A survey of trust and reputation systems for online service provision

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

There is a rapidly growing literature around trust and reputation systems, but the activity is not very coherent.

“There is also a rapidly growing literature around trust and reputation systems, but unfortunately this activity is not very coherent.”

Probably not a substantial claim: may not pass S4 (based).

There is considerable confusion around the terminology used to describe trust and reputation systems.

“There is considerable confusion around the terminology used to describe these systems”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Reputation systems are already being used in successful commercial online applications.

“Reputation systems are already being used in successful commercial online applications.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Authors often propose new trust and reputation systems from scratch without extending previous proposals, indicating a lack of coherence.

“There also seems to be a lack of coherence in this area, as indicated by the fact that authors often propose new systems from scratch, without trying to extend and enhance previous proposals.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

P2P networks introduce security threats, as they can spread malicious software and bypass firewalls.

“P2P networks introduce a range of security threats, as they can be used to spread malicious software, such as viruses and Trojan horses, and easily bypass firewalls.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Most experts on AllExperts receive ratings close to 10 on all aspects.

“Most experts receive ratings close to 10 on all aspects, so the General Prestige is usually close to 10×the number of questions received.”

May combine two claims.

Most analysed reputation systems compute scores by summation or average of ratings, while two use the flow model.

“The computation is mostly based on the summation or average of ratings, but two systems use the flow model.”

May combine two claims.

On eBay, less than 1% of ratings are negative, less than 0.5% are neutral, and about 99% are positive.

“Of all ratings provided, less than 1% is negative, less than 0.5% is neutral and about 99% is positive.”

• cites [55] Resnick, 2002, Trust among strangers in internet transactions: empirical analysis of eBay's reputation systemNot checked yet

  Not checked in the free run.

May combine two claims.

On eBay, buyers rate sellers 51.7% of the time and sellers rate buyers 60.6% of the time.

“Buyers provide ratings about sellers 51.7% of the time, and sellers provide ratings about buyers 60.6% of the time [55].”

• cites [55] Resnick, 2002, Trust among strangers in internet transactions: empirical analysis of eBay's reputation systemNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

In Slashdot's early days when the community was small, the signal to noise ratio was very high.

“In the early days when the community was small, the signal to noise ratio was very high.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Reviewers who reached Amazon Top 100 have reported a sudden increase in negative votes, reflecting a cat fight to get into top reviewer ranks.

“Also, reviewers who have reached the Top 100 rank have reported a sudden increase in negative votes which reflects that there is a cat fight taking place in order to get into the ranks of top reviewers.”

May combine two claims.

Commercial and live implementations have settled around relatively simple schemes, while the academic community proposes many systems with advanced features.

“The commercial and live implementation seems to have settled around relatively simple schemes, whereas a multitude of different systems with advanced features are being proposed by the academic community.”

Probably not a substantial claim: may not pass S4 (based).

The longevity of agents means it should be impossible or difficult for an agent to change identity or pseudonym to erase past behaviour.

“The longevity of agents means, for example, that it should be impossible or difficult for an agent to change identity or pseudonym for the purpose of erasing the connection to its past behaviour.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Transitivity is explicit in trust systems, whereas reputation systems usually only take transitivity implicitly into account.

“Secondly, transitivity is an explicit component in trust systems, whereas reputation systems usually only take transitivity implicitly into account.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Trust systems usually take subjective and general measures of reliability trust as input, whereas reputation systems use ratings about specific and objective events such as transactions.

“Finally, trust systems usually take subjective and general measures of (reliability) trust as input, whereas information or ratings about specific (and objective) events, such as transactions, are used as input in reputation systems.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The act of referring negative general and subjective trust in an entity can lead to accusations of slander.

“Another potential problem is that the act of referring negative general and subjective trust in an entity can lead to accusations of slander.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Some authors have proposed discrete trust models.

“This is also valid for determining trust measures, and some authors, including [1,8,9,44], have proposed discrete trust models.”

