open pdf file in asp net c# : How to add text fields to pdf Library SDK class asp.net wpf web page ajax 173_Dellarocas_Word_of_Mouth0-part490

For more information, 
ebusiness@mit.edu or 617-253-7054 
please visit our website at
http://ebusiness.mit.edu 
or contact the Center directly at
A research and education initiative at the MIT 
Sloan School of Management
t
THE DIGITIZATION OF WORD-OF-MOUTH:  
PROMISE AND CHALLENGES OF ONLINE 
FEEDBACK MECHANISMS
Paper 173 
Chrysanthos Dellarocas
October 2003 
How to add text fields to pdf - insert text into PDF content in C#.net, ASP.NET, MVC, Ajax, WinForms, WPF
XDoc.PDF for .NET, providing C# demo code for inserting text to PDF file
how to enter text into a pdf; add text boxes to pdf document
How to add text fields to pdf - VB.NET PDF insert text library: insert text into PDF content in vb.net, ASP.NET, MVC, Ajax, WinForms, WPF
Providing Demo Code for Adding and Inserting Text to PDF File Page in VB.NET Program
how to add text fields in a pdf; add text to pdf online
The Digitization of Word of Mouth: Promise
and Challenges of Online
Feedback Mechanisms
Chrysanthos Dellarocas
Sloan School of Management, Massachusetts Institute of Technology,Cambridge, Massachusetts 02139
dell@mit.edu
O
nlinefeedbackmechanismsharness the bidirectional communication capabilities of the
Internet to engineerlarge-scale, word-of-mouth networks.Bestknown sofar as a tech-
nology for building trust and fostering cooperation in online marketplaces, such as eBay,
thesemechanismsare poisedto havea much wider impact on organizations.Their growing
popularity has potentially important implications for a wide range of management activi-
ties such as brand building, customer acquisition and retention, product development, and
quality assurance. This paper surveys our progress in understanding the new possibilities
andchallenges that these mechanisms represent.It discusses some important dimensions in
which Internet-basedfeedback mechanisms differfrom traditional word-of-mouth networks
and surveys the most important issues related to their design, evaluation, and use. It pro-
vides an overviewofrelevantworkingametheoryandeconomicson thetopicofreputation.
Itdiscusses howthisbodyof work is beingextendedandcombinedwith insights fromcom-
puter science, management science, sociology, andpsychologytotake intoconsideration the
special properties of online environments. Finally, it identifies opportunities that this new
area presents foroperations research/management science (OR/MS) research.
(Reputation Mechanisms;Online Feedback; Electronic Markets; Trust; Internet;Game Theory)
1. Introduction
Oneofthe most importantcapabilitiesofthe Internet
relativetopreviousmasscommunicationtechnologies
is its bidirectionality. Through the Internet, not only
can organizations reach audiences of unprecedented
scaleatalowcost,butalso,forthefirsttimeinhuman
history,individuals can maketheirpersonal thoughts,
reactions,andopinions easily accessible tothe global
community of Internet users.
Word of mouth, one of the most ancient mech-
anisms in the history of human society, is being
given new significance by this unique property of
the Internet. Online feedback mechanisms, also known
as reputation systems (Resnick et al. 2000), are using
the Internet’s bidirectional communication capabili-
ties to artificiallyengineerlarge-scale, word-of-mouth
networks in which individuals share opinions and
experienceson awiderangeoftopics,includingcom-
panies, products, services, and even world events.
Table 1 lists several noteworthy examples of such
mechanisms in usetoday.
Perhaps thebest-known application ofonline feed-
back mechanisms to date has been their use as a
technology for building trust in electronic markets.
This has been motivated by the fact that many tra-
ditional trust-building mechanisms, such as state-
enforced contractual guarantees, tend to be less
effective in large-scale, online environments (Kollock
1999).Online feedback mechanisms have emerged as
aviable mechanism for fostering cooperation among
strangers insuch settings by ensuring that thebehav-
ior of a trader toward any other trader becomes
0025-1909/03/4910/1407
1526-5501electronicISSN
ManagementScience © 2003 INFORMS
Vol. 49, No. 10, October2003, pp. 1407–1424
VB.NET PDF Form Data Read library: extract form data from PDF in
featured PDF software, it should have functions for processing text, image as Add necessary references Demo Code to Retrieve All Form Fields from a PDF File in
how to add text to pdf; adding text to pdf in acrobat
C# PDF Form Data Read Library: extract form data from PDF in C#.
Able to retrieve all form fields from adobe PDF file in C# featured PDF software, it should have functions for processing text, image as Add necessary references
add text to pdf in preview; add text pdf file
DELLAROCAS
TheDigitizationofWordofMouth
Table1
SomeNoteworthyExamplesofOnlineFeedbackMechanisms(inUseasofMarch2003)
Website
Category
Summaryoffeedbackmechanism
Formatofsolicitedfeedback
Formatofpublishedfeedback
BBC
Worldnews
BBC’sTalkingPoint:Reader
Readerssendtheiropinionson
Selectivelistof(potentiallyedited)
forumonworldevents.
specifictopicsselectedby
opinionssubmittedbyreaders;
TalkingForumstaffinfreetext
noquantitativestatisticsprovided.
format;readerspropose
possibletopicsofinterest.
