APPLYING CASE-BASED REASONING FOR IDENTIFYING
THE NEGOTIATION PROFILE OF ELECTRONIC
NEGOTIATION SYSTEM USERS
Jakub Brzostowski
1
and Tomasz Wachowicz
2
1
Institute of Mathematics, Silesian University of Technology, ul. Kaszubska 23, 44-100 Gliwice, Poland
2
Department of Operations Research, University of Economics in Katowice, ul.1 Maja 50, 40-287 Katowice, Poland
Keywords: Negotiation profile, Conflict modes, Case-based reasoning, Electronic negotiation systems, Reputation
systems.
Abstract: In this paper we analyze the problem of identifying the negotiation profile of the electronic negotiation
system users. Usually such a profile is identified by means of the specific questionnaire (e.g. the Thomas-
Kilmann questionnaire), however it requires from the negotiator answering many troublesome questions
which is tiring and may lead to unreliable results. On the other hand many behavioural and psychological
studies confirm that there is a set of demographical and sociological characteristics that influence the human
general behaviour. Deriving from these studies we try to determine such a profile by analyzing the general
information provided by the pre-negotiation questionnaire the users fill while creating their negotiation
accounts. Having the historical data of Inspire negotiation system we try to find links between a set of the
data that describes the negotiators demographical features and their final negotiation profile using the notion
of Gilboa and Schmeidler case-based reasoning (CBR). To determine all the parameters required for the
case-based reasoning the statistical correspondence analysis on the set of the historical data is conducted in
advance. The results of CBR-based profile identification are also presented and discussed.
1 INTRODUCTION
The negotiation profile of a negotiator, as we see it,
is a set of negotiator's features, such as
cooperativeness, selfishness, assertiveness etc., that
describe negotiator’s behaviour in conflict
situations. It describes in fact a bargaining style of a
negotiator, which is a relatively stable, personality-
driven cluster of behaviours and reactions that arise
in negotiating encounters (Shell, 2001). This
bargaining style is, in turn, determined by
negotiator’s individual characteristics such as:
cultural (Adair and Brett, 2004) or demographical
(Jehn et al., 1997) ones or the visible personal
characteristics like age and gender (Kray and
Thompson, 2005). Some research confirm the
impact of these characteristics on the negotiation
process and outcomes (Sternberg and Dobson, 1987;
Thompson, 1990; Kersten et al., 2003) as well as the
impact of other factors, such as the motivation
styles, abilities or enduring dispositions (Elfenbein
et al., 2008). Thus it is important to be aware of
one’s negotiation profile, since it may influence the
negotiation process and the atmosphere of the
forthcoming negotiations. Having a particular
negotiation profile and using the corresponding
bargaining style makes the negotiators to be more or
less willing to use thus-or-such negotiation strategy
and tactics and consequently, to represent different
attitude towards the negotiation problem and to their
negotiation counterpart. Moreover, knowing the
counterpart’s negotiation profile may allow
negotiator preparing better to the forthcoming
negotiation, identifying their own strategies that fit
the counterpart’s position and style the best and
allow to influence them most efficiently to achieve
negotiator’s goals.
Negotiation profile can be identified twofold.
First approach is to use a psychometric instrument
based on the series of questions posed to the
negotiator and derive their negotiation profile by
analyzing different combinations of their answers.
Second approach is based on the extraction of the
levels of the profile’s features from the negotiation
thread by analyzing the communication process
between the parties. An example of the first
48
Brzostowski J. and Wachowicz T..
APPLYING CASE-BASED REASONING FOR IDENTIFYING THE NEGOTIATION PROFILE OF ELECTRONIC NEGOTIATION SYSTEM USERS.
DOI: 10.5220/0003755100480058
In Proceedings of the 1st International Conference on Operations Research and Enterprise Systems (ICORES-2012), pages 48-58
ISBN: 978-989-8425-97-3
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
approach is the Thomas-Kilmann conflict mode
instrument (TKI) proposed by Kilmann and Thomas
(1983). An application of such an instrument results
in levels of belonging of the negotiator to five
regions spanned on a plane with axes corresponding
to assertiveness and cooperativeness. The second
approach aims at measuring the degree of profile’s
features based on the types of responses given by a
negotiator during the process of exchanging
messages between two sides of the encounter (see
Brzostowski and Wachowicz, 2010). The type of
response, which may be positive, negative or
neutral, is scored with positive sign for positive
messages and negative sign for negative messages.
The scores also depend on the degree of importance
of a request and the degree of response satisfaction.
The problem with the first approach is that filling the
questionnaire and answering the series of question
may be time consuming, tiresome and discouraging
for the negotiator, while in the second approach
there is no knowledge allowing for building
negotiator's profile if the negotiator did not start any
encounter, yet.