• cites [1] Abdul-Rahman, 2000, Supporting trust in virtual communities sourceNot checked yet

  Not checked in the free run.

• cites [8] Buchegger, 2003, The effect of rumor spreading in reputation systems for mobile ad-hoc networksNot checked yet

  Not checked in the free run.

• cites [9] Cahill, 2003, Using trust for secure collaboration in uncertain environments, Pervasive Computing sourceNot checked yet

  Not checked in the free run.

• cites [44] Manchala, 1998, Trust metrics, models and protocols for electronic commerce transactions sourceNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

The term collaborative sanctioning has been used for reputation systems because their purpose is to sanction poor service providers and incentivize quality services.

“In this sense the term”

• cites [49] Mui, 2001, Ratings in distributed systems: a Bayesian approachNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Reputation systems typically use public information to reflect the community's opinion.

“Reputation systems are typically based on public information in order to reflect the community's opinion in general, which is in line with Def.3 of reputation.”

May combine two claims.

Most reputation systems cannot distinguish between variations in service provider performance and variations in the observer's taste, which can lead to unreliable and misleading reputation scores.

“most reputation systems will be unable to distinguish between variations in service provider performance, and variations in the observer's taste, potentially leading to unreliable and misleading reputation scores.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Some reputation systems take both public and private information as input.

“Some systems take both public and private information as input.”

May combine two claims.

Reputation systems provide an incentive for good behaviour and tend to have a positive effect on market quality.

“A natural side effect is that it also provides an incentive for good behaviour, and therefore tends to have a positive effect on market quality.”

• Also stated in: § 4 Collaborative filtering and Collaborative Sanctioning · plain

  “the purpose is to sanction poor service providers, with the aim of giving an incentive for them to provide quality services.”

• cites [49] Mui, 2001, Ratings in distributed systems: a Bayesian approachNot checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

It is often hard to assess the trustworthiness of remote entities because computerised communication media remove familiar interaction cues.

“However, it is often hard to assess the trustworthiness of remote entities, because computerised communication media are increasingly removing us from familiar styles of interaction.”

May combine two claims.

For transitive trust derivation to be valid, Alice must trust Bob to recommend Claire for a particular purpose, and Bob must trust Claire for that same purpose.

“In addition there are semantic constraints for the transitive trust derivation to be valid, i.e. that Alice must trust Bob to recommend Claire for a particular purpose, and that Bob must trust Claire for that same purpose [33].”

• cites [33] Jøsang, 2005, Semantic constraints for trust transitivityNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

In trust transitivity, if Alice trusts Bob, Bob trusts Claire, and Bob refers Claire to Alice, Alice can derive a measure of trust in Claire based on Bob's referral combined with her trust in Bob.

“The idea behind trust transitivity is that when Alice trusts Bob, and Bob trusts Claire, and Bob refers Claire to Alice, then Alice can derive a measure of trust in Claire based on Bob's referral combined with her trust in Bob.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The second property also depends on participants' willingness to provide ratings, for which there must be some incentive.

“The second property also depends on the willingness of participants to provide ratings, for which there must be some form of incentive.”

May combine two claims.

The second property of reputation systems depends on the rating protocol; usually not a problem for centralised systems but a major challenge for distributed systems.

“The second property depends on the protocol with which ratings are provided, and this is usually not a problem for centralised systems, but represents a major challenge for distributed systems.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Trust and reputation scores can be computed from own experience, second-hand referrals, or a combination of both.

“Seen from the relying party's point of view, trust and reputation scores can be computed based on own experience, on second hand referrals, or on a combination of both.”

Probably not a substantial claim: may not pass S3 (load-bearing).

In distributed reputation systems, direct experience with a target party can be taken into account as private information, possibly carrying a higher weight than received ratings.