Citysearch
Entertainmentguide
Usersraterestaurants,bars,
Usersratemultipleaspectsof
Weightedaveragesofratingsper
clubs,hotels,andshops.
revieweditemsfrom1–10and
aspectreflectingbothuser
answeranumberofyes/no
andeditorialratings;user
questions;readersratereviews
reviewscanbesortedaccording
as“useful,”“notuseful,”andsoon.
to“usefulness.”
eBay
Onlineauctionhouse
Buyersandsellersrateone
Positive,negative,orneutral
Sumsofpositive,negative,and
anotherfollowingtransactions.
ratingplusshortcomment;
neutralratingsreceivedduring
rateemaypostaresponse.
past6months(see§3).
eLance
Professionalservices
Contractorsratetheirsatisfaction
Numericalratingfrom1–5plus
Averageofratingsreceived
marketplace
withsubcontractors.
comment;rateemaypost
duringpast6months.
aresponse.
Epinions
Onlineopinionsforum
Userswritereviewsabout
Usersratemultipleaspectsof
Averagesofitemratings;percent
productsandservices;
revieweditemsfrom1–5;
ofreaderswhofound
othermembersratethe
readersratereviewsas
areview“useful.”
usefulnessofreviews.
“useful,”“notuseful,”andsoon.
Google
Searchengine
Searchresultsareordered
AWebpageisratedbased
Noexplicitfeedbackscoresare
basedonhowmany
onhowmanylinkspoint
published;orderingacts
sitescontainlinksthatpoint
toit,howmanylinkspoint
asanimplicitindicator
tothem(BrinandPage1998).
tothepointingpage,andsoon.
ofreputation.
Slashdot
Onlinediscussionboard
Postingsareprioritizedorfiltered
Readersratepostedcomments.
Noexplicitfeedbackscoresare
accordingtotheratingsthey
published;orderingacts
receivefromreaders.
asanimplicitindicator
ofreputation.
publicly known and may, therefore, affect the behav-
ior of the entire community towardthattrader in the
future. Knowing this, traders have an incentive to
behave well towardeach other,even if their relation-
ship is a one-time deal.As I discuss in §3,a growing
bodyofempiricalevidenceseems todemonstratethat
these systems have managed to provide remarkable
stability in otherwiserisky trading environments.
The application of feedback mechanisms in online
marketplacesisparticularlyinteresting,becausemany
of these marketplaces would probably not have
come into existence without them. It is, however, by
no means the only possible application domain of
such systems. Internet-based feedback mechanisms
are appearing in a surprising variety of settings. For
example, Epinions.com encourages Internet users to
rate practically any kind of brick-and-mortar busi-
ness such as airlines, telephone companies, resorts,
and so on. Moviefone.com solicits and displays
user feedback on new movies alongside professional
reviews,andCitysearch.comdoesthesameforrestau-
rants, bars, and performances. Even news sites, per-
haps thebest embodiment of the unidirectional mass
media of theprevious century,arenow solicitingand
publishingreaderfeedbackonworldeventsalongside
professionallywritten news articles (see, for example,
BBC’s Talking Point Web forum).
The proliferation of online feedback mechanisms
is already changing people’s behavior in subtle but
important ways. Anecdotal evidence suggests that
people now increasingly rely on opinions posted on
such systems to make a variety of decisions ranging
from whatmovieto watchtowhatstocks to investin
(Guernsey 2000). Only five years ago, the same peo-
ple would primarily base those decisions on adver-
tisements orprofessional advice.
1408
ManagementScience/Vol.49, No. 10, October 2003
VB.NET PDF Convert to Text SDK: Convert PDF to txt files in vb.net
Convert PDF to text in .NET WinForms and ASP.NET project. Text in any PDF fields can be copied and pasted to .txt files by keeping original layout.
how to add a text box in a pdf file; how to add text to a pdf document
C# PDF insert image Library: insert images into PDF in C#.net, ASP
C#.NET PDF SDK - Add Image to PDF Page in C#.NET. How to Insert & Add Image, Picture or Logo on PDF Page Using C#.NET. Add Image to PDF Page Using C#.NET.
add text to pdf document in preview; adding text pdf files
DELLAROCAS
TheDigitization of Wordof Mouth
Such trends have important repercussions for OR/
MS. Managers of today’s networked organizations
need to understand how the growing popularity
of large-scale, online feedback mechanisms affects a
wide range of activities within their organizations.
