NEW MECHANISM DESIGN IN THE C2C ONLINE
REPUTATION EVALUATION OPTIMIZING
Yi Yang, Wenjun Zhu and Xiliang Zhang
School of Ecomomics and Management, Xi'an University of Technology, Yanxiang Road, Xi'an, P.R. China
Keywords: Online trade, C2C, Reputation evaluation mechanism, Mechanism, Design.
Abstract: In view of the assumption of the trader’s bounded rationality, our research analyzed the defects of current
reputation evaluation mechanism and the trust problem in the online trade. The mechanism was redesigned
from the three aspects of the evaluation process, the accounting method of the reputation-limit and
reputation rating scores, then, the system of "tell the truth" has been added to the mechanism to realize the
improvement of incentive function in evaluation mechanism. In order to prove the effectiveness of the new
mechanism, an algorithm example was given based on the sequential game analysis and Harsanyi
transformation of the bounded rationality trader’s decision-making. The results show that the new
mechanism can effectively provide the decision-making information for the trading parties in the process of
reputation evaluation, and encourage both parties of evaluation to select the "tell the truth" strategy
achieving the maximum reputation score.
1 INTRODUCTION
Shorter distance and lower store costs have made the
online trading change the traditional sales channels
and consumption patterns, while, a unique
"credibility crisis" caused by the market liquidity
and transaction anonymity is increasing, become an
obstacle to the healthy development of the online
trading (Jennifer, 2006). Online Reputation
Evaluation Mechanism (OREM, also known as
online reputation evaluation system) came into being,
in order to relive the "reputation crisis", enhance
trust and other demands for traders.
Kim et al (2003)
discussed the problem of increasing consumer
confidence, believes the key of trust should be start
with personal information, product quality and price,
and through the 10 known websites have proven the
suitability, built the foundation of further empirical
research. Anyone can easily enter or leave, change
the identity in the C2C website that not only affect
the dealer’s trust, but also affect the continuation of
online trading market that led the OREM become an
important study issue; Yamamoto et al (2004) gave
important suggestions to improve the confidence
through the reputation evaluation using computer
simulation.
Mikhail et al (2002) thought the seller's
reputation that can help online auction bidders to
determine the quality through statistics of gold coin
auction; This proved "favourable" reputation has
positive impact to the seller from one side.
Jeffrey
(2005) confirmed fatherly the seller's reputation and
its marginal revenue appears inverse relationship
based on the Mikhail’s study.
Kamins et al. (2004)
analyzed the effects on the closing price from the
interaction between the starting price and the seller’s
reputation, the results show that asymmetric
information will increase the benefits of high
credibility of the seller, while there was no
significant relationship between the credibility and
the seller proceeds within a similar amount of
information. Cabral et al (2004) obtained that when
the seller receives the first negative evaluation, the
sales and selling prices will drop under the eBay's
reputation mechanism, subsequently increased in the
rate of negative feedback , the poor record seller
may withdraw from (and may re-enter under a new
identity), while the good record seller will get more
and better trading opportunities, yet, Dahui et al
(2004) analyzed the negative reputation on the
impact of price and sales transactions, made an
empirical study based on the data collected from
eBay, and worked out the seller's life-long negative
points, concluded the risk of the negative reputation
was not big as Cabral expected. Dellarocas et al
(2001, 2006a, 2006b, 2006c) proposed that the
OREM has become a promising mechanism for trust
53
Yang Y., Zhu W. and Zhang X..
NEW MECHANISM DESIGN IN THE C2C ONLINE REPUTATION EVALUATION OPTIMIZING.
DOI: 10.5220/0003486800530062
In Proceedings of the 13th International Conference on Enterprise Information Systems (ICEIS-2011), pages 53-62
ISBN: 978-989-8425-56-0
Copyright
c
2011 SCITEPRESS (Science and Technology Publications, Lda.)
management, utilizing computer science, marketing,
psychology and other field knowledge, analyzed
how to promote the construction of credibility
evaluation system, and discussed the reliability of
the evaluation mechanism from the number of
Buyers and sellers, the property of the participants,
Market microstructure, anonymous and certification
system evaluation, the future development of online
reputation evaluation depends on the reliability of
evaluation results.
In fact, the accused has not stopped on the
OREM, mutual abusive messages can be seen
anywhere on the web, mapping out there are flaws in
the reputation of the mechanism in some extent.
Chris et al. (2004) obtained a negative evaluation
will lead undesirable consequences for their
development through the empirical analysis and
pointed out the contradiction between the reputation
with the dissatisfaction about the transaction and
explain the causes of conflict, but not deeply
analyzed the inadequate of OREM. Christina et al
(2008) analyzed the seller change their strategy