In this paper we address a problem of identifying
the negotiation profile of the new electronic
negotiation system users with no negotiation history
behind them, but avoiding the specific psychometric
questionnaires (like the TKI ones). The main goal of
this paper is to prepare the formal mechanism that
would allow for eliciting the profile from the basic
descriptive information the user is giving while
registering to the system (creating an user account)
and filling the pre-negotiation questionnaire. To
solve this problem we propose to reason from the
case base that describes the historical negotiations
within which both the descriptive information of the
parties and their negotiation profiles were recorded.
It will allow us to conduct the case-based reasoning
(CBR) using the Gilboa and Schmeidler approach
(1995).
The paper has 6 more sections. The research
context that was the initial point for this paper is
presented in Section 2. Then, in Section 3 the
description of the case-based reasoning approach is
given. In Section 4, the set of historical data required
for the case-based reasoning is described, while in
Section 5 the application of CBR main ideas for
determining the negotiator’s bargaining profile
(based on the historical data) is shown. The issue of
determining the weight parameters required for case-
base reasoning is discussed in section 6. In section 7
the results of our approach are presented and
discussed. We conclude with some remarks on the
future work.
2 RESEARCH CONTEXT
The problem we have risen in the previous section is
a part of the research we are carrying out in order to
build a comprehensive electronic negotiation system
that would support its users throughout the whole
negotiation process, starting with the pre-negotiation
preparation and ending with the post-negotiation
optimization of the negotiated agreement
(Brzostowski and Wachowicz, 2009). The system,
called NegoManage, is designed to be a negotiating
platform that would allow negotiators to define the
negotiation problem, find the suitable counterpart
and negotiate the contract. NegoManage is a kind of
distributed system with the core deployed on the
Web and responsible for supporting communication
among the users and sharing the public information
about them (NegoManage Communication Unit).
There are also the satellite sub-systems responsible
for various functionalities such as the preference
elicitation subsystem, data visualization unit,
reputation subsystem, etc. Some of them are also
deployed on the Web, while the others need to be
installed on the users desktop computers. Such a
structure was chosen due to the security reasons.
Some information are strategic (like the negotiators
preferences) and should not be revealed to the
counterparts, therefore the subsystems responsible
for processing such information are deployed on the
users computers (NegoManage Individual Unit) and
there is no an external access to them from the level
of Web based units. The general architecture of the
system is presented in Figure 1.
Figure 1: The architecture of the NegoManage system.
One of the key elements of NegoManage system
is the reputation subsystem. According to the
definition of Howison (2009) the reputation in the
context of electronic interaction may described such
features of the systems and users as trustworthiness,
quality or any other characteristic specific to the
analyzed domain. NegoManage reputation
subsystem describes the negotiator’s profiles using
APPLYING CASE-BASED REASONING FOR IDENTIFYING THE NEGOTIATION PROFILE OF ELECTRONIC
NEGOTIATION SYSTEM USERS
49
two major negotiator’s characteristic: assertiveness
and cooperativeness, derived directly form the Dual
Concerns Model (Blake and Mouton, 1964). As
addressed in Section 1, such a profile could be
elicited by means of TKI and the Wharton-TKI
Bargaining Styles Grid (Shell, 2001) that allows to
recalculate TKI five-characteristic scores into the
two-dimension space of assertiveness and
cooperativeness. However, in NegoManage system
we aim to build the reputation system that reflect the
true negotiation profile the system user presents
during the subsequent encounters. We decided then
to determine the negotiators’ profiles based on the
analysis of messages exchanged between the players
in the process of negotiation. Deriving form the
Searle’s and Stile’s speech act taxonomies (Searle,
1969; Stiles, 1992) we proposed a new negotiation-
context depended speech act taxonomy (Brzostowski
and Wachowicz, 2010), based on which the
exchanged messages are classified and used later on
by the profiling mechanism. The mechanism, that
works according to formal algorithm, checks how
the message receiver is responding to the request of
message sender. Next, the mechanism determines
the degrees of the negotiator’s features (i.e.
assertiveness and cooperativeness). The final degree
of a feature is computed as an average degree of a
feature for multiple past negotiations. The
negotiation profile calculated this way is then
displayed by the reputation subsystem to all
NegoManage users and may be used by them for
browsing the most appropriate (from the behavioural
point of view) counterpart and for adequate
preparation of the negotiation strategy in the
forthcoming negotiations.
The only problem that occurs in our approach is
how to identify the initial negotiation profile of the
new system user, that did not conduct the
negotiation in NegoManage system yet. Since we
wanted to avoid using the TKI or similar solutions,
such as the Kraybill Conflict Style Inventory
(Kraybill, 2005) or Myers-Briggs Type Indicator
(Briggs and Myers, 1980), we decided to identify
such a profile based on the visible demographic
characteristics and psychological description of the
user that are available within their pre-negotiation
questionnaires. We assume then, having derived
from the results of behavioural research on
negotiation and the five-factor model (Mershon and
Gorsuch, 1988; Paunonen and Ashton, 2001;
Herrmann, 2004; Patton and Balakrishnan, 2010),
that there is any relation between some personal
characteristics of the negotiators and their
negotiation profiles. To find the relation and
describe it formally we will apply the case-based
reasoning (see Section 3). For such a reasoning the
historical data of the previous negotiation
experiment is needed that would provide all the
information required to build both the input and
output data. The case base we will use would be
comprised with the Inspire electronic negotiation
system data (see Section 4).