Some examples include:
• Brand building and customer acquisition. Online
feedback mechanisms can serve as a low cost and,
potentially, effective channel for acquiring and re-
taining customers, complementary to advertising
(Mayzlin 2003). At the same time, they quickly dis-
seminate bad news that can potentially harm brand
equity.
• Product development and quality control. Online
feedbacknetworks canassistan organizationtobetter
understandconsumerreactions toits current product
line. At the same time, they reveal this information
to competitors and they also accelerate the dissemi-
nation of information about product defects.
• Supplychainqualityassurance.Industry-widefeed-
back mechanisms can assist organizations to better
assess prospective first-time suppliers; they can also
actasapowerfuldiscipliningmechanismthatensures
fulfillment of contractual obligations and can poten-
tially lowerthe legal costs of doing business.
There is currently little work studying these and
other related possibilities. The rising importance of
onlinefeedbackmechanismsnotonlyinvites,butalso
necessitates, rigorous OR/MS research on their func-
tioningandconsequences. How dosuch mechanisms
affect thebehaviorof participants in thecommunities
where they are introduced? To what extent can their
operators and participants manipulate them? How
can communities protectthemselves from such poten-
tial abuse? Which mechanism designs work best in
which settings?Thisisjustasmallsubsetofquestions
thatinviteexciting and valuable research.
Incommon withotherInternettechnologies,online
feedbackmechanismsintensifythe interdependencies
between organizations, their customers, their part-
ners, and their competitors. Managers will,therefore,
findthatproperdecisionmaking relatedtotheimple-
mentation and use of feedback mechanisms requires
careful consideration, not only of their own actions,
but alsoofthe likely responses ofother players inter-
connected through them. Accordingly, the tools of
game theory play a prominent role in the study of
thesemechanisms.1
This paper surveys our progress so far in under-
standing the new possibilities and challenges that
thesemechanismsrepresent.Section2 discussessome
important dimensions in which Internet-based feed-
back mechanisms differ from traditional word-of-
mouth networks. Section 3 presents an overview of
eBay’sfeedback mechanism,perhapsthebest-studied
online feedback mechanism to date. It summarizes
initial field evidence on the mechanism’s proper-
ties and formulates the most important open ques-
tions relating to designing, evaluating, and using
such mechanisms. The next two sections survey our
progress in developing a systematic discipline that
can help answer those questions. Section 4 provides
anoverviewofrelevantpastworkingametheoryand
economics. Section 5 then discusses howthis stylized
body of work is being extended to take into consid-
eration the special properties of online environments.
Finally, §6 summarizes the main points of this paper
and discusses the opportunities that this new area
presents forOR/MS research.
2. Online Feedback Mechanisms:
An Ancient Concept
in a Modern Setting
Word-of-mouth networks constitute an ancient solu-
tion to a timeless problem of social organization: the
elicitation of good conduct in communities of self-
interestedindividualswhohaveshort-termincentives
to cheat one another. The historical appeal of these
networks has been their power toinduce cooperation
without the need for costly enforcement institutions.
Before the establishment of formal law and central-
ized systems of contract enforcement backed by the
sovereignpowerofastate,mostancientandmedieval
1OR/MSisnot alone inrealizingthat thehigherdegreeoforga-
nizationalinterdependencebroughtaboutbytheInternetincreases
the need to incorporate game-theoretic concepts and techniques
insystemdesignmethodologies.Papadimitriou(2001)providesan
insightful discussion of how the properties of the Internet have
generatedsubstantialinterest among computerscientists inincor-
poratinggametheoryintoalgorithmandcomputersystemdesign.
Management Science/Vol. 49, No. 10, October2003
1409
VB.NET PDF Text Extract Library: extract text content from PDF
With this advanced PDF Add-On, developers are able to extract target text content from source PDF document and save extracted text to other file formats
adding text field to pdf; how to add text to a pdf in preview
VB.NET PDF insert image library: insert images into PDF in vb.net
try with this sample VB.NET code to add an image As String = Program.RootPath + "\\" 1.pdf" Dim doc New PDFDocument(inputFilePath) ' Get a text manager from
adding text fields to a pdf; how to add text to a pdf file in acrobat
DELLAROCAS
TheDigitizationofWordofMouth
communities relied on wordof mouth as theprimary
enablerofeconomic and social activity (Benson 1989,
Greif1993, Milgrom etal. 1990),andmany aspectsof
socialandeconomic lifestilldosotoday(Klein 1997).
What makes online feedback mechanisms differ-
ent from the word-of-mouth networks of the past
is the combination of (1) their unprecedented scale,
achieved through the exploitation of the Internet’s
low-cost, bidirectional communication capabilities,
(2) the ability of their designers to precisely control
andmonitor theiroperation through the introduction
of automated feedback mediators, and (3) new chal-
lenges introduced by the unique properties of online
interaction,such as the volatile nature of online iden-
tities and the almost complete absence of contextual
cues that would facilitate the interpretation of what
is, essentially, subjectiveinformation.
