: x1 , x2 , . . . , xn are n non-negative
decision variables (or trust factors); m in
equation (A2) is a set of goal constraints
modeling the personnel preferences;
equations (A2) and (A3) are the total
preference score averages by the ith type of
personnel and the jth type of trust factor from
the trust survey (values found in Table II);
c = 1, . . . , C in equation (A3) are a set of goal
constraints permitting the prioritization of
select ``categories'' of trust factors; wi in
equation (A1) are the pre-emptive weighting
subjectively incorporated to weight the
importance by selective personnel of
categories of trust factors; wAHP j in equation
(A4) are the AHP weightings; ai and aic are
right-hand-side values in equations (A2) and
(A3) that represent the desired preference
score for each type of personnel goal
constraint;  in equation (A5) is the
[ 593 ]
Larry L. Radcliffe and
Marc J. Schniederjans
Trust evaluation: an AHP and
multi-objective programming
approach
Management Decision
41/6 [2003] 587-595
ij , times the maximum possible
ij parameters are found in the
[ 594 ]
maximum preference score possible for each
of the decision variable values; and all di‡ and
di deviation variables represent the possible
deviation from desired goal targets.
In this model we seek to minimize total
negative deviation (represented by Z) from all
desired goal targets. The goals are ordinally
ranked by priorities, Pk in equation (A1),
where k = 1, 2, . . . , K ordinal ranks and
Pi > P2 > PK . The ai and aic are right-hand-
side values representing goal targets in
equations (A2) and (A3) and are calculated for
each goal constraint by multiplying the value
of the largest total preference score average
value,
‡ 3:821x5 ‡ 3:670x6 ‡ 3:439x7 …A7†
preference score (i.e. 5 out of a 1-to-5 scale),
times the number of variables in that
particular constraint. The purpose of this
computation is to establish a right-hand-side
parameter representing an idealistically high
goal target for each goal constraint to seek.
Specifically, to force the model's solution to
heavily weight the decision variables with
the largest total preference score average
value. By placing the subjectively wi
pre-emptive weighting at the higher Pk
priority than the analytically determined
AHP weightings of the individual trust
factors an additional emphasis is placed on
the selection of whole categories of trust
factors, before their individual AHP weighted
importance is considered in the solution.
In summary, the objective function in
equation (A1) seeks to minimize all negative
deviation from the targeted goals, allowing
the xj to take on a positive weighting value;
the higher the weight, the more desirable the
trust factor. Equation (A2) seeks to weight
heavily those xj trust factors that offer the
largest total preference score average. In
equation (A3) weighting of the xj trust factors
that offer the largest total preference score
average again takes place, but this time with
the added management preference ranking,
wci incorporated to weight the importance of
selective ``categories'' of trust factors. Since
all of equations (A2) and (A3) are at the first
P1 priority, the wi weighted goal constraints
will receive pre-emptive attention in the final
solution, over the total preference score
averages. In equation (A4) the AHP
weightings are included to force the model's
solution to seek values for xj trust factors that
reflect the pair-comparisons performed by all
participants in the trust survey. Equation
(A5) seeks to set minimum weighting equal to
the maximum score value used in the survey
(i.e.  has a maximum value of 5 in this study).
The development of the GP model for the
SNS management problem consisted of seven
decision variables, three priorities and 20
goal constraints (i.e. three management
constraints based on equation (A2) and nine
trust factor category constraints based on
equation (A3), all at P1 ; one AHP weight
constraint based on equation (A4) at P2 ; and
seven maximum preference score constraints
based on equation (A5) at P3 ). Considerations
in developing the model include the following
points:
. There were three types of personnel
constraints using equation (A2). An
example of the first management goal
constraint is:
3:833x1 ‡ 4:038x2 ‡ 3:906x3 ‡ 4:042x4
‡ d1 d1‡ ˆ 141:47:
Note all the
average survey preference scores from
Table II. Note also that the maximum of
these constraint parameter values is 4.042,
which, when multiplied by the maximum
possible score of 5 and by the seven
decision variables in the constraint,
results in the right-hand-side goal target
value of 141.47 (i.e. 4.042  5  7).
. There were nine trust ``category''
constraints in the model using
equation (A3). While we only have three
types of personnel constraints, there are
three types of trust categories (C),
resulting in the need for nine different
goal constraints. An example of the
first trust category goal constraint
(for co-locate team trust for management
personnel) is:
3:833x1 ‡ 4:038x2 ‡ 3:906x3 ‡ d4 d4‡
…A8†
ˆ 60:57:
Note we again are using the management
personnel average preference scores from
Table II and the right-hand-side value is
now computed by multiplying the largest
parameter (i.e. 4.038) times the same
maximum score of 5, times the three
variables in the constraint, or 60.57.
. There was only one AHP goal constraint
using equation (A4). This goal constraint
utilized the AHP values from Table III and
is expressed as:
0:192x1 ‡ 0:193x2 ‡ 0:188x3 ‡ 0:134x4
‡ 0:116x5 ‡ 0:107x6 ‡ 0:071x7 …A9†
‡
‡ d13 d13 ˆ 1:
Note the right-hand-side target value of 1
(based on the logic that the sum of the
AHP weights is equal to 1) which is only
used here to help bind the resulting
solution decision variable values in line
with the proportions of the AHP
weightings.
Larry L. Radcliffe and
Marc J. Schniederjans
Trust evaluation: an AHP and
multi-objective programming
approach
Management Decision
41/6 [2003] 587-595
. The objective function from equation (A1)
. There were seven trust factor constraints
in the model using equation (A5). An
example of the first trust factor goal
constraint is:
x1 ‡ d14 ‡
d14 ˆ 5: …A10†
had to reflect the desire to permit the
technical and support staff's goal
constraints to be viewed as being as
important (at the first priority) as
management's goal constraint. This was
accomplished by attaching wi of 1 to each
of the di variables representing the
``personnel'' goal constraints in the
objective function. Management also
wanted to assign wi weights of 8, 5 and 1 to
constraints associated with co-located team
trust, virtual team trust and swift trust
``categories''. These weights corresponded
to the sub-criteria preferences established
using the AHP process.

Application questions
1 Why is the concept of trust in large-scale 3 How did the decisions given by the AHP
projects important? method differ from the GP model? What
2 When would you consider using two might have caused the difference?
analytic methods in your selection
process rather than just one?

[ 595 ]