3 CASE-BASED REASONING
AND CASE-BASED PROBLEM
SOLVING
The idea of case-based problem solving is based on
the postulate that similar problems have similar
solutions (see Aamodt and Plaza, 1994; Gilboa and
Schmeidler, 1995). In the classical case-based
problem solving (Leake, 1996) we use the past
solutions of past problems to solve a new problem.
The idea of case-based problem solving is illustrated
in Figure 1.
The CBR mechanism retrieves from the case base
the relevant cases and adapts them to fit a new
problem. In the first stage of CBR features of the
current situation that are really relevant are
determined. In the next stage the CBR mechanism
retrieves from the most relevant prior cases or case.
Then the retrieved case or cases is adapted to fit the
new situation. After applying the solution suggested
by CBR the new case is stored in the case base.
Our application of the CBR concept differs from
the typical application. Instead of finding the
solution to a new problem we use the CBR concept
to predict the outcome of a new situation. We do not
consider the whole CBR cycle concerning the full
methodological framework of this approach
(Aamodt and Plaza, 1994). Instead, we focus on the
inference stage within the overall process of CBR. In
our particular application context we can consider a
decision problem of selecting the negotiation partner
with desired conflict resolution style. The decision
of selecting the negotiation partner is based on
similar encounters in the past. The concept of Case-
Based Decision Taking (Gilboa and Schmeidler,
1995) is used when each case may be split into three
components: the decision problem (situation), the act
that was chosen by the decision maker in this case
and the outcome received by the decision-maker.
The three abstract sets corresponding to these three
component may be introduced:
1. P - the set of decision problems. In our case
the decision problem is described by the
ICORES 2012 - 1st International Conference on Operations Research and Enterprise Systems
50
Figure 2: Leak's model of CBR accroding to (Leake, 1996).
demographic description of the negotiation
partner in the past negotiation.
2. A - the set of possible acts. The set of
potential negotiation partners. The act in this
case is a choice of one of those partners.
3. R – the set of conceivable results. The set of
resulting conflict resolution styles of the
chosen negotiator partner. In other words, it
is the description of encountered negotiation
behaviour of the chosen partner.
The product of these three sets gives us the set of
all possible cases
RAPC ×
×
=
. The given case
base M is a subset of the set C (
CM
).
4 INSPIRE’S CASE BASE
Inspire (Kersten and Noronha, 1999) is an online
system that has been used for years, mainly for
training and teaching negotiations (Paradis et al.,
2010). It supports negotiating parties in bilateral
negotiation process: pre-negotiation preparation,
actual conduct of negotiation and post-negotiation
phase. In the pre-negotiation phase the Inspire users
are asked to fill the questionnaire giving the basic
personal information about themselves and their
attitude towards the negotiation problem, process
and counterpart. Then they take the Thomas-
Kilmann test and have their bargaining style
identified this way. Inspires provides also its users
with the preference elicitation, that allow building
their own negotiation offers scoring systems. In the
actual negotiation phase the system supports the
communication between the parties, helps them to
evaluate negotiation offers and visualizes the
negotiation progress. In the post-negotiation phase
Inspire analyzes the compromise (if achieved) and
browses for its improvements using the notion of
Pareto-efficiency.
The case base required for our analysis is
comprised of the data collected by Inspire system
within the pre-negotiation phase. The base contains
the description of the users’ demographic and
personal features provided by the pre-negotiation
questionnaire and the TKI’s conflict style regions
(i.e. competing, collaborating, compromising,
avoiding and accommodating) identified by means
of TKI test. The set of users’ personal information
with the pre-defined resolution levels for the closed-
form questions are shown in Table 1.
APPLYING CASE-BASED REASONING FOR IDENTIFYING THE NEGOTIATION PROFILE OF ELECTRONIC
NEGOTIATION SYSTEM USERS
51
Table 1: The user’s personal characteristics recorded by Inspire system.
Description Options
Program of study (ed_field)
Arts and Fine Arts (1); Business & Management (2); Communication (3); Computer Science &
Information Technology (4); Education (5); Engineering (6); Health, Medical & Nursing (7);
Humanities (8); International Studies (9); Law & Criminal Justice (10); Mathematics & Statistics
(11); Psychology (12); Public Affairs, Administration (13); Social & Behavioural Sciences (14);
Others (15)
Level of stud (ed_level) High School (1), Undergraduate (2), Graduate (3)
Gender (gender) Female (1), Male (2)
Age group (age) 20 or less (1); 21-25 (2); 26-30 (3); 31-40 (4); 41-50 (5); 51 or more (6)
In which country was the user born? (c_born) Two letters UN symbol (numerical code)
In which country does the user currently reside?