Scale Enables New Applications. Scale is essen-
tial to the effectiveness of word-of-mouth networks.
In an online marketplace, for example, sellers care
about buyer feedback primarily to the extent that
they believe that it might affect their future profits.
This can only happen if feedback is provided by a
sufficient number of current customers and commu-
nicated to a significant portion of future prospects.
Theory predicts that a minimum degree of partic-
ipation in word-of-mouth communities is required
before reputation effects can induce any cooperation.
Once this threshold is reached, however, the power
of reputation immediately springs to life and high
levels of cooperation emerge in a discontinuous fash-
ion (Bakos andDellarocas 2002).Therefore,the vastly
increased scale of Internet-based feedback mecha-
nisms is likelytorenderthempowerfulinstitutions in
environments where traditional word-of-mouth net-
works wereheretoforeconsideredineffectivedevices.
The social, economic,andperhaps even political con-
sequences of such a trend deserve careful study.
Information Technology (IT) Enables Systematic
Design. In pre-Internet societies word of mouth
emergednaturallyandevolvedin waysthatweredif-
ficult to control or model. The Internet allows this
powerful social force to be precisely measured and
controlled through proper engineering of the infor-
mation systemsthatmediateonlinefeedbackcommu-
nities. Such automatedfeedback mediators specify who
can participate, what type of information is solicited
fromparticipants,howitisaggregated,andwhattype
of information is made availabletothem about other
community members. Through the proper design of
these mediators, mechanism designers can exercise
precise control over a number of parameters that
are difficult or impossible to influence in brick-and-
mortar settings.Forexample, feedbackmediators can
replace detailed feedback histories with a wide vari-
ety of summary statistics; they can apply filtering
algorithmstoeliminateoutlierorsuspectratings;they
can weight ratings according to some measure of the
rater’s trustworthiness, etc. Such degree of control
can impact the resulting social outcomes in nontriv-
ial ways (see §§5.2–5.4). Through the use of informa-
tion technology, what had traditionally fallen within
the realm of the social sciences is, to a large extent,
being transformed to an engineering design prob-
lem. Understanding the full space of design possi-
bilities and the impacts of specific design choices on
theresultingsocialoutcomes isanimportantresearch
challengeintroducedby thesenewsystems.
Online Interaction Introduces New Challenges.
Thedisembodiednatureofonlineenvironmentsintro-
duces several challenges related to the interpretation
and use of online feedback.Some ofthesechallenges
have their roots in the subjective nature of feedback
information. Brick-and-mortar settings usually pro-
vide a wealth of contextual cues that assist in the
properinterpretation of opinions andgossip (such as
familiarity with the person who acts as the source of
that information, the ability to draw inferences from
the source’s facial expression or mode of dress, and
so on). Most of these cues are absent from online
settings. Readers of online feedback are, thus, faced
with the task of evaluating the opinions of complete
strangers. Otherchallengestofeedback interpretation
have their roots in the ease with which online iden-
tities can be changed. This opens the door to various
forms of strategic manipulation. For example, com-
munity members can build good reputations, milk
it by cheating other members, and then disappear
and reappear under new online identities and clean
records (Friedman and Resnick 2001). They can use
fake online identities topost dishonest feedback and,
thus, try to inflate their reputation or tarnish that of
1410
ManagementScience/Vol.49, No. 10, October 2003
VB.NET PDF Password Library: add, remove, edit PDF file password
VB: Add Password to PDF with Permission Settings Applied. This VB.NET example shows how to add PDF file password with access permission setting.
add text field pdf; add text to pdf file online
C# PDF Text Extract Library: extract text content from PDF file in
How to C#: Extract Text Content from PDF File. Add necessary references: RasterEdge.Imaging.Basic.dll. RasterEdge.Imaging.Basic.Codec.dll.
how to enter text in pdf; how to insert a text box in pdf
DELLAROCAS
TheDigitization of Wordof Mouth
their competitors (Dellarocas 2000). Finally,the medi-
ated nature of online feedback mechanisms raises
questions relatedto the trustworthiness of their oper-
ators. An important prerequisite for the widespread
acceptance of online feedback mechanisms is, there-
fore,a betterunderstandingof howsuch systemscan
be compromised and the development of adequate
defenses.
3. A Case Study: eBay’s
Feedback Mechanism
eBay’s feedback mechanism is, arguably, the best-
studied online feedback mechanism to date. This
sectionsummarizesinitialfieldevidenceon themech-
anism’s properties and motivates the need for a
systematic discipline of online feedback mechanism
design andevaluation.
Founded in September 1995, eBay is the leading
online marketplace for the sale of goods and ser-
vicesbyadiversecommunityofindividualsandbusi-
nesses.Atthebeginningof2003,theeBaycommunity
numbered 49.7 million registered users, and was the
most popular shopping site on the Internet when
measuredby total user minutes.2
One of the most remarkable aspects of eBay is that
the transactions performed through it are not backed
up by formal contractual guarantees. Instead,cooper-
ation and trust are primarily based on the existence
of a simple feedback mechanism. This mechanism
allows eBay buyers and sellers to rate one another
following transactions and makes the history of a
trader’s past ratings public to the entire community.