“In case the relying party has had direct experience with the target party, the experience from that encounter can be taken into account as private information, possibly carrying a higher weight than the received ratings.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Objective measures have the advantage that the correctness of ratings can be verified by others or automatically generated based on automated monitoring of events.

“An advantage with objective measures is that the correctness of ratings can be verified by others, or automatically generated based on automated monitoring of events.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Trust systems produce a score reflecting the relying party's subjective view of an entity's trustworthiness, whereas reputation systems produce a public reputation score as seen by the whole community.

“Trust systems produce a score that reflects the relying party's subjective view of an entity's trustworthiness, whereas reputation systems produce an entity's (public) reputation score as seen by the whole community.”

Probably not a substantial claim: may not pass S3 (load-bearing).

A party relying on a remote party's reputation score is trusting that party through trust transitivity.

“A party who relies on the reputation score of some remote party, is in fact trusting that party through trust transitivity [33].”

• cites [33] Jøsang, 2005, Semantic constraints for trust transitivityNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

In centralised reputation systems, transactions with reputable participants are likely to result in more favourable outcomes than transactions with disreputable participants.

“The idea is that transactions with reputable participants are likely to result in more favourable outcomes than transactions with disreputable participants.”

May combine two claims.

Trust and reputation systems can protect against deceitful information providers, which traditional security mechanisms cannot.

“Trust and reputation systems on the other hand can provide protection against such threats.”

Probably not a substantial claim: may not pass S4 (based).

Physical encounter and traditional communication allow people to assess a wider range of trustworthiness cues than computer mediated communication.

“Physical encounter and traditional forms of communication allow people to assess a much wider range of cues related to trustworthiness than is currently possible through computer mediated communication.”

Probably not a substantial claim: may not pass S4 (based).

Only trust and reputation systems, as soft security mechanisms, are useful for deriving provision trust.

“only trust and reputation systems (i.e.soft security mechanisms) are useful tools for deriving provision trust.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

Personal experience typically carries more weight than second hand trust referrals or reputation, but in the absence of personal experience, trust often has to be based on referrals from others.

“Personal experience typically carries more weight than second hand trust referrals or reputation, but in the absence of personal experience, trust often has to be based on referrals from others.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Bayesian systems may be too complex for average persons to understand.

“The advantage of Bayesian systems is that they provide a theoretically sound basis for computing reputation scores, and the only disadvantage that it might be too complex for average persons to understand.”

Probably not a substantial claim: may not pass S2 (checkable).

Trust plays a crucial role in computer mediated transactions and processes.

“This example shows that trust plays a crucial role in computer mediated transactions and processes.”

Probably not a substantial claim: may not pass S3 (load-bearing).

A general problem with all subjective measures is that it is difficult to protect against unfair ratings.

“A general problem with all subjective measures is that it is difficult to protect against unfair ratings.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Reputation systems assume that some participants will misrepresent service quality for profit and provide misleading ratings.

“Reputation systems, on the other hand, assume that some participants will try to misrepresent the quality of services in order to make more profit, and to lie or provide misleading ratings in order to achieve some specific goal.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Collaborative filtering systems assume all participants are trustworthy and sincere, reporting their genuine opinion.

“To be specific CF systems assume all participants to be trustworthy and sincere, i.e. to their job as best they can and to always report their genuine opinion.”

Probably not a substantial claim: may not pass S3 (load-bearing).

An individual's subjective trust can be derived from a combination of received referrals and personal experience.

“An individual's subjective trust can be derived from a combination of received referrals and personal experience.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Identity trust is a condition for trusting a party behind the identity with more than baseline provision trust.

“It can be observed that identity trust is a condition for trusting a party behind the identity with anything more than a baseline or default provision trust that applies to all parties in a community.”

Probably not a substantial claim: may not pass S3 (load-bearing).

A slightly more advanced scheme computes reputation as the average of all ratings.

