Is the Wikipedia Neutral?

Joseph M. Reagle Jr.

Claims of neutrality and accusations of bias are common themes of contemporary discourse about the media, government, education, and technology. In this essay I extend earlier work on the collaborative culture of Wikipedia (an on-line and free encyclopedia) to specifically focus on the fundamental but often misunderstood notion of neutrality. The Neutral Point of View (NPOV) policy recommends that "[a]rticles should be written without bias, representing all views fairly" (Wikipedia 2004npv). As I argued previously, the importance of this norm is that "while the perception is that NPOV is the source of much debate, it may act rather as a heat shield: reducing conflict and otherwise channeling outstanding arguments in the productive context of the primary goal of developing an encyclopedia that is representative of many viewpoints" (Reagle 2005).

Misunderstandings about this notion arise in part because, as the Wikipedia article itself admits, "the terms 'unbiased' and 'neutral point of view' are used in a precise way that is different from the common understanding" (2004npv). People are acknowledged to be subjective beings (i.e., "inherently biased"), but when used in the Wikipedia context articles are considered without bias when they "describe the debate fairly rather than advocating any side of the debate" (2004npv). This notion of neutrality is difficult because it seems impossible to explain without recourse to an equally problematic constellation of concepts. If neutral means unbiased, and unbiased means fair, might fair mean impartial, or something else? Another source of confusion is the subject of the alleged neutrality: the platform, processes and policies, people, practices, or the resulting articles? Can bias in one contaminate the neutrality of another? Additionally, the use of the prefixes "un" and "non-" with words such as "bias", "fair", and "neutral" is indicative of one more problem associated with the concepts. While we might find a clear definition of what bias is, for example, that definition might not be as equally as useful when we wish to understand what it means to be unbiased . Take, for example, the acronym "POV" which has acquired a derogatory connotation as the seeming opposite of "NPOV." Yet, when the acronym is expanded, to accuse someone of having a point of view seems rather ridiculous, even to those who advocate the neutrality policy.

This essay is inspired by earlier debates on neutrality of technical standards, literature on bias in technical systems, my present fascination with this Wikipedia norm and a change in my belief that while an important concept, the label of neutrality was an unfortunate coinage in the Wikipedia context. In this essay I hope to make a compelling argument that the label of neutrality is appropriate. I first consider a number of cases in which notions of neutrality are central (other reference works, playing games, technical systems and standards, content regulation, and war), define the terms "objective," "bias," and "neutral," and provide a framework for considering neutrality. I conclude by arguing that the term "neutral" is the right word for discussing the aspirations and results of Wikipedia contribution.

1 Existing cases

1.1 The Reference Work

The Wikipedia's concern with how to reconcile disparate views is not unique. McArthur (1986:55) notes that compilers of reference always have to "come to terms with the worldviews dominant in society at any time, or may need to seek a compromise among the worldviews that conflict." He found evidence of five different common approaches:

1. Dogma: the path of faith which insists on the truth of its ideology; "users of the work may then agree or disagree, and find the compilation useful or useless" (p. 55).
2. Rationalism: the path of reason and secular humanism which privileges assertions that achieve a consensus by means of evidence and logical arguments.
3. Neutrality and fairness: the path of "would-be even-handedness" that offers a "menu of various viewpoints," such as both faith and reason.
4. Compromise: the path of propaganda and expediency which attempts to reconcile - in the open or concealed - a dominant view with compatible elements of other views.
5. Pragmatic: the path of not "worry[ing] about cosmic underpinnings" but getting on with it, though this naivety might be deceptive to all involved.

While this is an interesting taxonomy, it is not easy to place the Wikipedia within it for two reasons. First, McArthur's analysis applied to finished products that were the results of a coherent and small editorial team with a homogeneous editorial policy, relative to the perspectives of the rest of the world. The Wikipedia is the product of thousands of contributors with differing worldviews interacting directly. Second, one can find elements of each of these worldviews in Wikipedia. For example, some might say that Wikipedia's neutrality policy is dogmatic; WikiInfo was founded for just this reason and permits users to associate differing viewpoints on any given article rather than developing a single "neutral" article. Additionally, Wikipedia's discursive culture is based on rationalism, concerned with consensus and encourages any claims to reference published materials rather than "original" (potentially crackpot) research. And, as I will show, notions of neutrality and fairness are very important. Finally, compromises do occur, and in keeping with the forever incomplete and changing nature of the Wikipedia, it is nothing if not a pragmatic undertaking.