through the generation of expected return and the
results of dishonest conduct under an effective
reputation mechanism, and the effectiveness of the
percentage of 12-month evaluation results in eBay,
the final design of the mechanism should start from
raising the level of parameters, but lack of analysis
of the behaviour of traders.
Any design and optimization of mechanism are
complex systematic thinking process, not only
analysis from the objective, but also sort out the
characteristics of the participants. Some traders In
the virtual web where more complex than the
actuality due to difficult to "face to face"
communicate may make non-rational behaviour (Yi
Yang, 2009), therefore, the OREM design is to
bound the irrational behaviour of traders and allow
them make rationally decisions. However, in reality
many people are not entirely rational "economic
man" (Qing Wang, 2009), after Simon (1955)
developed the concept of bounded rationality and
made a satisfactory criteria, the bounded rational
"social man" are gradually replaced fully rational
"economic man" either in theory or in practical
applications, for it is more close to reality. On et al.
(2002) found in the virtual network environment,
because they do not by any constraints, human
reason is very limited. The secret identity of online
transactions, a single expression characteristic
determines the behaviour of its unique
characteristics. Ariely et al (2003) pointed out that
based on the decision-making motivation of
different, "desperate" emotion in the online auction
process will distort value judgments of the auction
and then exert an influence on strategic options. Yi
Yang et al. ( 2007 ) took www.kongfz.com as an
example, analyzed the process of online bidder’s
mental accounts changing and found the default
rates were different: the small starting price and fare
increase higher than the large one’s, Participants’
were randomly greater than the experts’.
All above
results directly or indirectly confirmed the bounded
rational character of online traders.
Therefore, this article assumes that traders are
bounded rationality: a. Although the trader pursuit
their own credit score to maximize in the reputation
evaluation process, but if the other party meet own
expectations will be meet in the deal and the
evaluation; b. Traders will be impacted by the
evaluating competitors, that is forgive a little faults
in the process of exchange with each other because
of compassion.
2 NEW DESIGN PROCEDURES
OF OREM
2.1 Design of the Evaluation Process
False reputation and retaliation are two outstanding
issues of OREM. False reputation is collusive
behaviour that traders who are familiar with each
other through false transactions, thus achieving a
false evaluation of the reputation. In reality, for
avoiding this problem, both parties are required to
submit their performance.
Retaliation was mainly
due to the uncertainty caused by information
asymmetry the order of evaluation.
Currently the
seller and the buyer evaluated each other, after the
transaction is completed (Figure 1).
In fact, after each deal, buyers will generate their
own expect reputation E
B
in the payment level, then,
give opposite side a credit rating R
B
as his
compliance; And sellers results a credit rating R
S
of
buyer from the receipt of payment, then begins
shipments and generates its own credit expectations
E
S
. When the transaction is completed, if the seller
was the first evaluator, and R
S
< E
B
, the buyer will be
discontented, and retaliate evaluation R
B
< R
S
against
the seller; Likewise, if the buyer was the first
evaluator and R
B
< E
S
, seller will revenge and give
the evaluation R
S
< R
B
.
Therefore, first valuator will
always give the opposite side a good reputation for
other party can give his/her a good reputation, the
result makes the OREM forfeit the role of boosting
trust.
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
54
Figure 1: The current process of OREM.
The online bounded rational participants can
transform the strategy of himself/herself and the
adversary through information swapping.
Participants all have variant frames of reference on
reputation evaluation; they generally divided
opposite traders into several ranks on the contract’s
duty performing: Give higher rank to the
performance in good shape, give worse rank to the
performance in bad shape and the defaulters.
Therefore, designing the sequence of the reputation
evaluation could commence from trader’s
information swapping, and then perfecting the
OREM through confirmation on reputation scores.
At the same time by way of better collecting results
of reputation evaluation and providing trust
sustentation for potential trading, website should
definitely request all participants after trading
successfully evaluating the reputation of the other
party, who cannot steer next transaction within 120
days without giving evaluation.
Therefore, we re-
designed the order of the reputation evaluation
mechanism (such as figure 2).
After the trade achieved, Buyer B pays first
and submit performance testament (the electronics
file of remittance receipt) on the website, then looks
into the trading reputation history of S and submit
the expectant rank
B
E
and justification of
requisition;
Seller S starts dispatch after
receiving other party disbursement, then hands in
own dispatch testament (post article of electronics