“A slightly more advanced scheme proposed in e.g.[63] is to compute the reputation score as the average of all ratings, and this principle is used in the reputation systems of numerous commercial web sites, such as Epinions, and Amazon described in Sec.9.”

• cites [63] Schneider, 2000, Disseminating trust information in wearable communities sourceNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Reputation systems are based on the assumption that all community members should judge a transaction partner's performance or product quality consistently.

“This must not be confused with reputation systems which are based on the seemingly opposite assumption, namely that all members in a community should judge the performance of a transaction partner or the quality of a product or service consistently.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Bayesian systems provide a theoretically sound basis for computing reputation scores.

“The advantage of Bayesian systems is that they provide a theoretically sound basis for computing reputation scores, and the only disadvantage that it might be too complex for average persons to understand.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

According to Resnick et al., reputation systems must have three properties to operate at all: entities must be long lived, ratings about current interactions must be captured and distributed, and ratings about past interactions must guide decisions about current interactions.

“According to Resnick et al. [56], reputation systems must have the following three properties to operate at all:”

• cites Resnick et al. [56] Resnick, 2000, Reputation systems, Communications of the ACM sourceNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

P2P networks represent an environment well suited for distributed reputation management.

“Peer-to-Peer (P2P) networks represent a environment well suited for distributed reputation management.”

Probably not a substantial claim: may not pass S3 (load-bearing).

PageRank can be described as a reputation system because hyperlinks to a page are public information that can be combined to derive a reputation score.

“This can be described as a reputation system, because the collection of hyperlinks to a given page can be seen as public information that can be combined to derive a reputation score.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Assuming the only way to improve a page's PageRank is to buy domain names, Clausen (2004) shows a lower bound to the cost of obtaining arbitrarily good public PageRank.

“Assuming that the only way to improve a page's PageRank is to buy domain names, Clausen (2004) [11] shows that there is a lower bound to the cost of obtaining an arbitrarily good PR.”

• cites [11] Chen, 2001, Computing and using reputations for internet ratingsNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

AskMe does not publicly provide details of how its system works.

“Ask Me does not publicly provide any details of how the system works.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Belief theory is a framework related to probability theory where probabilities need not sum to one and the remaining probability is interpreted as uncertainty.

“Belief theory is a framework related to probability theory, but where the sum of probabilities over all possible outcomes not necessarily add up to 1, and the remaining probability is interpreted as uncertainty.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Google does not specify exactly how public PageRank is computed.

“Although Google do not specify exactly how the public PageRank is computed, the source rank vector E can be defined over the root web pages of all domains weighted by the cost of buying each domain name.”

May combine two claims.

Epinions reviewers can potentially earn as much for a positive review as for a negative review, which avoids incentivizing biased reviews for profit.

“Reviewers can potentially earn as much for helping someone make a buying decision with a positive review, as for helping someone avoid a purchase with a negative review.This is important in order not to give an incentive to write biased reviews just for profit.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Slashdot's reputation system directs and stimulates massive collaborative moderation effort.

“The Slashdot reputation system directs and stimulates the massive collaborative effort of moderating thousands of postings every day.”

May combine two claims.

Manchala's scheme and the REGRET reputation system fall in the fuzzy models category.

“The scheme proposed by Manchala (1988) [44] described in Sec.2 as well as the REGRET reputation system proposed by Sabater & Sierra (2001,2002) [59-61] fall in this category.”

• cites [44] Manchala, 1998, Trust metrics, models and protocols for electronic commerce transactions sourceNot checked yet

  Not checked in the free run.

• cites [59] Sabater, 2001, REGRET: a reputation model for gregarious societiesNot checked yet

  Not checked in the free run.

• cites [60] Sabater, 2002, Reputation and social network analysis in multi-agent systems sourceNot checked yet

  Not checked in the free run.

• cites [61] Sabater, 2002, Social ReGreT, a reputation model based on social relations, SIGecom ExchangesNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Some theoretical models also permit looped and/or infinite transitivity.

