Anda di halaman 1dari 20

International Journal of Production Research,

Vol. 46, No. 5, 1 March 2008, 13151333


Using a computational model to compare objective negotiations
in real and virtual environments
B. KRAWCZYK-BRYLKAf and M. PIOTROWSKI*
fFaculty of Management and Economics, Gdansk University of Technology, Gdansk, Poland
Faculty of Electronics, Telecommunications and Informatics,
Gdansk University of Technology, Gdansk, Poland
(Revision received January 2007)
Business transactions and other human activities involve many negotiation
procedures. The most popular ones are objective negotiations where the
coordinated behaviour of participants is based on to the point arguments and
mutual understanding. In this paper the computational model of such
negotiations is presented. It takes into account many parameters describing
initial conditions, negotiation processes and final contract outcomes. One can
observe the behaviour of negotiators and create face-to-face (f 2 f ) and chat
negotiation dances. To support the monitoring of negotiation processes and data
collecting, a suitable test bed is proposed and implemented. Moreover the tool
allows the analysis of huge amounts of data and simulates many aspects
of negotiations. The obtained results allow recognising and capturing properly
many human activities based on negotiations in natural and Internet
environments.
Keywords: Negotiations; Negotiation dances; Negotiation characteristics;
Model, Quality attributes; Experiments
1. Introduction
Due to modern information technologies individuals, groups, organisations and
societies can communicate with each other and be active in different areas regardless
of time zones and geographic locations. As a result, more and more goods and
services are accessible with no need for face-to-face (f 2 f ) contact with other people.
Using Web technology (for instance service oriented architecture (SOA); Singh 2005)
we can combine services into logical components aligned directly with human
activity operations (Brown 2000). In business negotiations we can orchestrate such
services into many configurations (Maamar 2001) creating a network of services
supporting virtual environments, organisations and societies (Igbaria 1999).
At present, classical services and e-services are available for users. The latter take
advantage of virtual environments and technology advancements to stay competitive
and customer-focused in a turbulent business landscape. The simple models
of human service cooperation are: provider/customer (for business seller/buyer)
*Corresponding author. Email: bastian@eti.pg.gda.pl
International Journal of Production Research
ISSN 00207543 print/ISSN 1366588X online 2008 Taylor & Francis
http://www.tandf.co.uk/journals
DOI: 10.1080/00207540701224475
and provider/broker/customer (Rust 2003). In such models minimal communication
language includes: offering services, responding to the offers, negotiating and
invoking services. Technologies and available environments alone do not guarantee
the viability of the virtual society. Moreover, available technical means must be used
intelligently and deliberately. Therefore basic human activities should be deeply
recognised, completely understood and well organised.
As mentioned previously, most business transactions or other human activities
involve negotiation procedures (Krawczyk-Brylka 2006). Negotiation is the process
that occurs between at least two parties (negotiators) and concerns a certain subject
(buying goods/products, taking decisions, preparing solutions, executing services).
Since real and virtual societies coexist, f 2 f (in natural environment) and desk-to-
desk (d2d) (indirect one, i.e. supported by communication devices) negotiations are
used in parallel. The former occurs in natural communication environment, where
people meet and converse, the latter takes place when different telecommunication
means are used. The simplest example of d2d negotiations is negotiations via e-mail
or chat (Picard 2002). In the case of more advanced technologies people can create
agents, who seek out potential buyers or sellers and negotiate with them on behalf of
their owners (Mares 1999). In our opinion proper implementation of agent to agent
(a2a) or d2d negotiation strategies requires good understanding of f 2 f negotiations,
which makes transforming traditional business into e-business much easier. This
paper takes into account two kinds of negotiations (f 2 f and d2d) and focuses on
the so-called objective negotiation strategy. The objective negotiation means
coordinated behaviour of both participants (parties) based on to the point
arguments. The main principles of this style of negotiations are: separating people
from problems, concentrating on business itself not on the position, looking for
solutions useful for both sides and adopting objective criteria for evaluation of each
proposal (Fisher 1982). This style lies between competitive and cooperative
negotiations, where negotiators either try to achieve their goals to the maximum
degree or try to find a compromise which would be acceptable for both sides. The
success of such negotiations is determined by their results and a similar level of
satisfaction on both sides. Such negotiation strategies are more useful for providers
of services (they help establish close relations with customers) and for customers
(who are aware of conscientiously and honestly done transactions).
In the time of transition from traditional to electronic business (e-business),
companies should redesign their business processes in order to better satisfy
customer needs (Patrie 2004). For this reason profound knowledge of such processes
and market requirements should be acquired and deeply analysed (Peterson 1997).
Many researches based on specially defined questionnaires make a large-scale study
to access e-business drivers facilitating achieving high levels of operational
performance. These studies identify the critical links between e-business drivers,
financial indicators and operational excellence measures (Barua 2000). In addition,
several descriptive theories and models try to describe selling/buying processes
and improve negotiation procedures involved in such processes (Benenicke 2003).
In general, a buying/selling transaction comprises the following six fundamental
stages:
1. Identification of the user needs and recognition of the buyer motivations
for buying a product.
1316 B. Krawczyk-Brylka and M. Piotrowski
2. Product brokering and information retrieval for consideration of different
buying alternatives.
3. Merchant brokering and first choice of the best alternatives using such buyer-
provided criteria as prices, warranty periods, type of payments, delivery
options, service availability, producer reputation, product quality.
4. Direct negotiation of different aspects including the above criteria to make
selling/buying transaction acceptable for both sides.
5. Contract definition for purchase, delivery and maintenance to finalise the
negotiated transaction.
6. Past-purchase product life-phases and evaluation of user acceptation.
In practice, the above stages can overlap and migrate from one to another and
can be implemented in different ways (Nguyen 2004). One implementation technique
is filtering (steps 1, 2, 3) the attractive products by their features described on the
Web. Another one is collaborative filtering which tries to recommend products based
on various user options and ranking of their alternatives. Besides, data mining
techniques (Adomavicius 2001) can be used to discover the best patterns in customer
purchases and to apply these patterns to help decisions taking by buyers (steps 4,
5, 6). In many cases, agent-based technologies can be used to support all stages of the
buying process.
In general, the negotiation process can be described by a sixth tuple, as follows
(Krawczyk-Brylka 2006):
N = S, P, G, D, C, E)
where S is the subject of negotiation, as it was presented above; P is a set of parties
participating in negotiations; P={P
1
, P
2
, . . . , P
i
, . . . , P
I
}
The cardinality of P is at least two, i.e. I _2, and each P
i
(i =1, 2, . . . , I ) includes
at least one negotiator. The parties first verbalise contradictory demands and then
move towards agreement by searching for new alternatives.
G is a set of goals describing attributes of the subject being under negotiation,
i.e. G={g
1
, g
2
, . . . , g
j
, . . . , g
J
}, where g
j
( j =1, 2, . . . , J ) presents one attribute,
for instance in the case of selling/buying negotiations: product cost, delivery
time, warranty conditions, etc. can be considered. The concrete value of g
j
is
denoted by v
j
.
D is a sequence of demands/requests formulated step by step by the parties
during the negotiation as a modification of previous demands or a presenta-
tion of new proposals, which should be discussed and modified later.
D= (D
1
, D
2
, . . . , D
k
, . . . , D
K
), where K means the number of negotiations steps and
D
k
is a set of values of attributes: D
k
={v
1
k
, v
2
k
, . . . , v
j
k
, . . . , v
J
k
}. The set is a proposal
of attribute values given by parties in kth step of the negotiation. D
k
represents a
point in J-dimensional space and sequence D corresponds to the path of transitions
from point D
1
to the point D
K
. Such a graphical representation of negotiation steps
in J-dimensional space is called negotiation dance. For each party all values v
j
k
should satisfy the following acceptation condition:
min
k
j
_ v
k
j
_ max
k
j
for all j, j = 1, 2, . . . , J, k = 1, 2, . . . , K:
Using a computational model to compare objective negotiations 1317
C denotes the contract for the subject, when all parties accept the demand
formulated in kth step of the negotiation. Then D
K
is called the outcome of the
negotiation (contract values), i.e.:
NO = D
K
= {v
K
1
, v
K
2
, . . . , v
K
j
, . . . , v
K
J
] = {c
1
, c
2
, . . . , c
j
, . . . , c
J
]
E is an environment, where the negotiation is being conducted. In the case of natural
environment, f 2 f negotiation takes place. Using communication via the Internet, we
have chat (d2d) negotiation, and using agent technology we organise negotiation in
more automatic way (i.e. a2a negotiation or e-negotiation). E may play an essential
role in achieving the required level of negotiation quality and implementation of
e-business activities.
The organisation of this paper is as follows. Section 2 presents the computational
model of negotiations, where initial conditions, processes and the obtained contract
are described. Furthermore, main parameters of these items are introduced. Section 3
presents a test bed for experiments. In section 4 the results of experiments and
simulations are presented and discussed. These results show the way for under-
standing and improving some aspects of negotiation processes. The difference
between f 2 f and d2d negotiations is given and an impact of technology progress
on the negotiations is suggested. Finally, the summary and open problems are
formulated to stress the complexity of this kind of human activities.
2. Computational model of negotiations
Referring to formal definition of negotiation model presented in the Introduction,
one can assume that the computational model of negotiations (CMN) allows us to
present their main parameters: initial task conditions (IC) (defined before the start
of the negotiation process), different negotiation strategies (NS) or negotiation styles
and the analysis of the obtained negotiation outcomes (NO) given in the contract.
Then:
CMN = IC, NS, NO)
Without the loss of generality to describe IC, NS, NO graphically one can assume
that I =2 and J =1.
In figure 1, one can see some parameter metrics, responsible for description of
initial conditions. Let min
i
and max
i
be the minimum and the maximum values
of parameter g
i
expected by ith party (negotiator). Then
mv
i
= 0:5(min
i
max
i
) i = 1, 2
bv = 0:5(min
2
max
1
)
mv
1, 2
= 0:5(mv
1
mv
2
)
cv = v
k
1
Parameters cv, bv, mv
1,2
are strictly connected to the contract of the negotiation.
The concept of the model follows from our empirical experiments, performed for
many parties and many different goals of negotiations. To present a more detailed
1318 B. Krawczyk-Brylka and M. Piotrowski
description of the IC the following limitations and parameters are defined
(see figure 1):
min
1
5max
1
. min
2
5max
2
min
1
5min
2
. max
1
5max
2
min
2
_ bv _ max
1
mv
1
_ mv
1, 2
_ mv
2
As shown in figure 1, one can distinguish three intervals for value g
1
: the first one
acceptable only for P
1
(
1
), its distance being
1
, the second one only for P
2
(
2
)
and the third one for both P
1
and P
2
(
3
), i.e.:

1
= max
1
min
1

2
= max
2
min
2

3
= max
1
min
2
Proceeding from these basic metrics one can express distances between the balanced
value and the mean expected values of g
1
for each of negotiators P
1
and P
2
. They
are equal to
4
,
5
respectively. The distance between the balanced value and the
contract value is equal to
6
. Then:

4
= [bv mv
1
[

5
= [bv mv
2
[

6
= [bv cv[
In a similar way one can calculate distances between the mean values of the mean
participants expectations, and the balanced and contract outcome, i.e.:

7
= [mv
1,2
mv
1
[

8
= [mv
1,2
mv
2
[

9
= [mv
1,2
bv[

10
= [mv
1,2
cv[
Metrics
4

10
are quality metrics. Choice of
x
, x=1, 2, . . . , 10 allows us to define
different negotiation positions. Besides, one can define different negotiation
strategies: competitive, cooperative and objective (as competitive/cooperative
ones). As was demonstrated in the Introduction, the cooperative strategy focuses
bv
min
2
P
1
max
2
min
1
max
1
mv
1
mv
1,2
cv mv
2
P
2
Figure 1. A graphical representation of an initial negotiation condition. (Participants can be
aware or unaware of their initial conditions. In CMN it is assumed they may know the basic
parameters shown in figure 1.)
Using a computational model to compare objective negotiations 1319
on common effort to find the best satisfaction and the highest consistent outcomes
for both sides. Consequently, one can expect that outcomes belong to interval
min
2
, max
1
). In the case of competitive strategies, the main goal is to find the best
results for one side without any correspondence to the satisfaction of the second one.
Therefore, the obtained results are closer to either min
1
or max
2
. The objective
strategies try to find compromised result from the point of view of each negotiating
party and obtain compromised satisfaction for each side. Hence the results are closer
to either mv
1
or mv
2
. In order to achieve better satisfaction, the results focus on
mv
1,2
, which were confirmed by experiments. It is easy to note that if
1
=
2
then
bv =mv
1,2
. In many cases, initial conditions can have an impact on the outcomes but
using suitable negotiation strategies can change the final result.
Parties can formulate demands in different order and all demands can be
presented by all parties either at once or sequentially one by one. There are no rules,
in which order demands should be presented. Moreover, parties can remain either
passive (when they only accept or reject the demands made by others) or active
(when they can change the current proposals to make them more likely to be
accepted). Each single change of a demand means a new step of the negotiation.
If current demands are not acceptable for the other parties, these parties can return
to the previous proposal. Therefore, the track of some of the proposals (D
/
D)
should be kept and recovery mechanisms should support such return operations.
Taking into account the way of demand formulation and presentation, negotiation
strategies can also be competitive (the parties focus on the best outcomes for
themselves), objective (the parties are looking for a compromise) and collaborative
(the parties try to understand the reasoning of other sides). Negotiation strategies
used by one party can vary over time according to the assumed tactics and current
negotiator satisfaction and feelings. A graph of negotiation dance very well suggests
types of negotiation strategies used for contract completion.
Graphical representation of a negotiation dance for objective strategy is shown
in figure 2. It consists of nine steps, where the first steps are characterised by much
higher changes of attribute g
1
than the next steps. In other words, negotiators decide
to discuss more with each other when their proposals oscillate near balanced values
(bv, mv
1,2
). In the competitive strategy the amplitude of the changes at all steps
is similar to the ones obtained in the last phases of the cooperative strategy.
The negotiation dance and consequently the quality of the negotiation contract can
be described by the following set of parameters:
K The number of steps (K), where:

11
Mean value of the change of the proposal for one step of the negotiation.

12
Minimal value of the change of the proposal for one step of the
negotiation.

13
Maximal value of the change of the proposal for one step of the
negotiation.
For I42 and J41 the number of parameters defined above and their dimension
increases so dramatically that their analysis is not possible without a special
program. Therefore, the proposed model is called the computational model.
The contract value can be described by cv, bv, mv
1,2
and its quality parameters
such as
6
,
10
. These parameters are suitable for all negotiation strategies.
1320 B. Krawczyk-Brylka and M. Piotrowski
Additionally, two other parameters seem to be suitable for evaluation of competitive
strategies:

14
= [cv min
1
[

15
= [max
2
cv[
and two others for objective strategies:

16
= [cv mv
1
[

17
= [mv
2
cv[
Beside the quality of contract (negotiation results) can also be evaluated by means
of some parameters connected to the sets shown in figure 2, where three squares are
defined and pairs of vertices located on the contract line ( y =x, value (P
1
) =value
(P
1
)) are shown. The first square contains such points: A(min
1
, min
1
) and B(max
2
,
max
2
) and it defines the set of possible proposals (SPP), which may occur
(be proposed by both parties) during the negotiation processes. The second square
is determined by the following points C(min
2
, min
2
) and D(max
1
, max
1
) and is called
min
1
min
1
min
2
min
2 max
1
max
1
max
2
max
2
mv
1,2
mv
1,2
mv
1
mv
1
mv
2
mv
2
bv
bv
cv
cv
D
1
D
2
D
3
D
5
D
7 E
SBP
D
F
CL
B
SPP
P
1
P
2
SRP
D
8
D
4
D
6
D
9
Figure 2. Graphical representation of a negotiation dance, presenting a sequence of demands
D=(D
1
, D
2
, . . . , D
K
) for objective strategy. The notations are as follows: SPP, set of possible
proposals; SBP, set of balanced proposals; SRP, set of reasonable proposals; CL, contract line.
Using a computational model to compare objective negotiations 1321
the set of balanced proposals (SBP). The third square is defined by points
E(mv
1
, mv
1
) and F(mv
2
, mv
2
) and is called the set of reasonable proposals (SRP).
It is obvious that SRPCSBPSPP. We can calculate the length of contract line [IC[
included in these sets which is expressed in the following formula:
|XCL[=[CLSXP[, where X=either P, B or R. Therefore:
PCL [ [ =

2
_
max
2
min
1
( ) =

2
_

1

2

3
( )
BCL [ [ =

2
_

3
RCL [ [ =

2
_
mv
2
mv
1
( ) =

2
_

8

7
( )
Following this formula one can also classify the outcomes into three categories:
impossible (belonging to the line interval PCL BCL RCL), symmetrical
(belonging to BCL RCL) and reasonable (belonging to RCL).
3. Test bed for experiments executions
Understanding human activities such as negotiation processes is a big challenge even
despite the technology progress. The main problems, as mentioned in section 2, are
the complexity of human decision process, difficulties of taking into account various
aspects of negotiations, and the impossibility of considering numerous existing
alternatives. To eliminate some of these difficulties the GAJA system were designed
and implemented to support performing the experiments (Piotrowski 2006).
In particular this system supports the design and execution phases of negotiation
experiments as well as data collection and analysis. The main idea of such a toolset
is presented in figure 3. Its architecture consists of the following layers:
. Graphical interface and computer network-Internet based environment for
contacts and interactions of all toolset users and simulation of the chosen
scenarios.
. Management and monitoring tools for registration of the required
information about negotiation processes, occurring events and human
decisions.
. Database for storing the collected data and some procedures for the results
analysis compatible with external tools.
. Control infrastructure (administrator functions library) to manage the
toolset and to change its configuration.
Each negotiation experiment means one negotiation process with a pair of
negotiators. To execute certain experiments one should define the suitable scenarios,
the execution of which can be supported by the functionality of the GAJA system.
The types of functionality are as follows:
. Type 1. For defining various kinds of human activities (tasks, e.g.
customer/provider negotiations, object ranking, brain storming, enterprise
coordination).
. Type 2. For defining a set of questionnaires, measurement procedures
and data necessary for evaluation of participants personality, effectiveness
of negotiation strategies, acceptability of negotiation outcomes, etc.
1322 B. Krawczyk-Brylka and M. Piotrowski
. Type 3. For defining simulation and data mining methods to discover some
rules hidden in the collected data and to compare certain results of human
activities, e.g. negotiations taking place in the real and virtual environment.
Three specific procedures are presented below: one, which is rather more general,
and two others, which are more specialised. They are strictly connected to the
observation and analysis of negotiation experiments. Certain administrative
procedures which allow defining the suitable environment for experiments have
also been defined. However, the latter are not described in this paper.
The experimental procedure consists of 10 steps:
1. Define an experiment and make a description of the initial conditions for the
experiment participants.
2. Examine the negotiators (experiment participants) personalities, experiences
and motivation by means of specially prepared questionnaires.
3. Inform each of the negotiators separately about the aim of negotiation,
the role of each negotiator and the alternative solutions (if they exist).
4. Begin the negotiation in a given environment (f 2 f, d2d) or simulate the
negotiation dance for suitable parameters.
5. Monitor and register the negotiation process by camera and system logs
(including chat log for d2d negotiations) and store values of suitable
parameters describing the process.
scientists
experts
task definition
initial assumption
scenario execution
Graphical interface and communication
environment
demands
answers