file) on the website, then looking into the reputation
expectation and the expectation justification of
decide whether satisfied B’s expectation, submit
evaluation
S
R
and the justification, then looks into
the trading reputation history of B and submits the
expectant rank
S
E
and justification of requisition.
After received goods, B decides whether satisfy
S’s expectation on S’s duty perforation and the
expectation and justification, then submit evaluation
rank and justification.
After the termination of
evaluation, website show at the same time
S
R
and
B
R
, finally get
B
G
and
S
G
, the reputation
scores of B and S, through reckoning and switching
by a square function.
Figure 2: The flow of optimized evaluation.
The prevent OREM offer the function of leaving
message after the evaluation, the optimized
mechanism is variant, via the form of submitting
own expectation and evaluating reputation for other
party to realize communication of both parties'
during the period of evaluation.
Online
communication is assistance and supplementary of
the Realistic exchange ( Peris, 2002 ) and help to
alleviate both parties mood ( Bark, 2006 ).
So the
form of information exchanging has significant
advantages compared with the existing evaluation
mechanism:
The process of submitting expect
grounds added opportunities to justify defects in
their services and help to reflect on the inadequacies
of their services, also can affect the opponent's
selection of evaluation and achieve the ultimate
objectives to reduce abuse and purify network.
NEW MECHANISM DESIGN IN THE C2C ONLINE REPUTATION EVALUATION OPTIMIZING
55
2.2 Determination of Reputation Limit
Reputation limit, paipai (www.paipai.com) called as
transactions weight, is given the different
transactions a different reputation scores or weight,
then use the level of evaluation to achieve the
incentive role of evaluation,
we take reputation limit
replace the weights in order to reduce unnecessary
confusion generated by adding the weight levels of
evaluation.
The prevent OREMs have not classified
reputation limits, while generally determining the
limit in two ways:
Online trading platform, such
as, eBay (www.ebay.com) and Taobao
(www.taobao.com) have overlooked the existence of
different transactions, reputation limit be identified
as 1, and then use the other's evaluation to determine
reputation score, such as "favourable" 1 point ,"bad "
by 1 point, "moderate " no points;
Paipai’s single
reputation score related two factors: Reputation
evaluation and transaction amount, the formula:
Reputation score= Reputation evaluation* Transaction weights
.
The weights of transaction amount are divided into 5
levels according to different intervals (Table 1),
thus,
the transaction of 0-1RMB (including 1) is no longer
accumulate in the reputation score.
if received
"moderate" regardless of the amount of transaction,
the transaction score is 0; if received "bad" and
using the TenPay trading, the score will be minus
based on the transaction amount, the more amount,
the higher
score deducted, and "bad "will affect the
rate of the “favourable” evaluation; Similarly, in the
received "favourable" circumstances, the higher the
corresponding amount, the more bonus points, such
as reputation evaluation is "favourable" and the
transaction amount is 200 Yuan, then the reputation
of the transaction as a "favourable"( +1 ) Multiplied
by the transaction amount weight 2, gain the
reputation score 2.
In order to achieve the mechanism of incentive
role, introduced basic reputation limit and adjustable
reputation limit.
Basic reputation limit is determined
a unique basic reputation limit for each different
transaction amount,
adjustable reputation limit is
based on basic reputation provide an incentive
compatible reputation limit on the other side’s
different evaluation level.
"Good", "medium" and "bad", three levels
evaluation, in certain extent is a hierarchy of the
trader’s service. However, this classification does
not fully describe the different people’s perception.
Table 1: Paipai’s weights of different transaction amount.
Transaction amount( RMB ) Weights
0-0.99 0
1-199.99 1
200-999.99 2
1000-5000 3
>5000 4
Generally, levels of the fuzzy evaluation is
divided into excellent, good, medium and bad, so in
order to better reflect the difference feelings of
evaluator to the reputation difference of being
evaluated party.
This article is divided into the level
of Reputation evaluate "excellent", "good",
"medium" and "bad ", "terrible" five levels.
On the
basis of determining the level of evaluation, we can
determine the different limit of the evaluating level
in specific transaction amount.
While, use
2
R
1
R
0
R
1
R
2
R
represent "excellent", "good",
"medium", "bad" and "terrible".
These levels not
only may make the participants have better
perception of the quality of goods and the fulfilment
of obligations of trading partners, but also promote
the participants to improve service quality for a
higher Reputation score.
The incentive role of evaluating level should
achieved by adjusting reputation limit, on the basis
of determining the basic reputation limit
1
C
,determine the corresponding reputation limit
adjustment of different transactions combining
evaluating level and evaluating process. For
adjustable reputation limit can use the following
formula:
1
Cr rC
(1)
Among,
210 1 2
2, 1, 0, 1, 2rr r r r r