“Some theoretic models including [35,38,73] do also allow looped and/or infinite transitivity.”

• cites [35] Jurca, 2003, An incentive compatible reputation mechanismNot checked yet

  Not checked in the free run.

• cites [38] Kurbel, 2001, A framework for multi-agent electronic marketplaces: analysis and classification of existing systemsNot checked yet

  Not checked in the free run.

• cites [73] Ziegler, 2004, Spreading activation models for trust propagation sourceNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Altavista's proposal to combine keywords using binary logic was too complex for users and not a good solution.

“Altavista's proposal for handling this problem was to offer advanced ways to combine keywords based on binary logic.This was too complex for users and therefore did not represent a good solution.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

Kuro5hin's problems could possibly have been avoided by using Slashdot's principle of only allowing longstanding members to moderate.

“It is possible that the problems experienced by Kuro5hin could have been avoided had they used Slashdot's principle of only allowing longstanding members to moderate because throwaway accounts would have been less effective as an attack tool.”

May combine two claims.

Kuro5hin's Mojo underwent major changes in October 2003 because it could not effectively counter noise postings from throwaway accounts and because attackers rated down targeted members' comments to make them lose reputation.

“It underwent major changes in October 2003 because it was unable to effectively counter noise postings from throwaway accounts, and because attackers rated down comments of targeted members in order to make them lose their reputation scores.”

Probably not a substantial claim: may not pass S3 (load-bearing).

PageRank ranks search results based on a page's reputation, roughly according to how many other pages point at it.

“PageRank proposed by Page et al. (1998) [52] represents a way of ranking the best search results based on a page's reputation.Roughly speaking, PageRank ranks a page according to how many other pages are pointing at it.”

• cites [52] L. Page, S. Brin, R. Motwani, T. Winograd, The PageRank Citation Ranking: Bringing Order to the Web. Technical report, Stanford Digital Library Technologies Project, 1998Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Some flow models assume a constant trust/reputation weight for the whole community, which can be distributed among members.

“Some flow models assume a constant trust/reputation weight for the whole community, and this weight can be distributed among the members of the community.”

May combine two claims.

PageRank alone cannot blacklist web pages because it only permits positive referrals and has no negative referrals.

“Negative referrals do not exist in PageRank so that it is impossible to blacklist web pages with the PageRank algorithm of Eq.( 4) alone.”

Probably not a substantial claim: may not pass S3 (load-bearing).

PageRank appears to have reduced keyword-stuffing spam because a page now also needs a high PageRank, not only matching keywords, to be shown to users.

“Although this still can occur, PageRank seems to have reduced that problem because a high PR is also needed in addition to matching key words in order for a page to be presented to the user.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Early web search engines like Altavista presented every page matching keywords, often resulting in too many irrelevant pages.

“The early web search engines such as Altavista simply presented every web page that matched the key words entered by the user, which often resulted in too many and irrelevant pages being listed in the search results.”

May combine two claims.

In flow models, a participant's reputation generally increases with incoming flow and decreases with outgoing flow.

“In general, a participant's reputation increases as a function of incoming flow, and decreases as a function of outgoing flow.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Discrete measures do not easily lend themselves to sound computational principles, so heuristic mechanisms like look-up tables must be used.

“The disadvantage of discrete measures is that they do not easily lend themselves to sound computational principles.Instead, heuristic mechanisms like look-up tables must be used.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

As mailing lists and discussion fora grow rapidly, spam and low quality postings become a major problem, forcing Slashdot to introduce moderation.

“As is the case with all mailing lists and discussion fora where the number of members grow rapidly, spam and low quality postings emerged to become a major problem, and this forced Slashdot to introduce moderation.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Advogato's reputation system has no direct purpose other than boosting members' egos and stimulating social and professional networking.

