Essays & conversations about constitutional moments on the Net collected by the Berkman Center.

Is the lack of web link and search engine accountability the elephant in the room of online reputation?

Essay by Chris Dellarocas, October 17, 2008 in response to Is reputation obsolete?

The majority of debate on online reputation and free speech has focused on questions that relate to content authorship and hosting (see for example, this book and related discussion here, here and here). There has been far less discussion about the responsibilities of those who link to harmful content as well as about the accountability of search engines, whose page ranking algorithms – themselves based on counting links – largely determine the extent of such content’s impact. In this essay I argue that page ranking algorithms and people’s linking decisions are at least as important components of online reputation formation as content itself and deserve to be made more visible and, perhaps, more accountable.

Links constitute the true currency of reputation on the web. Even the most malicious online content will remain largely unnoticed unless others choose to link to it. Links are all the more important since search engines, that ultimate arbiter of online relevance, use a page’s link counts as the primary determinant of that page’s ranking within a set of search results.

Linking to a piece of content constitutes a judgment on the part of the linker that this information is worth noticing. By the same token, a search engine’s choice to employ a link-based page ranking algorithm constitutes a judgment that link counts are a fair method of determining web content’s merit of being read.

Although linking is as deliberate and consequential an action as authoring, our social norms and legal structures have paid much less attention to it. Many web users who would never dream of posting certain types of content have far fewer qualms about linking to them. In cases of slander only the original creator of content bears legal responsibility. Section 230 of the Communications Decency Act of 1996 provides almost blanket immunity to the people who helped make this content visible by linking to it. Similarly, no responsibility is borne by the search engines, whose algorithms chose to list the content near the top of search results and greatly contributed to its negative impact.

Viewed from this perspective, the current lack of accountability with respect to linking and page ranking constitutes an important shortcoming of our fledgling reputation economy. On the one hand it encourages irresponsible and sometimes malicious behavior. On the other hand it misses a great opportunity to turn the millions of web users into more intelligent and responsible information gatekeepers.

Let me be upfront in that I am not advocating more litigation. I believe that a lot can be accomplished through education and implementation of the right incentives into the technical architecture of the web.

The first step is education. Most people do not fully realize the implications and responsibilities that come from their choice to link to a piece of online content. Even fewer people fully grasp the way in which web links, stripped out of their original context and aggregated en masse, affect the decisions of page ranking algorithms. For example, a blogger who links to a racist article from inside a posting in which she strongly condemns it, is at the same time boosting that article’s PageRank, improving its visibility on search engines and exacerbating its negative impact. Fully grasping the consequences of an individual’s linking decisions is the first step towards using this powerful staple of our networked society with responsibility.

The second step is implementing incentives for responsible linking and page-ranking into the architecture of the web. For example, one can envision a set of mechanisms that keep track of the linking actions of websites (and, to the extent possible, individuals), and, on the basis of such actions, assign to them a publicly visible score that roughly translates to their “quality of judgment”. Linking to content that proves to be beneficial increases the score; linking to content that proves to be harmful decreases it.

In a reputation economy, a person’s quality of judgment is as valuable and important a trait as a person’s reputation on any other dimension. In small communities people who spread false rumors quickly acquire a reputation for bad judgment and become ostracized or irrelevant. On the other hand, people who exhibit good judgment grow in esteem and are welcome everywhere. We need to build a similar set of checks and balances for the web.

Search engines must be subject to similar scrutiny. Their choices of page ranking algorithms are deliberate and, therefore, accountable. Plus they have very real consequences. It is my hope that public measurement of a search engine’s “quality of judgment” will induce the creation of more responsible algorithms. At the minimum, it will alert users that these all-powerful gatekeepers of reality are not infallible.

Implementing these ideas will not be easy. There are several difficult challenges for which there are no easy answers. Here are just a few: Who gets to decide what content is beneficial and what is harmful? In limited cases (for example, content that has been proven to be libelous in court) making such judgments with a fair degree of objectivity is feasible; in the majority of cases, however, such decisions will be subjective. How should one take into consideration the context of a link? For example, when a blogger lambasting a libelous posting ends up boosting its visibility on search engines, is this an instance of poor judgment on behalf of the blogger or a failure of page ranking algorithms to properly take the context of the link into consideration? Who should bear the responsibility (or get the credit) for anonymous links posted as comments on eponymous blogs?

Despite their difficulty, these are challenges that we cannot ignore. In our networked society, linking and page ranking carry just as much weight as authoring. All three need to be exercised with caution and responsibility. Similarly, any discussion of free speech and online reputation must focus on all three.