is the
evaluation given by counterparty. By the value of
R
and the relationship between the reputation
evaluation
R
and expectations
E
, we can arrive,
210 1 2
2 , 1, 0 , 1, 2 ,EE E E E E


if expect
"excellent" as
2
E
, expect "good" as
1
E
, expect
"medium" as
0
E
, expect "bad" as
1
E
, expect
"terrible" as
2
E
.
2.3 Incentive of Reputation Evaluation
The main role of prevent OREM is only to motivate
the participants improve their own Reputation scores
and grades, and then get more trading opportunities.
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
56
Although the Reputation evaluate level in some
extent reduce the opportunistic behaviour of traders,
but did not provide better incentives to promote
traders to improve their level of service, unmatched
that mechanisms design take information and
incentive as study objects.
One of successful application of mechanism
design theory is promoting the participants to tell the
truth in the auction mechanism, order to achieve a
balanced game, but the current mechanism could do
anything about it.
The reason is that Reputation
evaluation system is a service mechanism, has
essential differences with the trading mechanism;
In
the trading mechanism, strategy chosen by
Participants under the action of the mechanism not
only can affect the earnings of other participants, but
also affect their earnings,
but in the course of
optimizing the evaluation process and the exchange
of information, both strategies of evaluators only
affect counterparties income ( Reputation scores can
be seen as Reputation resources ) without affecting
benefits of themselves. This mechanism cannot
generate the game between the participants for
maximize their own return, so OREM require
introduce game analysis to solve this problem.
To
this end, by taking use of information exchange
functions of OREM, the mechanism could deal with
the participants’ submission of expectations and
evaluation through some kind of formula, achieving
the game balance of participants.
"Tell the truth" is one of the basic principles of
the effectiveness in mechanism design. The punitive
role of mechanism guarantees traders submit their
true expectation of reputation. If absence of punitive
function in OREM, the majority of participants
would choose to speculate in order to submit the
highest grade " excellent " expectation, to allow the
counterparty to give a higher evaluation, and
ultimately get more trading opportunities. Limited
rational trader in the face of punitive function will
considered submitting "false" expectation would
have loss of reputation, so that the punishment for
"lies" reputation has become a key of mechanism
design. In order to make trader submit their true
expectation in the evaluation process, formula of the
seller’s ultimate score
S
G
for each indicator and the
buyer’s ultimate score
B
G
for each indicator as
follows:if
,
SBBS
R
ER E
, then