“The Advogato reputation systems does not have any direct purpose other than to boost the ego of members, and to be a stimulant for social and professional networking within the Advogato community.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

According to the PageRank definition, R is in [0,1], but public PageRank values are scaled to [0,10] in increments of 0.25.

“According to Def.4 above R ∈ [0, 1], but the PageRank values that Google provides to the public are scaled to the range [0,10] in increments of 0.25.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Google's rapidly rising popularity at the cost of Altavista was caused by the superior search results delivered by PageRank.

“and the rapidly rising popularity of Google at the cost of Altavista was obviously caused by the superior search results that the PageRank algorithm delivered.”

Classified as probably not a substantial claim. May combine two claims.

Information asymmetry and uncertainty about supplier reliability can make online supply chain and subcontract agreements risky.

“However as described in Sec.1 the problem of information asymmetry and uncertainty about supplier reliability can make it risky to establish supply chain and subcontract agreements online.”

May combine two claims.

Opinions can be uniquely mapped to beta probability density functions, making the consensus operator equivalent to Bayesian updating.

“Opinions can be uniquely mapped to beta PDFs, and this sense the consensus operator is equivalent to the Bayesian updating described in Sec.8.2.”

May combine two claims.

Ballot stuffing seems to be a minor problem on eBay because participants can only rate each other after a monitored transaction.

“The problem of ballot stuffing, i.e. that ratings can be repeated many times, e.g. to unfairly boost somebody's reputation score, seems to be a minor problem on eBay because participants are only allowed to rate each other after the completion of a transaction, which is monitored by eBay.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Jøsang proposed a belief/trust metric called opinion.

“Jøsang (1999Jøsang ( ,2001) ) [28,29] has proposed a belief/trust metric called opinion denoted by ω A x = (b, d, u, a), which expresses the relying party A's belief in the truth of statement x.”

• cites [28] Ismail, 2003, Strong privacy in reputation systemsNot checked yet

  Not checked in the free run.

• cites [29] Jøsang, 1999, Trust-based decision making for electronic transactionsNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Slashdot introduced M2 metamoderation to address unfair moderations.

“To address the issue of unfair moderations, Slashdot has introduced a metamoderation layer called M2, (the moderation layer described above is called M1) with the purpose of moderating the M1 moderators.”

Probably not a substantial claim: may not pass S3 (load-bearing).

PageRank applies trust transitivity to an extreme degree, since rank values can pass through looped or arbitrarily long hyperlink chains.

“PageRank applies the principle of trust transitivity to the extreme because rank values can flow through looped or arbitrarily long hyperlink chains.”

May combine two claims.

BizRate's Customer Certified Merchant scheme does not capture frustrated customers who give up before checkout, so it tends to provide a positive bias of web stores.

“This scheme does not capture the frustrated customers who give up before they reach the check, and therefore tends to provide a positive bias of web stores.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Flow models compute trust or reputation by transitive iteration through looped or arbitrarily long chains.

“Systems that compute trust or reputation by transitive iteration through looped or arbitrarily long chains can be called flow models.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Trust and reputation can be represented as linguistically fuzzy concepts with membership functions.

“Trust and reputation can be represented as linguistically fuzzy concepts, where membership functions describe to what degree an agent can be described as e.g.trustworthy or not trustworthy.”

Probably not a substantial claim: may not pass S3 (load-bearing).

In PageRank, the source rank vector E is recommended to be chosen such that the sum of E(u) is 0.15.

“In [52] it is recommended that E be chosen such that u∈P E(u) = 0.15.”

• cites [52] L. Page, S. Brin, R. Motwani, T. Winograd, The PageRank Citation Ranking: Bringing Order to the Web. Technical report, Stanford Digital Library Technologies Project, 1998Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Epinions is a product and shop review site with a business model mainly based on cost-per-click online marketing.

“Epinions 24 founded in 1999 is a product and shop review site with a business model mainly based on so-called cost-per-click online marketing, which means that Epinions charges product manufacturers and online shops by the number of clicks consumers generate as a result of reading about their products on Epinions' web site.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Yu and Singh proposed using belief theory to represent reputation scores.