For an overview of eBay’s feedback mechanism, the
reader is referredtoResnickandZeckhauser (2002).
Summary of Empirical Evidence. eBay’s impres-
sive commercial success seems to indicate that its
feedback mechanism has succeeded in achieving its
primary objective: generate sufficient trust among
buyers to persuade them to assume the risk of
transacting with complete strangers (Ba and Pavlou
2002). Because sufficiently does not necessarily mean
efficiently, eBay’s success has generated substantial
2Source:MediaMetrix(2001).
interest in better understanding how its feedback
mechanismworks,howmuchitcontributestoitssuc-
cess, and how its success can be replicated in other
environments.
Afirst set of results comes from empirical studies.
Even a surface analysis of a representative eBay data
set can uncover some interesting properties (Resnick
andZeckhauser 2002):
• Most trading relationships are one-time deals:
89%ofall buyer-sellerpairs conductedjustonetrans-
action during the five-month period covered by the
data set.
• Buyers leftfeedbackon sellers 52.1%of the time;
sellers on buyers 60.6% of the time.
• Feedback is overwhelmingly positive; of feed-
back provided by buyers, 99.1% of comments were
positive, 0.6%werenegative,and0.3% were neutral.
Anumberof studies have delved deeperinto eBay
data sets to uncover additional properties. Resnick
et al. (2002) provide a comprehensive survey and
methodological critiqueofthese works.Theauthor is
aware of 14 such studies, as summarized in Tables 2
and 3. All follow a similar logic, though the details
vary in important ways. Apart from one laboratory
experiment, each is an observational study of a par-
ticular category of items.
The following points summarizetheprincipal con-
clusions derived from a collective reading of these
works:
• Feedback profiles seem to affect both prices and
theprobabilityofsale.However,thepreciseeffectsare
ambiguous; different studies focus on different com-
ponents ofeBay’scomplex feedbackprofile andoften
reach differentconclusions.
• The impact of feedback profiles on prices and
probabilityofsaleisrelativelyhigherforriskiertrans-
actions andmore expensive products.
• Among all different pieces of feedback informa-
tion that eBay publishes for a member, the compo-
nents that seem to be most influential in affecting
buyer behavior are the overall number of positive
and negative ratings, followed by the number of
recently (last seven days,lastmonth)postednegative
comments.
Towards a Systematic Discipline of Feedback
Mechanism Design. The initial evidence provided
Management Science/Vol. 49, No. 10, October2003
1411
C# PDF Password Library: add, remove, edit PDF file password in C#
C# Sample Code: Add Password to PDF with Permission Settings Applied in C#.NET. This example shows how to add PDF file password with access permission setting.
add text pdf professional; how to input text in a pdf
VB.NET PDF Text Add Library: add, delete, edit PDF text in vb.net
Extract; C# Write: Insert text into PDF; C# Write: Add Image to PDF; C# Protect: Add Password to PDF; C# Form: extract value from fields; C# Annotate: PDF Markup
adding text to a pdf document acrobat; add text to pdf reader
DELLAROCAS
TheDigitizationofWordofMouth
Table2
SummaryofPrincipalResults
Shorthand
Citation
Itemssold
Remarks
BP
BaandPavlou(2002)
Music,software,electronics
Positivefeedbackincreasedestimatedprice,
butnegativefeedbackdidnothaveaneffect.
BH
BajariandHortacsu(2003)
Coins
Bothpositiveandnegativefeedbackaffectprobabilityof
modeledbuyerentryintotheauction,butonlypositive
feedbackhadasignificanteffectonfinalprice.
DH
DewanandHsu(2002)
Stamps
Highernetscoreincreasesprice.
E
Eaton(2002)
Electricguitars
Negativefeedbackreducesprobabilityofsale,butnot
priceofsolditems.
HW
HouserandWooders(2000)
Pentiumchips
Positivefeedbackincreasesprice;negativefeedback
reducesprice.
KM
KalyanamandMcIntyre(2001)
PalmPilotPDAs
Positivefeedbackincreasesprice;negativefeedback
reducesprice.
KW
KauffmanandWood(2000)
Coins
Nosignificanteffects,butnegativefeedbackseemsto
increase priceinunivariateanalysis.
LIL
Leeetal.(2000)
Computermonitorsandprinters
Negativefeedbackreducesprice,butonlyforuseditems.
L
Livingston(2002)
Golfclubs
Positivefeedbackincreasesbothlikelihoodofsaleand
price;effecttapersoffoncearecordisestablished.
LBPD
Lucking-Reileyetal.(2000)
Coins
Noeffectfrompositivefeedback;negativefeedback
reducesprice.