SB
BS
GCR
GCR

(2)
if
,
SBBS
R
ER E
, then


1
22
BS
SB
BS
RE
GCCR
RE
GCR

(3)
if
,
SBBS
R
ER E
, then

1
22
SB
SB
B
S
GCR
RE
GCCR
RE

(4)
if
,
SBBS
R
ER E
, then


1
22
1
22
BS
SB
SB
B
S
RE
GCCR
RE
RE
GCCR
RE


(5)
Based on the above formula, traders see expectation
of the opposite and determine whether to meet the
expectation, If determined to meet the expectations
of all the indicators, choose " satisfied ", if not meet
the expectations of all or part of the indicators,
choose "dissatisfied", and then to re-evaluate the
opposite's indicators.
Bounded rational participants submit their true
expectations in the reference of the evaluating
history of the opposite, and as much as possible to
meet the expectation of the opposite.
Only in special
Satisfied circumstances or special dissatisfaction
there will be higher or lower evaluation than the
expectation of the opposite.
This mechanism
prevents the unnecessary dissatisfaction for both
traders, and makes traders can not only see the other
participants’ evaluation to potential traders, also see
their own position of potential traders. And
providing more information whether the trader
decide to trust a particular transaction,
contribute to
the trust of website trading environment.
3 TRADERS’ STRATEGY
CHOICE IN NEW MECHANISM
Although the two decisions are intertwined
throughout the course of evaluation and being
evaluated,
because evaluation only related with the
opposite’s reputation scores, whereas the
expectation related with own reputation scores,
then
following the attribution of reputation scores, the
whole process will be divided into two sub-game to
analyze. For each participant, the "evaluation" and
"being evaluated" are taken as two different
NEW MECHANISM DESIGN IN THE C2C ONLINE REPUTATION EVALUATION OPTIMIZING
57
processes.
Specifically, the evaluation process (after
seen the opponents’ expectation) is to submit their
evaluation to the opponent, but the being evaluated
process is to submit own expectation (to be
evaluation by opponent).
In both processes, the two
parties will adopt a different strategy in the
submission of their expectations and evaluation to
game between the traders. In part of the submit
expectation of " being evaluated " process,
each
participant will have five strategies ( "excellent",
"good", "medium", "bad" and "terrible" ) for the
different indicators to choose .
While, based on the
expectation and their own feelings of the services,
the opponent will select a strategy from the five
above strategies to cope with. Thus, five different
expectations strategies and five different assessment
strategies can be formed on a total of 25 potential
game results (Table 2).
3.1 Sequential Game Analysis
on Traders
Because the processes of game action one after the
other, that is, both games firstly submit their own
expectations for two being evaluated parties, then
the other side will submit evaluations after known
the expectations. The turn of decision time in the
two sub-game process make the two games can be
analyzed by sequential game:
Buyers cannot
accurately determine the evaluation grade of each
indicator the sellers will submit, because they cannot
exactly know the seller’s specific experience for the
level of buyer’s service in the course of submitting
their own expectations; The seller also can not
accurately determine which grade the buyer will
submit to buyer for the same reason.
In the new
mechanism ,there are two factors to determine the
reputation scores, own expectations and opposite
evaluation, expectations would be submitted in
advance known evaluation each other,
participants
received incomplete information in this expectations
submission, which makes the two sub-games are the
incomplete information game.
Then the evaluators
make evaluation based on the submission of
expectation.
This allows evaluators to take
advantage of information; making the evaluation and
decision may be influenced by the reasons
expectations and expectations. The Specific
performance is the evaluator likely to adopt a "meet"
strategy or may also take the "dissatisfied" policy.
In
the "dissatisfied" policy, the evaluator generally
submits the evaluation less than the being evaluated
expected. As bounded rationality of participants,
also may submit the higher evaluation than
expectation,
but this is less likely, because it may
make their own cost of evaluation rise. The both
sides are clearly aware of the above analyses of
bounded rational evaluation. Therefore, the
participants will also consider other possible
reaction,
so there will be the game process (Figure 3),
If affected by being evaluated, evaluator would
normally take the "meet" strategy, if not, then
evaluator may adopt the "dissatisfied" policy.
Table 2: Scores of different expectations and evaluation.
Expectation Evaluation Scores
E
2
R
2
(E
2
,R
2
)=2C
1
R
1
(E
2
,R
1
)=(3/4)C
1
R
0
(E
2
,R
0
)=–(1/2)C
1
R
-1
(E
2
,R
-1
)=–(7/4)C
1
R
-2
(E
2
,R
-2
)= –3C
1
E
1
R
2
(E
1
,R
2
)=2C
1
R
1
(E
1
,R
1
)=1C
1
R
0
(E
1
,R
0
)=–(1/4)C
1
R
-1
(E
1
,R
-1
)=–(3/2)C
1
R
-2
(E
1
,R
-2
)= –(11/4)C
1
E
0
R
2
(E
0
,R
2
)=2C
1
R
1
(E
0
,R
1
)=C
1
R
0
(E
0
,R
0
)=0C
1
R
-1
(E
0
,R
-1
)= –(5/4)C
1
R
-2
(E
0
,R
-2
)= –(5/2)C
1
E
-1
R
2
(E
-1
,R
2
)=2C
1
R
1
(E
-1
,R
1
)=C
1
R
0
(E
-1
,R
0
)=0C
1
R
-1
(E
-1
,R
-1
)=–C
1
R
-2
(E
-1
,R
-2
)= –(9/4)C
1
E
-2
R
2
(E
-2
,R
2
)=2C
1
R
1
(E
-2
,R
1
)=C
1
R
0
(E
-2
,R
0
)=–0C
1
R
-1
(E
-2
,R
-1
)= –(3/4)C
1
R
-2
(E
-2
,R
-2
)= –2C
1
Under the new mechanism, if own expectation
higher than the opposite evaluation, own reputation
scores will be subtracted a value greater than 0 based
on the opposite evaluation. Bounded rational
participants will try hard understand each other's
historical evaluation to guess the counter speculate
evaluation preference, and determine possible
evaluation then submit their expectation. Bounded
rational evaluators in the case of equal or less than
their expectations generally meet the expectations of
counterparties to reduce the evaluation time, while
in the case of higher than their expectations would
choose the dissatisfied strategy, and then submit the
real evaluation. Evaluators can be divided into
different "kind speaker" and "ticklish speaker",
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
58
"kind speaker" is vulnerable to be affected by the
submitted expectation and submit evaluation higher
than their own feeling, but "ticklish speaker" is
hardly being affected. In order to maximize their
reputation scores, the being evaluated generally
understand the types and preference of evaluators
through the evaluation mechanism, then submit their
own expectation.
Strategy 1
Strategy 2
Strategy 3
Strategy 4
Strategy 5