“Yu & Singh (2002) [70] have proposed to use belief theory to represent reputation scores.”

• cites [70] Yu, 2002, An evidential model of distributed reputation managementNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Amazon's #1 Reviewer usually posts more reviews than any living person could read, indicating group effort presented as one person's work.

“For example the Amazon #1 Reviewer usually is somebody who posts more reviews than any living person could possibly do if it would require that person to read each book, thus indicating that the combined effort of a group of people, presented as a single person's work, is needed to get to the top.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Amazon's reputation scheme is quite similar to BizRate's.

“Amazon's reputation scheme is quite similar to the one BizRate uses.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Epinions does not publish exact thresholds for becoming Top Reviewer or Advisor in order to discourage manipulation.

“Epinions does not publish the exact thresholds for becoming Top Reviewer or Advisor, in order to discourage members from trying to manipulate the selection process.”

May combine two claims.

Because ratings benefit the community, free-riding is possible when individuals let others provide the ratings.

“It serves the community to provide ratings and the potential for free-riding (i.e.letting the others provide the ratings) therefore exists.”

Probably not a substantial claim: may not pass S3 (load-bearing).

A party that has suffered a significant loss of reputation may change identity or pseudonym to start fresh.

“In case a party has suffered significant loss of reputation it might be in his interest to change identity or pseudonym in order to cut with the past and start from fresh.”

• AssumptionThe authors state this as an assumption: Reputation systems rely on identities and pseudonyms being long-lived so that past ratings about a party can be linked to the same party in the future.

May combine two claims.

eBay's Feedback forum appears to protect adequately against ballot stuffing because ratings can only be given after transaction completion and transaction fees make stuffing expensive.

“Among the commercial and live reputation systems, eBay's Feedback forum seems to provide adequate protection against ballot stuffing, because ratings can only be provided after transactions completion.Because eBay charges a fee for each transaction ballot stuffing would be expensive.”

Probably not a substantial claim: may not pass S3 (load-bearing).

When a relying party has private information, comparing it with public information can indicate how reliable the public information is.

“If the relying party has private information, then this information can be compared to public information in order to give an indication of the reliability of the public information.”

• AssumptionRests on an assumption: Private information from personal experience is normally more reliable than public information such as ratings from third parties.

Probably not a substantial claim: may not pass S3 (load-bearing).

From a business perspective, a non-robust reputation system can be desirable if it generally gives a positive bias, because positively biased ratings are more likely to promote sales.

“Another explanation is that, from a business perspective, having a reputation system that is not robust can be desirable if it generally gives a positive bias.After all, commercial web stores are in the business of selling, and positively biased ratings are more likely to promote sales than negative ratings.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Ratings are usually subjective, which opens the possibility of unfair ratings, but objective criteria would make high robustness much simpler to achieve.

“Ratings are usually based on subjective judgement, which opens up the Pandora's box of unfair ratings, but if ratings can be based on objective criteria it would be much simpler to achieve high robustness.”

May combine two claims.

Because commercial and online reputation systems have serious vulnerabilities, their reliability is sometimes questionable.

“Given that reputation systems used in commercial and online applications have serious vulnerabilities, it is obvious that the reliability of these systems sometimes is questionable.”

May combine two claims.

Even though a theoretical equilibrium exists, fluctuations will always occur, and conditions can be characterised under which oscillations are avoided or converge to equilibrium.

“Although a theoretic equilibrium exists, there will always be fluctuations, and it is possible to characterise the conditions under which oscillations can be avoided [65] or converge towards the equilibrium [26].”

• cites [65] Shapiro, 1982, Consumer information, product quality, and seller reputation, The Bell Journal of Economics sourceNot checked yet

  Not checked in the free run.