pause
role
(customer/
buyer)
personality
experience
position
role
(provider/
seller)
personality
experience
position
nagotiators
status
nagotiation
dance data
process
status
control data
(steps)
continue
write
read
Database and
evaluation process
stop
start
demands
answers
calculate
pattern scenario
and contracts
negotiators
self-evaluation
expert
remarks
strategies
Real and simulated
negotiated process
assumptions
Management and monitoring
negotiation process
P
1
:
:P
1
administrator
tool integration
component generation
access control
commands and macro
contract
cinditions
Figure 3. Test bed for supporting execution, monitoring and analysis of negotiation
experiments.
Using a computational model to compare objective negotiations 1323
6. Stop the negotiation if either the consensus was reached or the time finished.
7. Let both the negotiators and experts (if they observe the process) partially
evaluate the recurring processes by.
8. Collect evaluation questionnaires in the system database for further analysis
and data mining, in figure 4 there is a screenshot of the data analysis and
export module.
9. Design procedures and plan new experiments to continue research on given
problems.
10. Use external tools (e.g. Statistica) and present their results in a graphical
mode to visualise relationships and general rules.
Other very useful procedures are simulation and evaluation. The simulation pro-
cedure allows us to generate negotiation dances for specific initial conditions, different
behaviour of negotiators and to evaluate some negotiation cases, described by certain
values of parameters defined in section 2. This procedure is defined as follows:
Simulation procedure
1. Consider some statistical characteristics of human behaviour collected in
a real negotiation process and put them as simulation control driven
parameters.
2. Use control driven parameters that generate sequences of D for different
aims, initial conditions, and different behaviour of participants.
Figure 4. Example screenshot of the GAJA system, showing some gathered results from
negotiation experiments.
1324 B. Krawczyk-Brylka and M. Piotrowski
3. Collect simulation results in order to verify some drivers discovered in real
experiments (use the evaluation procedure).
4. Evaluate correctness and exactness of the collected data of the process under
assessment.
Having the set of results obtained during real and simulation experiments,
one can analyse the negotiation process for different negotiation goals, different
initial conditions, and different negotiation strategies. To evaluate the experiments
there is a new generalised procedure, called evaluation procedure, as follows:
Evaluation procedure
1. Use e.g. Statistica tool to find general relationships for a set of collected data.
2. Distinguish sets of data having values of parameters satisfying the given
criteria (e.g. participants with high and low positions, with large and small
negotiation experience, etc.).
3. Make deeper statistical analysis (including data mining) for the selected sets
of data (correlations factor analysis, comparison analysis).
4. Check some results using simulation procedures.
5. Plan new monitoring procedures to continue observation of certain tendencies
or relationships.
Simulation and evaluation procedures are related to each other. Such an approach
allows us to adopt a well known method of analysis known as hit and miss method.
Table 1 shows the main quality attributes and metrics taken into consideration in
this paper. To present quantitative metrics we consider selling/buying negotiations
for I =2 with one negotiator in each party, and J =5.
All the above formulas can be used for analysing different kinds of negotiations
including f 2 f and d2d ones. If our experiments consist of many rounds of
negotiations, the completeness and satisfaction can be expressed as either the
distribution of values of such metrics or as their mean values.
Table 1. Main quality attributes and metrics of negotiations.
Quality attributes Quality metrics
Parties characteristics Personality (personal feature of the negotiator measured
by professional sociology and psychology tests,
e.g. NEO PI-R).
Experience (low, medium and high, in business activities).
Role of participants (seller, buyer).
Position in negotiation (availability of alternative offers).
Communication channels f 2f (natural), d2d (virtual).
Negotiation strategies (styles) Competitive, objective, collaborative.
Outcome metrics Effectivenessthe percentage of negotiations with complete
contracts in the considered experiment.
Performancedistribution of negotiation times for complete
contract in the considered experiment.
Completenessthe percentage of goals taken into
consideration for each negotiation in the considered
experiment.
Satisfactionthe relative difference (%) between the
obtained outcome and the expected outcomes.
Using a computational model to compare objective negotiations 1325
Satisfaction of the negotiation related to the contract value (cv) is evaluated
by each negotiator separately. We distinguish three kinds of satisfaction correlated
to attributes bv(sat_bv
x
), mv(sat_mv
x
) and subjective satisfaction determined
by the questionnaire filled by participants (sat_quest
x
). They can be expressed by
the following formulas:
sat bv
x
=
0% if cv4max
x
1