1
11
11
E,R
E,R , E,R ,
E,R , E,R





1
20
-1 -2


2
22
22
E,R
E,R ,E,R ,
E,R ,E,R





2
10
-1 -2
Satisfied
Dissatisfied
Satisfied
Dissatisfied
satisfied
Dissatisfied


0
00
00
E,R
E,R ,E,R ,
E,R , E,R





0
21
-1 -2


-1
-1 -1
-1 -1
E,R
E,R ,E,R
E,R ,E,R





-1
21
0-2


-2
-2 -2
-2 -2
E,R
E,R ,E,R ,
E,R ,E,R





-2
21
0-1
Evalutor A
Being evaluated M
Evaluator A
Evaluator A
Satisfied
Dissatisfied
Satisfied
Dissatisfied
Evaluator A
Evaluator A
Figure 3: Sequential game of bounded rational traders.
3.2 The Harsanyi Transformation of
Sequential Game
The reasons that participants estimate the opponents’
possible evaluation in the process of submitting
expectations mainly should be analyzed from three
aspects: Maximization of reputation score and
grade. Scholars study already confirmed that: the
higher scores reputation seller will get more trading
opportunities than the lower. Understanding of the
opponent's preferences can be better to submit their
own expectation, and maximize their reputation
scores. Expectations higher than the evaluation of
the opponent whose reputation score will be reduced,
expectations equal to the evaluation of opponent will
make their own score maximum. Understanding
of the opponent's preference is actually a process of
reducing risk. Rational participant in any process of
economic activities
, if there is the opportunity to
reduce uncertainty, all will be trying to fight for.
Similarly, in the OREM process, participants all will
try to understand the opponent's evaluation of
preferences.
1
P
2
P
Mechanism
Strategy 1
Strategy 2
Strategy 3
Strategy 4
Strategy 5