(c_reside)
Two letters UN symbol
How long does the user reside in the country of
residence? (l_reside)
6 months or less (1); 6 months to 1 year (2); 1 year to 2 years (3); 2 years to 4 years (4); 4 years
to 7 years (5); 7 years or more (6)
First language (m_tongue) Name (numerical code)
Did the user participate in negotiation experiments
before? (p_before)
Yes (1); No (2)
Did the user use a decision support or negotiation
software before? (NSS)
Yes (1); No (2)
Did the user attend negotiation course/seminar before?
(course)
None (1); one (2); two or more (3)
How does the user rate her knowledge of negotiation?
(knowledge)
From novice (1) to expert (7)
How does the user rate her English proficiency?
(english)
From poor (1) to excellent (7)
In case-based decision theory each case consists of
two components: the situation and the outcome that
was experienced for this situation. For the
negotiation profile identification each case would
consist of demographic values as an input
(description of the situation) and the bargaining style
obtained as a result of TKI questionnaire as an
output (description of the outcome in terms of the
scores of the five regions of styles). The sample case
base input and output are shown in Table 2 and
Table 3.
Table 2: The case base (input) containing user description
in terms of demographic variables.
Vectors of input
age gen
der
c_r
esid
e
l_re
side
c_b
orn
m_t
ong
ue
eng
lish
cou
rse
ed_
leve
l
NS
S
kno
wle
dge
p_b
efor
e
ed_
fiel
d
2 2 1 3 16 1 6 2 3 1 2 1 2
2 1 1 6 1 1 6 1 2 1 2 2 2
2 2 1 6 1 1 5 1 2 1 2 1 2
1 2 1 6 1 1 7 2 2 1 5 2 2
3 2 1 6 1 1 4 1 2 1 4 1 2
Table 3: The case base (output) containing resulting
conflict mode regions descriptions.
Vectors of output
Competing Collaborating Compromising Avoiding Accomodating
4 3 10 7 4
5 2 9 7 5
6 6 9 5 2
9 3 10 2 4
4 5 10 6 3
5 APPLYING CBR FOR
IDENTIFYING THE
NEGOTIATORS BARGAINING
STYLE
The case-based prediction in our application aims at
estimating negotiator's conflict resolutions
(bargaining) style approach. Therefore as the
situations we have the negotiators demographic
features and as the outcome their conflict resolution
style descriptions. In other words, based on the
knowledge contained in the Inspire’s case base we
aim at predicting the conflict resolution styles based
ICORES 2012 - 1st International Conference on Operations Research and Enterprise Systems
52
on its demographic characteristics. In order to use
the case-based reasoning for identifying the profile
of any new negotiator we have to extract from the
case base the user demographic descriptions most
similar to the demographic descriptions of this new
negotiator (system user) first. Having selected these
descriptions we can estimate the bargaining style of
the new user by comparing the historical conflict
resolutions styles of the similar users in terms of
demographic characteristics.
Two questions arise here. First, how to form a
similarity metric for comparing the situations
(demographics descriptions) and second, how to
adjust the conflict resolutions style descriptions of
the similar negotiators to form a prediction about the
conflict resolution style of the assessed novice
negotiator.
Most of the demographic variables we use are
nominal, and therefore the similarity metric has to be
formed taking into account this feature. The
situation (the user demographics) is modelled here
as a vector of 12 variables. Since these variables are
nominal the similairty metric comparing the
situations x and y on each variable i should have a
simple form:
=
=
yx
yx
yxS
i
1
0
),(
,
(1)
meaning that if two values of a demographic
variable are compared the similarity metric assings 1
to a pair of values if they are identical, and 0 if they
are different. It does not make sense to fuzzify the
metric since the varaible is nominal. The similarity
metric comparing vectors of 12 demographic
characteristics X = (x
1
, …, x
12
) and Y = (y
1
, …, y
12
)
is in turn of the following form:
=
=
12
1
),(),(
i
iiii
yxSwYXS
.
(2)
As shown in the above formula, the similarity of
the demographic descriptions of two users is a
weighted sum of the similarity metrics, comparing
the situations on each coordinate of the demographic
user descriptions. The values
})12,,1{( Kiw
i
are
the levels of importance of all demographic
variables. These values are determined using the
correspondence analysis, as we will show in Section
6.
Having extracted the most similar cases from the
case base by means of the similarity metric S, we
may now adjust the conflict resolution styles
descriptions in the exctracted cases. This style is
described by five variables, namely: competing,
collaborating, compromising, avoiding and
accomodating. These variables are ordinal, which
allows us to use following procedure of adjustement.
Let us consider one variable which describes the
level of belonging to a particular conflict resolution
feature (for instance competing). Let us assume that
in the hypothetical situation we extracted from the
case base a set K of k most similar cases. In this
situation we have k cases in the form of the vectors
as follows:
),,,;,,,(
5211221
jjjjjj
cccddd KK .