6
0:5
3

100% if cv c min
x
, max
x
)
100% if cv5min
x
8
>
>
>
>
<
>
>
>
>
:
sat mv
x
=
0% if cv4max
x
1

16
max
x
mv
x
[ [

100% if cv c mv
x
, max
x
)
100% if cv5mv
x
:
8
>
>
>
>
<
>
>
>
>
:
However, the above formula should be adjusted according to the context of g and
additionally,
16
is suitable only for customer/buyer role and it should be changed
to
17
for provider/seller role. These formulas are normalised to the existing
outcome areas.
4. Obtained results
With use of the well defined procedures and the designed and implemented test bed
presented in section 3 many negotiation experiments were executed and all quality
metrics described in table 1 were analysed. The basic items of experiments are shown
in table 2.
All negotiators were randomly divided into two groups: sellers and buyers.
Then, their personality was tested with psychological tests (see table 1). Pairs of
negotiators (seller and buyer) were randomly assigned to one of two environments
(f 2 f and d2dchat). Next, each pair of negotiators received the same description
of the negotiation task, where their roles, position and initial conditions were given
(see table 2).
The presented results correspond to the group of 400 participants. The
distribution of their profession and professional experience is given in figure 5.
Among the participants of the experiment, students were the largest group 37%.
Office workers constitute 31%, technical staff 17%, tradesmen 5%, managers 5%.
Other subjects of study were manual workers and freelancers (altogether 5%).
The biggest group (37%) has been working in business for 1 to 5 years. The next
group (25%) has less than 1 years experience, which means that respondents were
students or last years graduates of our university. It was found that the time of
negotiations is nearly twice as short for negotiators with 10-years experience
compared with the time of negotiations for negotiators with no or little experience.
The negotiation process, satisfaction, motivation level, and subjective negotiation
position were assessed by the experiment participants as well as external experts.
1326 B. Krawczyk-Brylka and M. Piotrowski
The aim of this kind of double evaluation was to achieve more objective results.
It can be observed that the subjects of the study overvalued their satisfaction level,
i.e. they gave more positive assessment of themselves than the experts. The latter,
however, gave higher evaluation of the meetings fluency (exchange of negotiation
turns, communicative flexibility of the participants) and the participants motivation.
Moreover, it was found that the negotiation outcome depends more on the subjective
impression of its participants than on the factual progress of the meeting.
Business-like style discussions dominated the negotiation process, though there
were participants who used a competitive style, which led to breaking the negotiation
or reaching high satisfaction at the expense of another negotiator. The results of the
Table 2. Basic items of experiments.
Items Descriptions
Negotiators 200 pairs of negotiators, 50% female, 50% male, I =2. 30%
for f 2 f negotiations, 70% for d2d.
Roles Seller, buyer (given in advance in description of negotiation
tasks).
Positions Either equal or biased for each pair of negotiators, where
equal position means that both negotiators have no
alternative offer available and biased position means that
one of the participants has alternative business offers.
Subjects Selling/buying or achieving access to medical services.
Negotiation of medical equipment having specific
functionality and quality.
Attributes considered
under negotiations
Price, warranty period, bank credit duration, delivery time,
tutorial cost, J =5.
Evaluations forms Special questionnaires to be filled by experts or negotiators
to calculate the set of parameters
i
, i =1, . . . , 17.
Used procedures Experimental procedure, simulation procedure and evaluation
procedure (see section 3).
Initial conditions Defined by type of experiments using parameters shown
in figures 1 and 2 (min
1
, max
1
, min
2
, max
2
) when
055|bv mv
1,2
|
initial conditions are named extremum
conditions.
None
12%
1 year
25%
15 years
37%
610 years
16%
>10 years
10% Students
37%
Office workers
31%
Technical staff
17%
Tradesmen
5%
Managers
5%
Freelancers
and manual
workers
5%
(a) (b)
Figure 5. (a) Profession of negotiators. (b) Professional experience of negotiators.
Using a computational model to compare objective negotiations 1327
negotiations were also compared for two groups of students with different
educational profile: IT and economic. While more conscious of the positive aspects
of the win-win strategy, the students of the Faculty of Management and Economy
seemed to be less effective in the situation of a biased position. The computer science
students, who had a more analytical approach to the task, and were more motivated
by their winning situation, finally got better results.
While defining the personalities of the group under study, a high level of diligence
and extraversion was observed. The first quality implies serious engagement in a
deeply analytical decision-making process. It also entails high motivation and acting
according to plan. The second quality depicts an optimistic and active approach to
the surrounding reality, as well as positive and pro-social emotionally. Such
personality traits as liveliness, openness to new experiences, agreeability and
perseverance were at the average level (46 stens). The lowest results were attributed
to such characteristics as neuroticism and emotional reactivity.
To estimate the overall effectiveness of the negotiation it was important to
determine the weight of the particular negotiation elements. Thus it follows that
the price contributes to negotiation effect in 42%, the date of delivery 15%, the date
of payment 18%, the warranty conditions 15%, training costs 10%. It was found in
practice that the negotiators made accurate evaluation of the negotiation situations
by giving priority to the price factor. The original model of negotiation assessment
has been proved to be effective in practical usage and therefore it can be employed