MA
MelnikandAlm(2002)
Goldcoins
Positivefeedbackincreasesprice;negativefeedback
decreasesprice.
MS
McDonaldandSlawson(2002)
Dolls
Highernetscore(positivesandnegatives)increasesprice.
RZ
ResnickandZeckhauser(2002)
MP3players,Beaniebabies
Bothformsoffeedbackaffectprobabilityofsalebutnot
pricecontingentonsale.
RZSL
Resnicketal.(2002)
Vintagepostcards
Controlledfieldexperiment;establishedsellercommands
higherpricesthannewcomers;amongnewcomers,small
amountsofnegativefeedbackhavelittleeffect.
Note.AdaptedfromResnicketal.(2002).
by empirical studies, though useful, does not help
answerthemostimportantunderlyingquestion:How
well does eBay’s mechanism work? In fact, these
studiesraiseawholenewset ofinteresting questions.
Forexample,whyis thefraction ofnegative feedback
Table3
SummaryofPrincipalQuestions
Questionconsidered
Studies
Howdoesaseller’sfeedbackprofileaffectprices?
All
Howdoesaseller’sfeedbackprofileaffectthe
BH,E,L,RZ
probabilityofsale?
Doesfeedbackmattermoreforriskiertransactions
BP,LIL
ormoreexpensiveproducts?
HowdopricesoneBaycomparetopricesina
DH,KW
moreconventionalchannel?
WhatcomponentsofeBay’sfeedbackprofilebetter
DH
explainbuyerbehavior?
Note.ShorthandisdefinedinTable2,Columns1and2.
solow? Is this an indication of themechanism’s poor
functioning(buyers arereluctant toexpresstheirtrue
opinions fearing retaliatory bad ratings from sellers),
or perhaps a consequence of the mechanism’s suc-
cess (sellers are induced to behave well and, there-
fore,therearesimplyfewdissatisfiedbuyers)?Whyis
therelationship betweenfeedbackandpricesambigu-
ous? Is this an indication that the mechanism is not
well designed, or perhaps that many users do not
yet understand how to best process the information
it provides?
In the author’s opinion, the two most concrete
evaluation criteria of a feedback mechanism’s perfor-
mance ought to be (1) the expected payoffs of the
outcomes induced by the mechanism for the various
classes of stakeholders over the entire time horizon
that matters for each of them, and (2) the robustness
1412
ManagementScience/Vol.49, No. 10, October 2003
DELLAROCAS
TheDigitization of Wordof Mouth
of those outcomes against different assumptions
about the participants’ behavior.3
Calculationofpayoffsrequiresanunderstandingof
how eBay’s mechanism affects the behavior of buy-
ers and sellers and how these behaviors evolve over
time if all players are simultaneously pursuing their
own interests. The tools of game theory are, thus,
instrumentalindevelopingconceptualmodelsofsuch
systems.
Robustness considerations are especially important
on eBay becausethe essence of feedback mechanisms
relies on voluntary elicitation of behaviorand this, in
turn,relieson anumberofassumptions abouthuman
rationality and beliefs. Two issues stand out as par-
ticularlyimportant.First, becausefeedback provision
is currently voluntary, the impact of incomplete or
untruthful feedback needs to be better understood.
Second, thevulnerabilityof thesystemagainst strate-
gic manipulationandonlineidentitychanges mustbe
carefully studied.
Once we have sufficiently understood the proper-
ties and performance of eBay’s current mechanism,
the next obvious question is: How can this mecha-
nism be improved? Answering this question requires
abetterunderstanding of theunique design possibil-
ities of onlinefeedback mechanisms.A fewexamples
of a much larger setof possibilities follow.
• Online feedback mechanisms can precisely con-
trol the form of information they solicit: eBay asks
users to rate transactions as positive, negative, or neu-
tral. Would it have been better to ask users to rate
transactions on ascalefrom1–5(which is whatAma-
zon does)? Could a question with different phrasing
leadtoeven higherefficiency?
• Feedback mechanisms control how information
gets aggregated and what information is publicly
available in feedback profiles. Currently,eBay’s feed-
back profile is a relatively complex artifact that
includes the entire history of ratings together with a
numberofsummary statistics.Because differentusers
pay attention to different subsets of this informa-
tion, this complicates the modeling and predictions
3Otherplausible,butcurrentlylesswellunderstoodevaluationcri-
teriaincludeinducingoutcomesthatareperceivedas“fair”bythe
majorityofplayersandensuring theprivacyoftheparticipants.
of the induced outcomes. Would it be better to hide
some parts of this information (for example, the
detailed feedback history?). Would some other sum-
mary statistics (e.g., the fraction of negative ratings)
lead to even more efficient outcomes? Would it be
desirableto implementsomesort of automatedfilter-
ing of ratings that fail to satisfy some criteria?