(3)
where
},,1{ kj K
,
j
i
d corresponds to the ith
demographic characteristic of the jth case
(
}12,,1{ K
i ) and
j
m
c corresponds to the mth
bargaining style characteristic of the jth case
(
}5,,1{ K
m ). The cases were exctracted from the
case base based on their similarity to the current
situation for which we want to estimate the conflict
resolution style and which we denote as
)
~
,,
~
,
~
(
~
1221
dddD K= . The similarity degree of each
case j from the set K is determined by means of
formula (2) and is denoted by
j
s :
)
~
,( DDSs
jj
= for each },,1{ kj K .
(4)
Having the sequence of
j
s , describing the
similarities of the historical situations with the
current situation, we can estimate the value of
conflict resolution style variable for the current
situation using the concept of Gilboa and Schmeidler
(1995) case-based reasoning. According to the
Gilboa and Schmeidler ideas the estimation of the
bargaining style feature will be computed using the
following formula:
.
~
1
1
=
=
=
k
j
j
j
k
j
j
m
m
s
sc
c
(5)
In the formula (5) the estimation of a feature is
computed as the weighted sum of this feature in
historical situations. The more similar the historical
situation is to the current situation the higher the
contribution of the feature taken from historical
situation.
Let us assume, for example, that for the new
negotiator we have extracted from the case base
three the most similar cases with similarity degrees
as follows:
APPLYING CASE-BASED REASONING FOR IDENTIFYING THE NEGOTIATION PROFILE OF ELECTRONIC
NEGOTIATION SYSTEM USERS
53
7.085.095.0
321
=== sss
Furthermore, let us assume that the bargaining
style descriptions for these cases are as follows:
),4,10,3,8,5(),,,,(
1
5
1
4
1
3
1
2
1
1
=ccccc
),5,11,4,3,8(),,,,(
2
5
2
4
2
3
2
2
2
1
=ccccc
).3,7,10,5,2(),,,,(
3
5
3
4
3
3
3
2
3
1
=ccccc
We use formula (5) to estimate the style’s
features of the new negotiator. For instance, the
estimation of the competing feature (the first of five
bargaining style feature) is the following:
18.5
5.2
7.0285.0895.05
~
3
1
3
1
1
1
=
×+×+×
==
=
=
j
j
j
jj
s
sc
c
.
As we can see in the above formula the first
feature levels (5, 8 and 2) are simply aggregated
with weights corresponding to the similarity degrees
of the current situation and the three historical
situations. The levels of the rest of the bargaining
style features are determined similarly.
The only problem that needs to be solved to used
the case-based reasoning proposed above is to find
the weights describing the levels of similarity,
required for formula (2). In the next section we will
present how these weights may be derived from the
case base using simple statistical tool – a
correspondence analysis.
6 CALCULATING THE LEVELS
OF IMPORTANCE OF THE
DEMOGRAPHIC VARIABLES
In order to determine the levels of importance of
particular demographic variables we perform a
correspondence analysis (see Hill, 1974; Benzecri,
1992). Correspondence analysis is a variant of
principal component analysis aimed primarily at
categorical data. The method allows analyzing the
data table (a contingency table) and leads to a kind
of visualization of the rows and columns of this table
in the form of a map. Then it allows interpreting
these distances and relative positions of the points
from the map. The analysis that usually precede the
three-step correspondence analysis algorithm is the
chi-square test that allows to verify what is the
relation (association) between the variables that
comprise the table, if any. In our approach, by using
the correspondence analysis we will measure the
level of association of a demographic variable with
the bargaining style and treat this measure as a level
of importance of the demographic variable. In other
words, the higher the association degree of the
demographic variable with the conflict resolution
style the more important the demographic variable
is. From the given case base we will compute the
association degree using the Yule’s phi-coefficient
(Φ) that derives from the chi-squared test. The
computation procedures is as follows.
In order to determine the Φ value we start with
the computation of contingency table as shown in
Table 4.
Table 4: The contingency table.
Categories
of variable X
Categories of variable Y
Sum of
rows
Y
1
Y
2
... Y
J
X
1
n
11
n
12
... n
1J
n
1.
X
2
n
21
n
22
... n
2J
n
2.
... ... ... ... ... ...
X
H
n
H
1
n
H
2
... n
H
J
n
H
.
Sum of
columns
n
.1
n
.2
... n
.J
n
In our particular application context the variable
X corresponds to one demographic variable (age, for
instance) and the variable Y corresponds to one of
the variables defining the bargaining style (for
instance the competing feature). The values n
ij
correspond to the frequencies of occurrence of cases
derived from the database, falling into the ith
category of the first variable and jth category of the
second variable. While comparing the variable age
(X) with the competing feature (Y) we need to
analyze the 6 by 12 matrix, since there are 6
predefined categories of age recorder by Inspire
system and there is 12 possible levels of each
bargaining feature defined by TKI. Similar matrices
we build for each combination of the demographic
characteristic and the bargaining style feature.