11
11 11
11 11
E,R
E,R , E,R ,
E,R , E,R
p
pp
pp





1
20
-1 -2


12
12 12
12 12
E,R
E,R , E,R ,
E,R , E,R
p
pp
pp





2
10
-1 -2
Satisfied
Dissatisfied
Satisfied
Dissatisfied
Satisfied
Dissatisfied


10
10 10
10 10
E,R
E,R , E,R ,
E,R , E,R
p
pp
pp





0
21
-1 -2


1-1
1-1 1-1
1-1 1-1
E,R
E,R , E,R ,
E,R , E,R
p
pp
pp





-1
21
0-2


1-2
1-2 1-2
1-2 1-2
E,R
E,R , E,R ,
E,R , E,R
p
pp
pp





-2
21
0-1
Evaluator A
Being evaluated M
Evaluator A
Evaluator A
Evaluator A


21
21 21
21 21
E,R
E,R , E,R ,
E,R , E,R
p
pp
pp





1
20
-1 -2


22
22 22
22 22
E,R
E,R , E,R ,
E,R , E,R
p
pp
pp





2
10
-1 -2


20
20 20
20 20
E,R
E,R , E,R ,
E,R , E,R
p
pp
pp





0
21
-1 -2


2-1
2-1 2-1
2-1 2-1
E,R
E,R , E,R ,
E,R , E,R
p
pp
pp





-1
21
0-2


2-2
2-2 2-2
2-2 2-2
E,R
E,R , E,R ,
E,R , E,R
p
pp
pp





-2
21
0-1
Evaluator A
Evaluator A
Evaluator A
Being evaluated M
Satisfied
Dissatisfied
Satisfied
Dissatisfied
Satisfied
Dissatisfied
Satisfied
Dissatisfied
Satisfied
Dissatisfied
Satisfied
Dissatisfied
Satisfied
Dissatisfied
Evaluator A
Evaluator A
Evaluator A
Strategy 1
Strategy 2
Strategy 3
Stragey 4
Strategy 5
Figure 4: The Harsanyi transformation of Sequential
game.
Sequential game of incomplete information can
take Harsanyi transformation- introducing selection
mechanism of super-participant "nature ",
the game
will be shifted into the analysis of complete
information.
For traders, the super-participant is the
evaluation mechanism, not only promote the two
sides to "speak the truth", but also provide the
opponent's history of evaluation,
so as to provide the
decisive information of the opponent's preferences
of evaluation.
On the basis of confirmed participants
will understand opponents’ preference, the use of
sequential analysis result of the game through
Harsanyi transformation can analysis and evaluate
problems of the balance of different strategies by
two trading sides.
New mechanism reserves records
of each participant's evaluation: records of
evaluating grade, records of evaluating reasons, and
NEW MECHANISM DESIGN IN THE C2C ONLINE REPUTATION EVALUATION OPTIMIZING
59
records of expectations and reasons of expectations.
So that participants can easily get possible
evaluation of the opponents for their services, that is,
can guess the probability of different opponents’
strategies.
Bounded rational participants generally
determine the probability of the counterparty’s type
based on the opponent's evaluating record and
Corresponding evaluating reasons: "kind speaker"
1
p
and "ticklish speaker"
2
p
and
12
1pp
so the information of the being evaluated is also
being turned into complete information.
In the case
of comparing reputation score under various
strategies, the being evaluated can make better
strategic choices.
3.3 Analysis of Algorithm Example
New mechanism reserves transaction information
and evaluation of information of each participant.
Transaction information which includes: transaction
time, trading commodities, counterparty and
transaction amount, and so on.
Evaluation
information, including: records of selection of
evaluating strategies, records of evaluating reasons,
and records of expectations and reasons of
expectations.
Transaction information can confirm
the authenticity of the transaction in certain extent,
while the evaluation information provides a basis for
the determination of opponents’ preferences.
Participants can determine the probabilities of
opponents’ different strategy according the
preferences of evaluation and their own level of
service, then choose the strategy of submitting
expectation based on the probability. Different risk
preferences of participants
may adopt different
strategies,
the favourable risk participants may prefer
to take risky strategy, submit a higher expectation,
and the risk-neutral participants may submit the
greatest probability of strategy, as for the risk
aversion of participants may submit smaller
expectation.
Therefore, we only make brief
description of the strategic options to participants
with different risk preferences, without deeply
analysis and discussion.
Assume that there are evaluating histories of
evaluator A:
(1) The total of evaluation 100 times, 50 times as
"excellent", 30 times as "good", 15 times as "
medium ", 3 times for the " bad ", 2 times for "
terrible ";
(2) Among 10 times of the “dissatisfied’ strategy, 9
times were less than the opposite’s expectation, the
reasons are poor package made the product has been
damaged to some extent, the other 1 was out of the
submitted expectation, the ground is product being
packed well;
(3) Further 30 times evaluations meet the "excellent"
expectations, but being noted that there are small
defects in goods or other reasons.
The reputation limit of the transaction is 1.5, the
probability of A as "kind speaker" is [30/(30+10)]=
0.75, the probability for the "ticklish speaker" is
[10/(30+10)] = 0.25, under all possible strategies the
being evaluated M’s scores can be seen in Table 3-4.
And the evaluating history of A shows he has a
certain preference to the package of products, while
the probability of A’s evaluation higher than the
opposite’s expectation is (1/10) =0.1. Then assumed
the being evaluated spent more energy in the process
of mail package, through analysis of A’s historical
evaluation, he ensure their service achieved the
"good" level (R
1
).
Table 3: The reputation scores under kind evaluator.
Probability of kind evaluator0.75
Expectation Evaluation Scores
E
2
R
2
0.75×2×1.5=2.25
R
1
0.75×(3/4) ×1.5=0.84
R
0
–0.75×(1/2) ×1.5=–0.56
R
-1
–0.75×(7/4) ×1.5=–1.97
R
-2
–0.75×3×1.5=–3.38
E
1
R
2
0.75×2×1.5=2.25
R
1
0.75×1×1.5=1.13
R
0
–0.75×(1/4) ×1.5=–0.28
R
-1
–0.75×(3/2) ×1.5=–1.69
R
-2
–0.75×(11/4) ×1.5=–3.09
E
0
R
2
0.75×2×1.5=2.25
R
1
0.75×1×1.5=1.13
R
0
0.75×0×1.5=0.00
R
-1
–0.75×(5/4)×1.5=–1.41
R
-2
–0.75×(5/2)×1.5=–2.81
E
-1
R
2
0.75×2×1.5=2.25
R
1
0.75×1×1.5=1.13
R
0
0.75×0×1.5=0.00
R
-1
–0.75×1×1.5=–1.13
R
-2
–0.75×(9/4) ×1.5=–2.53
E
-2
R
2
0.75×2×1.5=2.25
R
1
0.75×1×1.5=1.13
R
0
–0.75×0×1.5=0.00
R
-1
–0.75×(3/4)×1.5=–0.08
R
-2
–0.75×2×1.5=–2.25
Then from the comparative analysis of Table 3,
we can clearly see in the case of determining the
evaluator A as the "kind speaker", in order to
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
60
maximize its own reputation scores, M has two
options in the process of submission of their
expectations: the first is to submit "Good" (E
1
), the
score should be
11
0.75 , 1.5 1.13ER
; The
second is to submit "excellent" (E
2
), the score should
be
22 21
0.1 , 0.9 , 0.75 1.5 0.98ER ER 