• Feedback submission is currently voluntary on
eBay. Furthermore, there is currently no quality
control of submitted feedback. Could eBay intro-
duce a carefully designed system of buyer fees and
rewards that elicits complete participation andtruth-
ful feedback?
Athirdsetofquestionsrevolvesaroundhowonline
feedbackmechanisms comparewithmoreestablished
institutions for achieving similar outcomes, such as
formal contracts and advertising. These comparisons
are important; their outcome will help determine
how wide an impact these mechanisms will ulti-
mately have.
An objective of any discipline of design is to even-
tually abstract from the study of specific cases and
reach some general principles and guidelines. In the
case of feedback mechanisms, this objective trans-
lates to recognizing general classes of settings where
feedback mechanisms may be usefully applied, iden-
tifying important families of feedback mechanism
architectures, and understanding what architectures
are best suited to what settings.
Finally, design involves a responsibility for detail;
this creates a need to deal with complications. In
the service of design, several established OR/MS
paradigms, such as decision theory and simulation,
and empirical and experimental studies, are natu-
ral complements to game theory, both for qualifying
these models and for adapting them to account for
thecomplexities of the“realworld” andthebounded
rationality of actual human behavior.4 The restof this
paper provides a survey of past work that can serve
as a starting pointfor answering the above questions
in a systematic way.
4SeeRoth(2002)forabroaddiscussionofthenew
methodological
challengesintroducedbytheincreasinguseofeconomics,notonly
foranalyzingmarkets, butalsofordesigningthem.
Management Science/Vol. 49, No. 10, October2003
1413
DELLAROCAS
TheDigitizationofWordofMouth
4. Reputation in Game Theory
and Economics
Given the importance of word-of-mouth networks in
human society, reputation formation has been exten-
sivelystudiedby economists using the tools of game
theory.Thisbodyofworkisperhapsthemostpromis-
ingfoundationfordevelopingananalytical discipline
of online feedback mechanism design. This section
surveys past work on this topic, emphasizing the
results that are most relevant to the design of online
feedback mechanisms. Section 5 then discusses how
this stylized body of work is being extended to
addresstheuniquepropertiesofonlineenvironments.
4.1. Basic Concepts
According to Wilson (1985), reputation is a concept
that arises in repeated game settings when there is
uncertainty about some property (the “type”) of one
ormoreplayersinthemindofotherplayers.If“unin-
formed” players have access to the history of past
stage game outcomes, reputation effects then often
allow informed players to improve their long-term
payoffs by gradually convincing uninformed players
thattheybelongtothetypethat bestsuits theirinter-
ests. Theydothis byrepeatedly choosing actions that
make them appear to uninformed players as if they
were of the intended type, thus acquiring a “reputa-
tion” for being of that type.
The existence of some initial doubt in the mind of
uninformed players regarding the type of informed
players is crucial for reputation effects to occur. To
see this, consider a repeated game between a long-
run player and a sequence of short-run (one-shot)
opponents. In every stage game, the long-run player
can choose oneof several actions,but cannotcredibly
commit to any of those actions in advance. If there
is no uncertainty about the long-run player’s type,5
rational short-run players will then always play their
stage game Nash equilibrium response. Such behav-
ior typically results in inefficient outcomes.
Consider, for example, the following stylized ver-
sion of a repeated “online auction” game. A long-
lived seller faces an infinite sequence of sets of
5Inother words,ifshort-runplayersareconvincedthat the long-
runplayerisarationalutility-maximizingplayerwhosestagegame
payoffsareknownwithcertainty.
identical one-time buyers in a marketplace where
there are onlytwokinds of products:
(1)low-qualityproductsthatcost0tothesellerand
areworth 1 to thebuyers, and
(2) high-quality products that cost 1 to the seller
and areworth 3 to thebuyers.
Each period the seller moves first, announcing the
quality of the product he promises to buyers. High-
qualityproducts are moreprofitable,so thesellerwill
always promise high quality. Buyers then compete
with one anotherin a Vickrey auction and, therefore,
bid amounts equal to their expected valuation of the
transaction outcome. The winning bidder sends pay-
ment to the seller. The seller then has the choice of
either “cooperating” (delivering a high-quality good)
or “cheating” (delivering a low-quality good). It is
easy to see that this game has a unique subgame-
perfect equilibrium. In equilibrium, the seller always
cheats (delivers low quality), buyers each bid 1, each
buyer’sexpectedpayoffis0,andtheseller’s expected
payoff is 1.
The ability to build a reputation allows the long-
run player to improve his payoffs in such settings.
Intuitively, a long-run player with a track record of
playing a given action (e.g., cooperate) often enough
in the past acquires a reputation for doing so and is
“trusted”bysubsequentshort-runplayerstodosoin
thefuture.However,whywouldaprofit-maximizing,
long-term playerbe willing to behave in such a way,
and why would rational short-term players use past
history as an indication of future behavior?