Based on each contingency table we compute the
value of chi-square metric:
∑∑
==
=
H
h
J
j
hj
hjhj
n
nn
11
2
2
ˆ
)
ˆ
(
χ
,
(6)
where
hj
n and
hj
n
ˆ
are the empirical and theoretical
frequencies of the contingency table, respectively.
The theoretical frequencies
hj
n
ˆ
are determined from
the following formula:
ICORES 2012 - 1st International Conference on Operations Research and Enterprise Systems
54
n
nn
n
jh
hj
..
ˆ
=
.
(7)
The chi-square metric will be used now for
determining the strength of relationship between the
analyzed variables. One of such a measures is Yule’s
phi-coefficient (there are also other similar
measures, such as the Pearson's Contingency
Coefficient, T-Czuprow's convergence coefficient or
V-Cramer's coefficient) computed as follows:
n
2
χ
=Φ .
(8)
As mentioned before we will treat the value of Φ
coefficient as a level of importance of the
demographic variable in the case-based reasoning
approach. For our particular Inspire’s case base we
obtain the values of phi-coefficients for pairs of
demographic variables and conflict bargaining
variables as shown in Table 5.
The last column of Table 5 is an average of the
Yule’s phi-coefficients for bargaining style variables
(it is normalized as well). As a result of this
computation we obtain levels of importance (w
i
) of
the demographic variables that may be used now to
initiate the weights in the formula (2) and conduct
the case-based reasoning for identification of the
negotiation profile of any new negotiator that wants
to use the negotiation support system like the
NegoManage one.
7 EXPERIMENTAL
EVALUATION
To evaluate the proposed approach we split the case
base into two parts. Our case base consists of 228
cases. The first part of the case base is the basis for
the reasoning process. The second part of the case
base consists of 25 cases. We use the second part of
the case base for testing the reasoning mechanism.
For each of the 25 cases in the second part of the
case base we estimate the conflict resolution style
description and compare it with the actual conflict
resolution style description. The comparison is done
by computing the distance between two vectors of
five conflict resolution style variables. As said
before the first vector describes the actual conflict
resolution style given in the testing part of the case
base, and the second vector describes the predicted
conflict resolution style. We perform the reasoning
on the level of similarity equal to 0.75. Table 6
shows the results of the reasoning.
As we can see in Table 2 the distances between
the actual and predicted outcomes are lower than the
max and min outcomes what indicates on good
performance of the CBR prediction mechanism. The
empty rows indicate that in the case base there was
no case that would be similar enough (on the level of
0.75) to the current situation
.
8 CONCLUSIONS
In this paper we proposed a novel approach for
Table 5: The Phi coefficients for different pairs of demographic variables and conflict resolution style variables.
Competing Collaborating Compromising Accomodating Avoiding Average
Normilized
(w
i
)
age 0.46 0.49 0.39 0.4 0.49 0.45 0.07
gender 0.22 0.15 0.23 0.23 0.17 0.2 0.03
c_reside 1.05 1.12 1.23 0.94 1 1.07 0.17
l_reside 0.54 0.44 0.5 0.42 0.33 0.45 0.07
c_born 1.4 1.7 1.31 1.27 1.22 1.38 0.23
english 0.5 0.52 0.48 0.51 0.53 0.51 0.08
course 0.35 0.45 0.34 0.28 0.29 0.34 0.06
ed_level 0.22 0.24 0.27 0.22 0.24 0.24 0.04
NSS 0.19 0.24 0.15 0.24 0.17 0.2 0.03
knowledge 0.5 0.44 0.48 0.46 0.47 0.47 0.08
p_before 0.18 0.25 0.14 0.14 0.21 0.18 0.03
ed_field 0.71 0.67 0.6 0.62 0.62 0.64 0.1
APPLYING CASE-BASED REASONING FOR IDENTIFYING THE NEGOTIATION PROFILE OF ELECTRONIC
NEGOTIATION SYSTEM USERS
55
identifying the negotiation profile, i.e. the bargaining
style of a negotiator. Instead of eliciting such a
profile by means of a psychometric test, like it is
usually done in many negotiation situations, in our
approach we decided to use the historical
information about negotiation processes stored in the
INSPIRE system case base and, based on the results
of some research on the behavioural aspects of
negotiations, to derive the bargaining profile of the
negotiator from the analysis of their personal
characteristics. The case base we operated with
contained the knowledge about the demographic
characteristics of a negotiator and its conflict
resolution style determined by means of TKI. Then,
based on the postulate that similar demographics of
the player yields similar conflict resolution style
(which is the fundamental to the case base reasoning
approach) we derive the estimation of a conflict
resolution style (a profile) of a new negotiator. To
support our case-based mechanism with all the data
required we have also implemented the elements of
the correspondence analysis, that allows finding the
links between the analyzed variables that are
described by means of the weak scales (in our
experiment some of the personal features had been
described by means of ordinal- or nominal-scale
variables).