.
Therefore, we can see from the above analysis,
the risk aversion being evaluated M should choose
to submit the "good" expectation; the preferable risk
being evaluated may choose to submit "excellent"
expectation.
4 CONCLUSIONS
About the current questioned online integrity and
inadequate reputation evaluation mechanisms, the
existing research results of bounded rationality as a
precondition, this paper took the behaviour of a
participant to optimize the defects of current crisis of
confidence in reputation evaluation mechanism. In
the process of Improve the existing evaluation
mechanism, this paper mainly start form three
aspects:
Optimization of the evaluation process, the
determination of adjusting reputation limit and
reputation scores,
in which optimization of the
process of the evaluation is based on the existing
processes to increase the new function of submitting
expectation, providing preconditions for the " tell the
truth " mechanism; Adjustment of reputation limit is
to increase the incentive function and to match the
level of evaluation, is an incentive reputation limit
(or weight of transaction amount) on the adjustment
of basic reputation limit. The paper divided the
whole evaluation process into two sub-processes of
"evaluation" and "being evaluated" in accordance
with the various participants, and introduced the
concept of "tell the truth" mechanism by
determination of reputation scores, improved the
incentive function of the mechanism. Finally, a
sequential game analysis of traders’ possible
strategy choice, divided the participants into "kind
speaker" and "ticklish speaker" and analysed with a
algorithm example of Harsanyi transformation
concluded that the best strategy choice of bounded
rational participants under the incentive function of
online trading system is to "tell the truth", in order to
achieve the maximization of their own reputation
scores, thus confirmed the validity of the new
mechanism.
Table 4: The reputation scores under ticklish evaluator.
Probability of ticklish evaluator0.25
Expectation Evaluation Scores
E
2
R
2
0.25×2×1.5=0.75
R
1
0.25×(3/4) ×1.5=0.28
R
0
–0.25×(1/2) ×1.5=–0.19
R
-1
–0.25×(7/4) ×1.5=–0.66
R
-2
–0.25×3×1.5=–1.13
E
1
R
2
0.25×2×1.5=0.75
R
1
0.25×1×1.5=0.38
R
0
–0.25×(1/4) ×1.5=–0.09
R
-1
–0.25×(3/2) ×1.5=–0.56
R
-2
–0.25×(11/4) ×1.5=–1.03
E
0
R
2
0.25×2×1.5=0.75
R
1
0.25×1×1.5=0.38
R
0
0.25×0×1.5=0.00
R
-1
–0.25×(5/4)×1.5=–0.47
R
-2
–0.25×(5/2)×1.5=–0.94
E
-1
R
2
0.25×2×1.5=0.75
R
1
0.25×1×1.5=0.38
R
0
0.25×0×1.5=0.00
R
-1
–0.25×1×1.5=–0.38
R
-2
–0.25×(9/4) ×1.5=–0.84
E
-2
R
2
0.25×2×1.5=0.75
R
1
0.25×1×1.5=0.38
R
0
–0.25×0×1.5=0.00
R
-1
–0.25×(3/4)×1.5=–0.03
R
-2
–0.25×2×1.5=–0.75
Mechanism design has an important guiding role
in improvement of mechanism, while the
implementation of incentive function determines the
effectiveness of the mechanism design and
optimization, thereby affecting its existence and
development. Therefore, with species diversity and
characteristic complexity of the mechanisms, how to
determine the correct classification and the
classification standards is the most basic of should
be the next main research directions in the field of
mechanism design.
REFERENCES
Ariely, D., Itamar, S., 2003. Buying, Bidding, Playing or
Competing? Value Assessment and Decision
Dynamics in Online Auctions. Journal of Consumer
Psychology, 13, pp: 113-123.
Bark, A., Bloch, N., 2006. Factors Related to Perceived
Helpfulness in Supporting Highly Distressed
Individuals through an Online Support Chat. Cyber
Psychology & Behavior, pp: 60-68.
Cabral Luis M. B., Hortacsu Ali, 2004. The Dynamics of
Seller Reputation: Theory and Evidence from eBay.
NEW MECHANISM DESIGN IN THE C2C ONLINE REPUTATION EVALUATION OPTIMIZING
61
NBER Working Paper No. W10363. March, Available
at SSRN: http://ssrn.com/abstract=516704.
Chris Snijders, Richard Zijdeman, 2004. Reputation and
Internet Auctions: eBay and Beyond. Analyse & Kritik
26. pp: 158-184.
Christina Aperjis, Ramesh Johari, 2010. Designing
Reputation Mechanisms for Efficient Trade. Available
at SSRN: http://ssrn.com/abstract=1596839.
Chrysanthos Dellarocas, F. Dini and G. Spagnolo2006a.
Designing Reputation (Feedback) Mechanisms.
Handbook of Procurement, Cambridge University
Press.
Chrysanthos Dellarocas, 2006b. Reputation Mechanisms,
in Hendershott, T. (Eds.): Handbook on Economics
and Information Systems. Elsevier Publishing. New
York, pp: 629–660.
Chrysanthos Dellarocas, 2001. Building Trust On-Line
The Design of Reliable Reputation Reporting
Mechanisms for Online Trading Communities. MIT
Sloan School of Management. Sloan Working Paper
Jan-80.
Chrysanthos Dellarocas, 2006c. How Often Should
Reputation Mechanisms Update a Trader’s Reputation
Profile? Information Systems Research, pp: 1-32.
Dahui Li, Zhangxi Lin, 2004. Negative Reputation Rate as
the Signal of Risk in Online Consumer-to-consumer
Transactions. Proceedings of ICEB, 2004.
Dongmin Kim, Izak Benbasat, 2003. Trust-related
Arguments in Internet Stores: A Framework for
Evaluation. Journal of Electronic Commerce
Research, VOL. 4, NO. 2, pp: 49-64.
Hitoshi Yamamoto, Kazunari Ishida, Toshizumi Ohta,
2004. Trust Formation in a C2C Market: Effect of
Reputation Management System. In: Proceedings of
The Workshop on Deception, Fraud and Trust in
Agent Societies at The Third International Joint
Conference on Autonomous Agents and Multi-Agent
Systems (AAMAS-2004). New York, pp: 126-136.
Jennifer Brown, John Morgan, 2006. Reputation in Online
Markets: Some Negative Feedback EB/OL. http:
// are.berkeley.edu/.
Jeffrey, A. L., 2005. How valuable is good reputation? A
Sample Selection Model of Internet Auctions. The
Review of Economics and Statistics. 87(3), pp: 453-
465.
Kamins, M., Xavier, D., Valerie, S. F., 2004. Effects of
Seller-supplied Prices on Buyers’ Product
Evaluations: Reference Prices in an Internet Auction
Context. Journal of Consumer Research. 30(4), pp:
622-628.
Mikhail, I. M., James, A., 2002. Dose a seller’s
ecommerce reputation matter? Evidence from EBay
Auctions. The Journal of Industrial Economics. (3),
pp: 337-349.
On, A., Dan, A., 2002. E-rationality: Rationality in
Electronic Environments? Advances in Consumer
Research. pp: 261-262.
Peris, R., Gimeno, M. A., Pinazo, D., et al, 2002. Online
Chat Rooms: Virtual Spaces of Interaction for Socially
Oriented People. Cyber Psychology & Behavior. pp:
43-51.
Simon H., 1955. A Behavioral Model of Rational Choice.
Quarterly Journal of Economics. pp: 99-18.
Qing Wang, Yong Zhao, 2009. A Bounded Rationality
Revealed Preference and Satisfying Decision Analysis.
Journal of Hua zhong University of Science and
Technology(Nature Science Edition). (8), pp: 5-7.
Yi Yang, Jinbao Qian, 2007. Analysis on the Influence of
Online Auction Result on Bidders’ Mental Accounting:
An Empirical Research from Kongfz Secondhand Book
Net. Economic Management. (6), pp: 43-47.
Yi Yang, Xiliang Zhang, Jinbao Qian, 2009. Behavior of
Competitors Online Auctions: Overview and Prospect.
Economic Perspectives. (1), pp: 86-89.
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
62