as the universal model to assess various complex human undertakings.
The extreme situation assumed three possible conditions of the negotiators
position, i.e. the balance between the buyer and the seller or one of the parties
is winning. Each experimental group consisted of one third of all the subjects.
The observer (the external expert) assessed the motivation of the group members
as high8, 3 (at the 19 scale). However, the subjective perception of the subjects
self-identification with the adopted roles was 7, 4. Hence, there came a conclusion
that the experiment fulfilled all the conditions of a case study.
On the whole, eight pairs of participants broke the negotiations (one from f 2 f
negotiations and the others from d2d negotiations). That was not due to personality
characteristics but the negotiation style: aggressive attacking of the interlocutor,
criticising his proposals, threatening to break the negotiations.
The negotiation time was measured for each of the experiments. The distribution
of negotiation time is shown in figure 6. While comparing the d2d and f 2 f
negotiations, one can notice that the negotiation time is nearly three times longer
for the d2d negotiations. This refers strictly to the negotiation process. Time
necessary for preparation towards the meeting is not taken into consideration.
The satisfaction of the negotiators approaches 50%, which indicates that people
tend to compromise and share the benefits of the negotiations. The satisfaction of the
buyers is higher than that of the sellers; similarly the satisfaction of the men is higher
than the women (figure 7). Satisfaction depends mainly on negotiation outcomes.
Additionally the satisfaction is determined by these six important factors: personal
skills, experience of the negotiator, personal engagement, adaptability, persons own
perception of the negotiation process and communication channel.
Comparing the f 2 f and d2d negotiations one can observe that the style of
d2d negotiations has more often aggressive elements (figure 8) than f 2 f negotia-
tions and the negotiators satisfaction is lower for d2d negotiations than for f 2 f
1328 B. Krawczyk-Brylka and M. Piotrowski
Number of
respondents
(in %)
Negotiation time
(in minutes)
0
5
10
15
20
25
30
35
40
45
5 15 25 35 45 55 65
f2f
d2d
Figure 6. Distribution of negotiation time (performance) for f 2 f and chat channels.
0
5
10
15
20
25
30
35
40
0 10 20 30 40 50 60 70 80 90 100
The number
of respondents
(in %)
Satisfaction Broken
negotiations
0
5
10
15
20
25
30
35
40
men women
Figure 7. Distribution of satisfaction for men and women.
f2f
d2d
Number of
respondents
(in %)
competitive/objective objective/cooperative
0
10
20
30
40
50
60
70
competitive objective cooperative
Figure 8. Popularity of negotiation strategies in real and virtual environments in the
considered experiments.
Using a computational model to compare objective negotiations 1329
negotiations (figure 9). Personality is the most important factor determining the
outcome of the Internet negotiations, whereas in f 2 f the most significant role is
played by the subjective perception of the negotiation process (friendly atmosphere).
Besides the empirical experiments, many simulations were also performed in
order to verify observed tendencies (see the simulation procedure in section 3).
However, the simulation procedure requires some statistical characteristics of human
behaviour, which are used as control parameters of the simulation. Referring
to the two-dimensional space of negotiations presented in figure 2, analysis of 200
negotiation experiments shows that 71% of the outcomes belongs to RCL,
20% to BCL, and 9% to PCL. Moreover, the distribution of outcomes values can
be described by the Gauss distribution with the mean value equal to either bv or
mv
1,2
and dispersion equal to either 6 =[BCL[ or 6 =[RCL[ respectively.
Additionally we determine the distribution of the outcomes. Figure 10 shows
how this distribution changes according to the level of experience (low, high).
Satisfaction
(in %)
Number of
respondents
(in %)
0
5
10
15
20
25
10 20 30 40 50 60 70 80 90 100
f2f
d2d
Figure 9. Distribution of satisfaction for f 2 f and chat negotiations.
The number
of outcomes
Low experience
High experience
Outcome values
45
35
20
10
0 10 20 30 40 50 60 70 80 90
Figure 10. Estimated curve shapes for outcome distributions coresponding to low and high
negotiators experiences.
1330 B. Krawczyk-Brylka and M. Piotrowski
For low experience of negotiators the curve is less dispersed than for high experience.
Similar tendency for satisfaction can be observed.
Some of the parameters taken into consideration in the simulation are presented
in tables 35. The impact of negotiators profession for d2d negotiations is shown
in table 3. Table 4 shows the impact of experience expressed as a number of
participations in the same kind of negotiations with some of the parameters different.
Correlation between negotiators satisfaction and parameters
i
, i =1, 2, . . . , 17
are shown in table 5. The contract value depends on initial conditions described
by bv, min
i
, max
i
, mv
i
(i =1, 2) and mv
1,2
. Satisfaction described by negotiators is
significantly correlated with satisfaction defined on the basis of bv. Moreover, the
latter is also related to cv, but for sat_quest and sat_mv the relation is not evident.
It seems that the mean value of cv (cv) is closer to mean value of bv (bv) when
negotiators have less experience, and are closer to the mean value of mv
1,2
(mv
1, 2
)
otherwise.
Simulation procedure was used to examine how initial conditions impact the
contract values (tables 6 and 7). One can note that our simulation model requires
further improvements for extremum initial conditions. For instance, comparing
experimental results shown in tables 6 and 7 it is easy to note that the obtained
contract values are more sensitive to changing the parameter min
2
than max
2
.
Table 5. Correlation coefficients for
i
, i =1, 2, . . . , 17 and negotiations outcomes.
Negotiation outcome Correlated parameters
cv sat_bv (0.47);
6
(1.0);
10
(0.79);
14
(0.92);
15
(0.89);