We believe the approach we proposed is more
user-friendly to the negotiator than the classic TKI
test (or the similar ones), since it does not enforce
them to fill the tiresome and difficult psychometric
questionnaires, but - what we are aware of - it needs
to be verified and tested on the significantly big
sample of the negotiators. The future work will
Table 6: Experimental evaluation of the CBR mechanism.
Case
Actual conflict resolution style Predicted resolution style Comparison
Competing Collaborating Compromising Avoiding Accomodating Competing Collaborating Compromising Avoiding Accomodating distance maximal minimal
1 9 5 8 3 3 6 5 9 5 4 0.1166 0.4333 0.4
2 9 5 8 2 4
3 3 4 12 3 5 4 4 9 5 4 0.1666 0.4333 0.4
4 2 3 8 9 8 6 5 9 5 4 0.25 0.45 0.3166
5 5 4 12 2 5 5 5 9 5 4 0.1333 0.4333 0.4
6 6 7 6 7 4 8 4 9 3 3 0.2166 0.4166 0.25
7 10 5 9 3 3 4 5 10 4 5 0.1666 0.4166 0.25
8 2 8 8 3 9 5 5 7 6 7 0.2 0.4166 0.25
9 1 4 8 5 10 5 5 8 5 6 0.15 0.4 0.3
10 6 6 10 4 2 5 5 8 6 5 0.15 0.4166 0.4
11 1 3 8 9 7 5 5 8 6 5 0.1833 0.4 0.35
12 12 6 6 2 2 6 4 8 6 4 0.2833 0.3833 0.3333
13 6 2 8 7 5 6 5 8 6 4 0.0833 0.4 0.3333
14 11 4 10 2 1 8 6 7 4 4 0.2166 0.3833 0.2333
15 6 4 8 5 7 5 5 8 6 5 0.8333 0.4166 0.3166
16 11 5 7 2 3 5 5 9 5 5 0.2166 0.4333 0.3833
17 3 3 11 8 3 6 5 9 5 4 0.1833 0.4333 0.4
18 8 8 7 1 4 5 4 7 6 6 0.2333 0.3888 0.3666
19 2 6 8 5 7 5 5 8 6 5 0.1166 0.4 0.3666
20 5 3 9 3 8 5 5 8 6 5 0.15 0.4 0.3833
21 7 2 10 5 4 6 5 9 5 4 0.0833 0.4333 0.4
22 3 6 7 6 8 4 7 7 7 5 0.1 0.4166 0.1
23 7 3 12 4 2
24 12 5 5 5 3
25 2 7 9 3 7 7 6 9 4 2 0.2 0.3833 0.25
ICORES 2012 - 1st International Conference on Operations Research and Enterprise Systems
56
focus then on the implementing our approach in the
NegoManage web-based core simultaneously with
the electronic TK test, which will allow to verify
whether the profiling mechanism that we built based
on the estimates from the case-based reasoning and
the correspondence analysis, is reliable enough and
may be used for eliciting the true negotiation profile
(i.e. the profile that is concordant with the results of
the TKI) of the new electronic negotiation system
user. After implementing our approach we also plan
to verify the results on the alternative psychological
research on negotiation (see Elfenbain et al., 2008),
which may modify the way we define the situation
in our case-based approach. It may appear that some
other psychological characteristic may (or should)
also be included in our analysis (such as the attitude
towards the problem, partner and process) that
would lead to the better estimation of the
negotiator’s profile.
Finally, we are aware of other alternative
methods that may be used in negotiators’ profiling,
that base on the classical statistical
clustering/classification approaches or apply AI
solutions like neural networks, and which we
rejected in our preliminary selection since they
usually requires metric data. Thus our future work
will also focus on analyzing the possible extensions
and modifications of these rejected methods that
could be used alternatively to our method and then
on comparing the clustering results they would lead
to with the results of our model.
ACKNOWLEDGEMENTS
We thank Gregory Kersten and the InterNeg
Research Centre Team for supporting us with the
Inspire database. This research is supported by the
grant of Polish Ministry of Science and Higher
Education (N N111 362337).
REFERENCES
Adair, W. L. and Brett, J. M. (2004). Culture and
negotiation processes. In Gelfand, M. J. and Brett J.
M. (Eds.), The Handbook of Negotiation and Culture
(pp. 158-176). Stanford: Stanford Business Books.
Aamodt, A. and Plaza, E. (1994). Case-based reasoning:
Foundational issues, methodological variations, and
system approaches. AI Communications, 7:, 39-59.
Benzecri, J. P. (1992). Correspondence analysis
handbook, New York: Marcel Dekker.
Blake, R. R. and Mouton, J. S. (1964). The managerial
grid. Houston: Gulf Publications.
Briggs, I. and Myers, P. B. (1980). Gifts Differing:
Understanding Personality Type. Mountain View:
Davies-Black Publishing.
