Click to Judge

The dramatic real-life consequences of online ranking, posts and algorithms.

It is unusual to see a work that treats reputation as a topic in its own right, but reputation matters and merits our attention. We rely on reputation—the social distillation of opinion about an individual, product or service—to decide whom to trust or what to buy, and our own reputation precedes us, shaping our interactions and opportunities as it does so. Reputation is not only important, but problematic; it is circulated through formal channels such as credit scores and also through erratic, unaccountable channels of gossip and rumour, breeding grounds for discrimination and incrimination.

The internet can magnify both the formal and informal channels for conveying reputation, in ways that are sometimes useful, sometimes dismaying. In this collection of essays, Lior Strahilevitz asks us to “imagine if every plumber, manufactured product, cell phone provider, home builder, professor, hair stylist, accountant, attorney, golf pro, and taxi driver were rated.” Strahilevitz sees many potential benefits: “In such a world, there would be diminished need for regulatory oversight and legal remedies because consumers would police misconduct themselves.” But others might be reminded of Gary Shteyngart’s 2010 novel Super Sad True Love Story, in which patrons in bars and clubs rate each other, casually yet uncomfortably, on the scale of “fuckability.”

We are in neither world just yet, but the spectrum of possibilities is becoming clear. From the widespread acceptance of online ratings for books and movies to events such as the suicide of American teenager Tyler Clementi after roommates recorded a video of his date with another male student and posted it online, there is no doubt that online reputation is increasingly “reshaping the offline world.”

Back in 2004, when eBay and Amazon were just beginning to shake up retail markets, an ambitious and opinionated “Manifesto for the Reputation Society” appeared in the online journal First Monday. Authors Hassan Masum and Yi-Cheng Zhang saw the innovations of these internet giants, aggregating the opinions of millions of buyers and sellers on their websites, as harbingers of greater things to come: “the potential utility of reputation services is far greater, touching nearly every aspect of society.”

Much has changed since that time. Algorithms for matching people and tastes have become intricate and sophisticated; commercial websites now assemble tens of millions of reviews of everything from restaurants to university lecturers; reputations are made and reputations ruined, sometimes tragically, by photographs, videos and gossip posted to social networking sites that did not exist when the manifesto was published. So I leapt at the chance to review a new volume co-edited by one of its authors. Let me be honest: despite—or perhaps because of—working in the computer software industry for the last 20 years, I expected to disagree with much of this book, but I looked forward to wrestling with their new data, arguments and insights.

What I did not expect was to be bored, but unfortunately that is what happened. The book’s audience is ill defined; it contains little new research; and it lacks a coherent focus. Most seriously, essay after essay simply fails to tackle the thorny issues that continue to bedevil internet reputation systems.

Scattered throughout The Reputation Society are cautionary tales, spelling out the consequences of poor reputation management in the “offline world.” Fearing lawsuits, “most knowledgeable employers refuse to provide any subjective recommendations of former employees, positive or negative.” Landlords spend good money screening out tenants with bad reputations, where “bad” might include involvement in any kind of landlord–tenant dispute, no matter how minor. Credit scores are forms of reputation so important that governments around the world must regulate them; and the failure of financial rating agencies (which are reputation managers of a sort) was a major contributor to the 2008 financial collapse. Reputation can easily become, as John Henry Clippinger writes, “not so much a thing as an ongoing contest.”

Can digital technologies help us smooth the flow of reputation and help us decide more effectively “whom to trust and what to believe” in today’s world? Masum and Tovey believe that “properly designed reputation systems have the potential to reshape society for the better by shining the light of accountability into dark places, through the mediated judgments of billions of people worldwide.”

The story of travel company TripAdvisor illustrates both the promises and challenges of reputation systems. Since its launch in 2000, this influential website has collected more than 50 million anonymous reviews of hotels, flights and vacation destinations, and it now boasts 45 million visitors each month. Clearly something about the site is working, yet there are also problems.

As a result of its popularity, TripAdvisor’s ratings can now make or break hotels, resorts and lodgings around the world. One might think that fake reviews—owners either rating themselves generously or rating competitors harshly—would be an obvious issue, but in a 2010 interview with The Economist, CEO Stephen Kaufer dismissed the problem: “One or two phoney reviews: who gives a shit?”

Kaufer may be singing a different tune now. In September 2011 the Advertising Standards Authority in the United Kingdom opened a formal investigation (still ongoing in January) into TripAdvisor following a claim that there may be between five million and ten million fake reviews on the site. That makes TripAdvisor look bad, but before jumping to conclusions you should know that the claim was made by KwikChex, a “reputation management” company that obviously has its own agenda. Also, TripAdvisor has responded: it now places a red flag on the pages of hotels it believes to be posting (or paying for) fake reviews. Maybe it will come out of the investigation with its own reputation intact. Not so fast: in reputation as in physics, for every action there is an equal and opposite reaction. Some hotels flagged by TripAdvisor are protesting their innocence and crying foul in turn, suing the company for further damaging their business, while other hotels allege that guests are threatening to write bad reviews unless they receive special treatment. Reputation as contest indeed.