• cites [26] Hood, 2001, The Literature of Bibliometrics, Scientometrics, and Informetrics, Scientometrics sourceNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). The quote only partly states this claim.

Methods for avoiding bias from unfair ratings can be broadly divided into endogenous and exogenous discounting.

“The methods of avoiding bias from unfair ratings can broadly be grouped into two categories described below.”

May combine two claims.

Discriminatory behaviour can occur both when services are provided and when ratings are provided.

“Discriminatory behaviour can occur both when providing services and when providing ratings.”

Probably not a substantial claim: may not pass S2 (checkable).

Criterion (3) will probably never be solved completely because new unforeseen attacks will always arise.

“Criterion (3) on the other hand will probably never be solved completely because there will always be new and unforeseen attacks for which solutions will have to be found.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Some exogenous discounting techniques would theoretically protect against this discrimination, but no simulations have proved it.

“Some of the techniques described in Sec.10.3.2 would theoretically be able to protect against this type of discrimination, but no simulations have been done to prove this.”

• AssumptionThe authors state this as an assumption: Raters with low reputation are likely to give unfair ratings, and raters with high reputation are not.

May combine two claims.

Resnick and Zeckhauser found that only 0.6% of buyer ratings and 1.6% of seller ratings were negative.

“In Resnick & Zeckhauser (2002) [55], it was found that only 0.6% of all the ratings provided by buyers and only 1.6% of all the ratings provided by sellers were negative, which seems too low to reflect reality.”

• cites Resnick & Zeckhauser (2002) [55]Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Unfair ratings and ballot stuffing are probably the hardest problems to solve in reputation systems based on subjective ratings from participants.

“The problems of unfair ratings and ballot stuffing are probably the hardest to solve in any reputation system that is based on subjective ratings from participants, and large number of researchers are working on this in the academic community.”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

Any reputation system with user participation depends on how people respond to it and must be designed with that dependence in mind.

“Any reputation system with user participation will depend on how people respond to it, and must therefore be designed with that in mind.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

Exceptional manual control will probably always be needed if a reputation system comes under heavy attack.

“Exceptional manual control will probably always be needed, should the system come under heavy attack.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

The current period can be seen as a period of pioneers.

“The period we are in can therefore be seen as a period of pioneers”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Scientometrics usually give reasonable indicators of quality and reputation, but they can sometimes produce misleading results.

“As pointed out by Makino (1998) [43], even though scientometrics normally provide reasonable indicators of quality and reputation, it can sometimes give misleading results.”

• cites Makino (1998) [43]Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Ballot stuffing is closely related to unfair ratings because it usually consists of too many unfair ratings.

“This problem is closely related to unfair ratings because ballot stuffing usually consists of too many unfair ratings.”

May combine two claims.

Ratings are usually submitted after a transaction, and the parties involved typically have no direct incentive to rate one another.

“Ratings are typically provided after a transaction has taken place, and the transaction partners usually have no direct incentive for providing rating about the other party.”

Classified as probably not a substantial claim.

Rather than one solution that works in all situations, there will be multiple techniques with advantages, disadvantages, and trade-offs.

“Instead of having one solution that works well in all situations there will be multiple techniques with advantages, disadvantages and trade-offs.”

May combine two claims.

Ballot stuffing is a common problem in many online reputation systems, and those systems usually have poor protection against it.

“This is a common problem in many online reputation systems described in Sec.9 and they usually have poor protection against it.”

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

The lack of incentives to provide ratings is a general problem that needs attention and may require specific incentive mechanisms.

“However lack of incentives for providing ratings is a general problem that needs special attention and that might require specific incentive mechanisms.”

Classified as probably not a substantial claim. May combine two claims.

Academic proposals for trust and reputation systems so far lack coherence, are usually designed from scratch, and only rarely build on other authors' proposals.

“A general observation is that the proposals from the academic community so far lack coherence.The systems being proposed are usually designed from scratch, and only in very few cases are authors building on proposals by other authors.”