16
(0.92);
17
(0.57)
sat_bv cv (0.47); sat_quest (0.36);
6
(0.47);
10
(0.51);
14
(0.46);

15
(0.45);
16
(0.46);
17
(0.32)
sat_mv
2
(0.6);
3
(0.6);
4
(0.6);
5
(0.6);
7
(0.6);
8
(0.6);

9
(0.6);
14
(0.37);
15
(0.4);
16
(0.37)
sat_quest sat_bv(0.36); sat_mv(0.13)
Table 3. Impact of negotiators profession on outcome values.
Computer scientists Economists
Profession Equal position Biased position Equal position Biased position
The mean values of cv 54.390 52.71 55.300 47.33
bv 54.250 45.00 54.250 45.00
mv
1,2
47.375 45.00 47.375 45.00
Table 4. Impact of the negotiators experiences on outcome values.
Experience
Lowparticipation in an
experiment one time only
Highparticipation in
similar experiments many times
The mean values of cv 46.38 52.71
bv 45.00 45.00
mv
1,2
50.00 50.00
Using a computational model to compare objective negotiations 1331
The increase of min
2
by two times leads to breaking the negotiations, whereas max
2
should increase more than four times for the negotiations to be broken.
5. Conclusions
This paper presents a quantitative model of negotiations. This model allows
evaluating the quality of negotiations in different environments and comparing
these environments in order to point out the most important drawbacks. Four
quality attributes are defined and analysed. It was shown that the effectiveness
and completeness are strictly related to negotiators personality and experience.
The performance and satisfaction depend on the personality of negotiators and their
positions, evidently in chat negotiations. Negotiation strategies also impact the
performance and satisfaction.
It was shown that the communication channel has impact mainly on the
performance of negotiation process. This means that modern communication
technology needs further improvements to increase negotiation quality. To increase
the completeness, effectiveness, performance and satisfaction of negotiations more
intelligent and flexible tools are required. Such tools should be able to check, control
and predict the behaviour of negotiators and register, analyse and predict the
negotiation dances. Then it will be possible to control and increase the negotiation
quality and in consequence to improve the quality of e-business transactions.
This paper presents the computational model of objective negotiations described
as initial conditions, negotiation dances and final contracts, and proposes the set of
metrics which thoroughly describe the negotiation process and have essential
impact on the outcomes. Due to this approach, the simulation of negotiation
processes is realisable with simulation results acceptable for typical initial conditions
(bv -mv
1,2
). The open problem is how to describe the behaviour of negotiators
for extremum conditions, e.g. when the negotiators position is different.
Table 7. Impact of max
2
on the contract parameters when min
1
, min
2
, and max
1
=const.
min
2
30 40 50 60 70
bv 45.00 50.00 55.00 60.00 65.00
mv
1,2
45.00 47.50 50.00 52.50 55.00
The mean values of Simulated cv 45.94 48.45 49.16 52.03 54.18
Experimental cv 52.69 53.71 56.83 No contract No contract
Table 6. Impact of max
2
on the contract values when min
1
, min
2
, and max
1
=const.
max
2
70 130 190 290
bv 45.00 45.00 45.00 45.00
mv
1,2
45.00 60.00 75.00 92.50
The mean values of Simulated cv 45.94 59.58 71.57 105.88
Experimental cv 52.69 54.14 56.24 No contract
1332 B. Krawczyk-Brylka and M. Piotrowski
A very interesting result of the experiments is that among the three main areas
of negotiators possible proposals of attributes considered under negotiations
(see figure 2), the SRP area, which is the smallest one, comprises more than 70%
of all proposals. The same remarks refer to the contract line, where RCL collects the
majority of outcome values. In other words, objective strategies even in extremum
conditions (bv ,=mv
1,2
) lead to the situation, where the contract values belong
to RCL. In our opinion SRP and RCL are really representative items for objective
negotiations.
The open problem is how to achieve better similarity between the simulated
negotiation dances and the ones registered during real experiments.
Acknowledgement
This work was supported under national MNiI grant no. N516035 31/3499.
References
Adomavicius, G. and Tuzhilin, A., Using data mining methods to build customer profiles.
IEEE Comput., 2001, February, 7482.
Barua, A., Konana, P., Whinston, A. and Yin, F., Making e-business pay: eight key drivers
for operational success. IT Prof., 2000, November/December, 2230.
Benenicke, M. and Langendorfer, P., Towards automatic negotiation of privacy contracts for
Internet services, in Proceedings of the 11th International Conference on Networks, 2003,
pp. 4954.
Brown, R., Dynamics Within and Between Groups, 2000 (Blackwell Publishing Ltd.: Oxford).
Fisher, R., The pragmatic perspective of human communication. A view from system
theory. In Human Communications Theory, edited by F. Dance, 1982 (Harper and
Row: New York, NY).
Igbaria, M., The driving forces in the virtual society. Comm. ACM, 1999, 42(12), 5470.
Krawczyk-Brylka, B. and Piotrowski, M., Analysis of human negotiations for e-business
improvements, in PRO-VE2006 Proceedings, Helsinki, 2006, pp. 429438.
Maamar, Z., Dorion, E. and Daigle, C., Toward virtual marketplaces for e-commerce
support. Comm. ACM, 2001, 44(12), 3539.
Mares, P., Guttman, R.H. and Moukas, A.G., Agents that buy and sell. Comm. ACM, 1999,
March, 8184.
Nguyen, T.D. and Jennings, N.R., Coordinating multiple concurrent negotiations.
AAHAS04, 2004 (New York, USA).
Patrie, C. and Sahai, A., Business processes on the Web. IEEE Inter. Comput., 2004, January/
February, 2829.
Peterson, E. and Thompson, L., Negotiation teamwork: the impact of information
distribution and accountability on performance depends on the relationship among
team members. Org., Behav. Human Dec. Proc., 1997, 72(3), 364383.
Picard, W., Multi-facet analysis of e-negotiations. PhD thesis. Ecole Nationale Superieure des
Telecommunications, Paris, 2002.
Piotrowski, M. and Krawczyk-Brylka B., The web based system for recording and analysing
different kinds of negotiations, in WEBIST 2006, Proceedings of the Second
International Conference on Web Information Systems and Technologies, Society,
e-Business and e-Government/e-Learning, Portugal, 2006.
Rust, R.T. and Kannan, P.K., E-service: a new paradigm for business in the electronic
environment. Comm. ACM, 2003, 46(6), 3642, June.
Singh, M.P. and Huhns, M.N., Service Oriented Computing. Semantics, Processes, Agents,
2005 (Wiley: New York).
Using a computational model to compare objective negotiations 1333

Anda mungkin juga menyukai