To explain such phenomena, Kreps et al. (1982),
Kreps and Wilson (1982), and Milgrom and Roberts
(1982) introducedthe notion of “commitment” types.
Commitment types are long-run players who are
locked into playing the same action.6 An important
subclass of commitment types are Stackelberg types:
long-run players who are locked into playing the
so-calledStackelberg action.The Stackelbergaction is
6Commitmenttypesaresometimesalsoreferred toas“irrational”
typesbecausetheyfollowfixed,“hard-wired”strategiesasopposed
to “rational” profit-maximizing strategies. An alternative way to
justify suchplayers is to consider them as players with nonstan-
dardpayoff structuressuchthatthe“commitment” actionis their
dominantstrategy giventheirpayoffs.
1414
ManagementScience/Vol.49, No. 10, October 2003
DELLAROCAS
TheDigitization of Wordof Mouth
the action to which the long-run player would credi-
blycommitif he could.Inthe “onlineauction” exam-
ple, the Stackelberg action would be to cooperate;
cooperation is the action that maximizes the seller’s
lifetime payoffs if theseller could credibly commit to
an action for the entire duration of the game.7 There-
fore,theStackelbergtypeinthisexamplecorresponds
to an “honest” seller who never cheats. In contrast,
an “ordinary” or “strategic” type corresponds to an
opportunistic seller who cheatswhenever itis advan-
tageous forhim to do so.
Reputation models assume that short-run players
know that commitment types exist, but are ignorant
ofthetypeofplayertheyface.An additional assump-
tion is that short-run playershave access totheentire
historyofpast stage gameoutcomes.8 Aplayer’s rep-
utation at any given time then consists of the condi-
tional posterior probabilities over that player’s type,
given a short-run player’s prior over types and the
repeated application of Bayes’ rule on the history of
past stage game outcomes.
In such a setting, when selecting his next move,
the informedplayer musttakeinto account, notonly
his short-term payoff, but also the long-term con-
sequences of his action based on what that action
revealsabout his type tootherplayers.As longasthe
promised future gains due to the increased (or sus-
tained)reputation thatcomes from playingtheStack-
elbergaction offset whatevershort-termincentives he
might have to play otherwise, the equilibrium strat-
egyforan“ordinary”informedplayerwillbetotryto
“acquire a reputation” bymasqueradingas a Stackel-
bergtype (i.e., repeatedly play the Stackelberg action
with high probability).
7Ifthesellercouldcommittocooperation(productionofhighqual-
ity), buyers would then each bid 2 and the seller’s expected per
periodpayoffwouldbe2.
8The traditional justificationforthisassumptionisthat past out-
comes are either publicly observable or explicitly communicated
among short-runplayers.Theemergenceofonlinefeedback mech-
anisms provides, of course, yet another justification (but see the
discussionofcomplicationsarisingfromtheprivateobservabilityof
outcomes insuchsystemsin§5.2).
In the “online auction” example, if the promised
future gains of reputation effects are high enough,9
ordinary sellers are induced to overcome their short-
term temptation to cheat and to try to acquire a
reputation for honesty by repeatedly delivering high
quality. Expecting this, buyers then place high bids,
thus, increasing theseller’s long-term payoffs.
In general, reputation effects benefit the most
patient player in the game: The player who has the
longesttimehorizon (discounts futurepayoffs less) is
usually the one who is able to reap the benefits of
reputation. Fudenberg and Levine (1992) show that
this result holds even when players can observe only
noisysignals of eachother’sactions,sothatthegame
has imperfect public monitoring. They prove that, if
short-run players assign positive prior probability to
the long-run player being a Stackelberg type, and if
that player is sufficiently patient, then an ordinary
long-run player achieves an average discounted pay-
off close to his commitment payoff (i.e., his payoff if
he could credibly commit to the Stackelberg action).
Toobtain this payoff,theordinaryplayerspendslong
periods of time choosing the Stackelberg action with
high probability.10
4.2. ReputationDynamics
In most settings where reputation phenomena arise,
equilibrium strategies evolve over time as informa-
tion slowly leaks out about the types of the vari-
ous players. In general, thederivation of closed-form
solutions in repeated games with reputation effects
is complicated. Nevertheless, a small number of spe-
cificcaseshavebeenstudied.Theyprovideinteresting
insight into the complex behavioral dynamics intro-
ducedby reputational considerations.
Initial Phase. In most cases, reputation effects
begin to work immediately and, in fact, are strongest
duringtheinitial phasewhenplayersmustworkhard
9Inthistypeofgame,thisrequiresthattheremaininghorizonof
the seller is long enough, and that the profit margin of a single
transactionishighenough,relativeto thediscountfactor.
10This
result also requires that the stage game is either a
simultaneous-movegameor in asequential-move game, that the
short-run players always observe whether or not theStackelberg
strategy hasbeenplayed.
Management Science/Vol. 49, No. 10, October2003
1415
Documents you may be interested
Documents you may be interested