Chris Dellarocas is an Associate Professor of Information Systems and Director of the Center for Complexity in Business at the Robert H. Smith School of Business of the University of Maryland. His research examines the implications of consumer-generated content and social web technologies on business and society. His work on online reputation formation has received international recognition and has been quoted in, among other places, CNN Headline News, The New York Times, The Wall Street Journal, Business Week, Washington Post and the Financial Times. He is an inventor with 3 patents and board member of several Web 2.0 companies. 

Comments (4)

    • Crosbie Fitch wrote:

      Nick, how refreshing to hear from someone who recognises that nature gives us the biggest clues in finding solutions to apparently intractable problems.

      Reputation and identity are indeed social constructs, emergent epiphenomena. In the offline world we do not refer to a central authority to ascertain someone’s reputation or identity. We corroborate these things socially, with our peers and acquaintances.

      Reputation and identity are problems that should be solved using decentralised, public, peer-to-peer technologies, not centralised, proprietary systems.

      I wrote on this a while ago here:
      http://www.digitalproductions.co.uk/index.php?id=69
      (Ideating Identity)

    • Nick DiGiacomo wrote:

      What’s missing in most discussions of online reputation is a workable model for the process by which reputation is created.

      According to Wikipedia, “reputation is the opinion (more technically, a social evaluation) of the public toward a person, a group of people, or an organization”. I would suggest that any authority relative to measuring online reputation should speak directly to this technical definition – i.e. to reputation being a social evaluation.

      To start, this definition tells us what reputation is not. Reputation is not just “the facts”. It’s not, for example, just some combination of how many tons of CO2 a company produces, how much it gives to charity, and how long the average wait is on a service call. Nor is it just an assessment or analysis by a person or organization – no matter how expert he/she/it might be – of a company’s behavior or performance (good or bad). Nor is it just the total of all the media stories and articles that journalists, bloggers and researchers publish about a company’s or an individual’s deeds and misdeeds.

      Even if we summed up all the facts, analyses, assessments, stories and articles about a company or an individual, we still wouldn’t have measured reputation. All that information has to be passed through two important filters. First are the beliefs and biases of the individuals (e.g. consumers) who are interpreting the stories, analyses and facts. And second is the complex and non-linear way those opinions move through the larger population.

      What comes out the other end of those filters – after that process of social evaluation – is what is recognized as reputation.

      The simplest example of the above process at work is gossip. This analogy does not trivialize reputation. Quite the opposite - it ties it to a mechanism on which humans have relied over the eons to establish trust.

    • Chris Dellarocas wrote:

      You are raising a very interesting point when you ask “what to do with criticism?” The answer is not clear to me either, however, here’s what’s different on the Web: because of the way search engines interact with links, it is possible that well-intentioned (and deserving) criticism might boost the visibility and impact of a harmful posting in different contexts than the one intended by the author of the critique. Example: You are a very popular blogger and you come across a libelous posting about Barack Obama that you really feel deserves criticism. You write an article in your blog about this posting; of course, you include a link. Since your blog is very popular, your link boosts the libelous posting’s Pagerank. This in turn causes Google to list it higher up when other people type “Obama” and helps lots of other people find it. Since these other people do not access the posting through your commentary but directly through Google, they do not benefit from your insights so many of them believe it. In this scenario, your well-intentioned attempt to criticize ends up increasing the damage that this posting causes. Would the right thing have been to ignore the libelous posting? I am not sure. However, this particular side-effect of linking is something that most people who participate in online discussions don’t even think about. My viewpoint is that this and other related phenomena deserve more attention and study.

    • Bertil Hatt wrote:

      I’m not sure that Google isn’t able to tell a negative link for looking at associated words around it; some people like me love to use a meliorative semantic field to criticize, or apparent stern words to praise — but for most links, “terrible” won’t be good. What to do with criticism, though? Controversy can be the best remedy to Balkanisation, and I sense much more of a threat there, thanks to the web at least in Politics (less so in Academia, at least for younger, Google Scholar-driven grad-students).

      However: how is that specific to the Web, ie. a word made of formal, one-way links? Doesn’t any commentator, be it in the press or at a diner party, by lambasting anything take the risk to launch a mob or trigger contradictory reactions?

      Some technical solutions exist: meta-tags, no-follow; but aren’t adopted by many. One could imagine to have more structured data (and I’ve seen very encouraging demo) but short of being simpler then HTML, they might never get the same adoption.

    Post new comment

    The content of this field is kept private and will not be shown publicly.
    CAPTCHA
    This question is for testing whether you are a human visitor and to prevent automated spam submissions.
    Image CAPTCHA
    Enter the characters shown in the image.