You will not find any of this in The Reputation Society; in fact, you find nothing beyond a paragraph about the workings and history of any customer-driven reputation system. Instead, the essays steadfastly avoid engaging seriously with conflict of interest, making only generic observations. Eric Goldman wonders if the marketplace itself will keep reputation systems honest, but his only answer is that “it remains an open question.” Randy Farmer writes that “good digital reputations should always be context-limited—the nature of the inputs should constrain the use of the reputation scores that are output.” In his take on “Managing Online Communities,” Cliff Lampe concludes that “reputation systems provide useful cues in constrained information channels and give feedback to users that might be difficult to detect in other ways.” Are your eyes glazing over yet? Essay after essay ends on a similar note. “Understanding and prioritizing a system’s business objectives should be a designer’s first task.” “Open and well-lit places for discussion may be a prerequisite to guiding the development of reputation systems in a positive direction.” Such pronouncements may have been the best one could do a decade ago, but with the multitude of real cases available today, they are no longer good enough.

In one of the few detailed case studies in the book Marc Maxson and Mari Kuraishi of GlobalGiving present their own “reputation system” for philanthropy: “a set of tools that guide donors in making funding decisions” to make the charitable sector more effective. Even though GlobalGiving is committed to technological approaches, its methods (along with others in the field such as GiveWell and Charity Intelligence Canada) end up owing more to traditional financial audits than to crowdsourced internet opinion. It is easy to see why: passing judgement on charitable projects demands verifiable information, not anonymous comments.

By 2007, [GlobalGiving] started vetting organizations from scratch. We began by determining whether an organization was equivalent to a U.S. 501(c)3 organization, as required by U.S. tax law … A document-based approach for vetting new members was defensible and minimized legal risk … In 2009, we started adding two crucial elements to our vetting process … We examined the organization’s social media footprint on the Internet and also required new organizations to raise fifty donations in the first month, thereby demonstrating that people support them … In 2010, we began experimenting with another approach: asking people in communities across Kenya to share a story about a “community effort” they had witnessed.

GlobalGiving’s story is not promising for the book’s thesis: “The evidence from our pilot activities around beneficiary feedback suggests that there is a challenge of scale—in particular, a challenge of finding a low-cost solution that will gather sufficient community feedback to provide a realistic view of community-based organizations.”

The authors continue undaunted in their commitment to technological approaches. To address the challenge of scale, GlobalGiving now encourages “Kenyan organizations to adopt Twitter and use it to build rapport with potential donors.” The authors then criticize a Kenyan project leader for failing to “turn his Twitter monologue into a conversation,” and conclude that “technological advances (Short Message Service—SMS) and technological penetration (six in ten people in the world have cell phones) are making new forms of interaction available, but there will have to be an accompanying shift in behavior to ensure sufficient, ongoing, and real-time feedback flow.” GlobalGiving’s belief in technological solutions is leading it to place new burdens on the recipients of philanthropy in order to build a reputation system.

As the most specific and mature case study in the whole volume, GlobalGiving’s continuing experimentation made this reader more skeptical than ever of algorithmic approaches to reputation systems.

A recent collection of essays edited by legal academics Saul Levmore and Martha Nussbaum, The Offensive Internet: Speech, Privacy and Reputation is a contrast to the optimistic tone of The Reputation Society.

The book was prompted in part by a chat room for American law students called AutoAdmit, widely read a few years ago. Brian Leiter claims of AutoAdmit that “half [of the conversations] had as their primary purpose racist, misogynistic, and anti-Semitic abuse or simply vicious harassment, defamation, and implied threats against named individuals, usually other law students.” While essays in The Reputation Society wonder how online communities can encourage appropriate behaviour and maintain community norms, the case of AutoAdmit raises questions about what to do when the norms of a community are themselves objectionable to the larger society.

Search engines such as Google act as “implicit” reputation systems. Instead of averaging explicit ratings, they pick out individual events or comments about an individual and highlight the most “relevant.” Search engines can shine an intense spotlight on the most dramatic or salacious comments about an individual. As AutoAdmit had a large audience, Google would pick up anonymous smears and make them “the top search results for anyone—a friend, a family member, a prospective employer, a new acquaintance—Googling their names.”

Internet-based reputation systems are too diverse to be simplistically labelled as good or bad, but it is surely undeniable that, along with many wonderful things and a remarkable ability to let us find information we need to know, “it is easy to spread false rumours about almost anyone” on the internet, and false rumours can gain traction through circulation among networks of people sharing similar views. Such problems prompt Frank Pasquale to observe that “although the emancipatory potential of digital connectivity is clear, critical Internet studies have illuminated its role in reinforcing old structures of unfair disadvantage and unearned privilege.”

Faced with cases such as AutoAdmit, debates about speech on the internet are becoming increasingly specific, moving away from generalities about freedom of speech, anonymity and privacy to more finely grained discussions about the responsibilities and rights of specific platforms and participants in information-sharing networks.

I was expecting these debates to be central topics of The Reputation Society, but it takes until essay 16 of 18, by Michael Zimmer and Anthony Hoffman, to address this elephant in the digital-reputation room, and once again the conclusion is frustratingly bland. “Future information-sharing platforms and reputational systems can be better designed. Users should be provided with greater control over their personal information flows, with particular attention paid to the contextual nature of information sharing practices.” By this time I was ready to throw the book against the wall. Any serious attempt to discuss reputation systems must surely address the questions of who is going to provide this control, who is going to make them do it, and what forms this control could take. The authors do not even attempt an answer. I was left to wonder if the fact that only two of 29 contributors to The Reputation Society are women contributed to the neglect of this dark side of internet reputation.

There is a need for inventive and serious thought about the issues of reputation and trust in an increasingly digital world. Our social and commercial interactions will be increasingly mediated by large-scale software systems, and we need ideas about how best to design, navigate and regulate these systems. Unfortunately, by avoiding real-world cases and thorny problems, The Reputation Society provides no answers.