Brzostowski, J. and Wachowicz, T. (2009). Conceptual
Model of eNS For Supporting Preference Elicitation
and Counterpart Analysis. In Kilgour, D.M. and Wang
Q. (Eds.), Proceedings of GDN 2009: An International
Conference on Group Decision and Negotiation (pp.
182-186). Waterloo: Wilfried Laurier University.
Brzostowski, J. and Wachowicz, T. (2010). Building
personality profile of negotiator for electronic
negotiations. In Trzaskalik, T. and Wachowicz, T.
(Eds.), Multiple Criteria Decision Making '09 (pp. 31-
46). Katowice: The Publisher of The University of
Economics in Katowice.
Elfenbein, H. A., Curhan, J. R., Eisenkraft, N., Shirako, A.
and Baccaro, L. (2008). Are some negotiators better
than others? Individual differences in bargaining
outcomes. Journal of Research in Personality, 42,
1463-1475.
Gilboa, I. and Schmeidler, D. (1995). Case-based decision
theory. The Quarterly Journal of Economics, 110(3),
605-639.
Herrmann, G. M. (2004). Haggling Spoken Here: Gender,
Class, and Style in US Garage Sale Bargaining.
Journal of Popular Culture, 38(1), 55-81.
Hill, M. O. (1974). Correspondence analysis: a neglected
multivariate method. Applied Statistics, 23, 340-354.
Howison, J. (2003). An Introduction to the Literature on
Online Reputation Systems for the MMAPPS project.
Retrieved July 27, 2011, from http://james.howison.
name/pubs/RepLitIntro.pdf
Jehn, K. A., Chadwick, C. and Thatcher, S. (1997). To
agree or not to agree: The effects of value congruence,
individual demographic dissimilarity, and conflict on
workgroup outcomes, International Journal of
Conflict Management, 8(4): 287-305.
Kersten, G. E. and Noronha, S. J. (1999). WWW-Based
Negotiation Support: Design, Implementation And
Use. Decision Support Systems, 25, 135-154.
Kersten, G. E., Koeszegi, S. and Vetschera, R. (2003). The
effects of culture in computer-mediated negotiations.
Journal of Information Technology Theory and
Application, 5(2), 1-28.
Kilmann, R. and Thomas, K. W. (1983). The Thomas-
Kilmann Conflict Mode Instrument.
Cleveland: The
Organizational Development Institute.
Kray, L. J. and Thompson, L. (2005). Gender stereotypes
and negotiation performance: An examination of
theory and research. Research in Organizational
Behavior, 26, 103-182.
Kraybill, R. (2005). Style Matters: The Kraybill Conflict
Style Inventory. Harrisonburg: Riverhouse ePress.
Leake, D. B. (1996). Case-Based Reasoning: Experiences,
Lessons & Future Directions. Menlo Park: AAAI
Press/MIT Press.
Mershon, B. and Gorsuch, R. L. (1988). Number of factors
in the personality sphere: does increase in factors
increase predictability of real-life criteria?. Journal of
Personality and Social Psychology, 55, 675–680
APPLYING CASE-BASED REASONING FOR IDENTIFYING THE NEGOTIATION PROFILE OF ELECTRONIC
NEGOTIATION SYSTEM USERS
57
Paradis, N., Gettinger, J., Lai, H., Surboeck, M. and
Wachowicz, T. (2010). E-Negotiations via Inspire 2.0:
The System, Users, Management and Projects. In: de
Vreede, G.J. (Ed.), Group Decision and Negotiations
2010. Proceedings (pp. 155-159). Omaha: The Center
for Collaboration Science, University of Nebraska at
Omaha.
Patton, C. and Balakrishnan, P. V. (2010). The impact of
expectation of future negotiation interaction on
bargaining processes and outcomes. Journal of
Business Research, 63(8), 809-816.
Paunonen, S. V. and Ashton, M. S. (2001). Big Five
factors and facets and the prediction of behavior.
Journal of Personality & Social Psychology, 81, 524-
539
Searle, J. R. (1969). Speech Acts: An Essay in the
Philosophy of Language. Cambridge: Cambridge
University Press.
Shell, G. R. (2001). Bargaining Styles and Negotiation:
The Thomas-Kilmann Conflict Mode Instrument in
Negotiation Training. Negotiation Journal, 17(2), 155-
174.
Sternberg, R. J. and Dobson, D. M. (1987). Resolving
interpersonal conflicts: An analysis of stylistic
consistency. Journal of Personality and Social
Psychology, 52, 794-812.
Stiles, W. B. (1992). Describing talk: a taxonomy of
verbal response modes. SAGE Publications.
Thompson, L. (1990). Negotiation behaviour and
outcomes: Empirical evidence and theoretical issues.
Psychological Bulletin, 108, 515-532.
ICORES 2012 - 1st International Conference on Operations Research and Enterprise Systems
58