In any case, it's not clear to me the extent to which McArthur felt these categories must be exclusive, or how they might be applied to new works. Consequently, I turn away from the study of reference works towards every day notions of neutrality to better understand what that term might mean in the Wikipedia context.

1.2 Playing fair

When I look up the word "fair" in my dictionary the first sense given includes the word "unbiased" - not terribly useful. However, it also notes an "accordance with the rules." This notion of playing by the rules is an important one. Many of us are familiar with the childhood protocol of determining who is "it" - perhaps the seeker in the game of hide-and-seek. And we also know that if clever and careful enough, the person reciting the rhyme (" one potato, two potato, three potato, four") can determine the outcome. Consequently, in the face of imperfect protocols and rules, fairness might require one to act in good faith, to assume the best, and grant the benefit of the doubt to others. It also sometimes requires a level of scrutiny and accountability, much like Reagan's aphorism on mutual nuclear disarmament with the Soviets, "trust but confirm." (Interestingly, last month I encountered a fair "it" protocol on a bocce court: the organizer says "go" and together each player throws her fist into the center with a random number of fingers extended; these are summed and then counted from the organizer around the circle until stopping on the "it.")

What, then, does the notion of playing fairly contribute towards understanding Wikipedia's notion of neutrality? Unfortunately this case, nor any other case - and perhaps not even their sum - provides a bright line by which one can simply proclaim neutrality. Instead, one encounters attributes that help one assess whether something is more or less neutral, and how one might be more so in the future. In this case, the notion of participants following a collection of rules under some sort of authority with some level accountability is key.

1.3 Bias in technological systems

A collection of scholars have noted that the ways in which we order (Bowker and Star 1999) and mechanize the world can be an exercise of power, or "political" (Winner 1986). Friedman and Nissenbaum (1996) define bias as systematic and unfair discrimination, not incidental (p. 23), and posit three possible sources. Pre-existing bias has its roots in social institutions, (e.g., "red-lining", or failing to offer home loans in certain neighborhoods); technical bias arises from decisions made in technical design (e.g., the size and way in which results are presented can favor those on the top of the first page); emergent bias arises in the context of use (e.g., educational software designed in a culture that rewards competitive and individualistic playing strategies may penalize students of a more cooperative culture).

The important notion here, for my purposes, is sensitivity to the possibility of systematic discrimination that is often hidden. For example, from the previous case of fair play, an "accordance with the rules" can only be as fair as the rules themselves; so we also need the larger system of rules to be open to scrutiny.

1.4 Policy neutrality in technical standards

A recent prominent debate about technological neutrality, in which I participated, arose in the context of the W3C's content labeling standard (PICS) and related policy statements. The W3C's policy neutral stance had its genesis, in part, in the earlier work of the X Consortium:

Bob Scheifler, a developer of the X Windows System, … is often quoted for his useful maxim of "mechanism not policy." The application of this statement was towards the rather technical decision of how applications should avoid setting X resources directly, but instead allow these "policy" decisions to be controlled by the user (Scheifler, 87). The result was a mechanism that allowed user control over graphical elements and [the] window system. (Cranor and Reagle 1997)

In this case, both the mechanism and policy spoken of were largely technical: the mechanism might specify the rendering of a window element (e.g., size and color) while the policy was a user controlled specific rendering (e.g., large and blue). In the context of the W3C projects, the policies were associated with the social issues of speech/censorship and privacy. However, substantively, the approach was the same as that of the X Consortium. The W3C policy statement declared:

Since the Web is global, its technology must support a wide range of policy options that encourage all cultures to use the Web. To support this mission in the context of a global information space, the W3C must be aware of and apply an understanding of public policy (which may inhibit or promote the growth of the Web) towards Web technology. W3C works on behalf of and with its membership to build and maintain an inter-operable network architecture. This architecture must allow local policies to co-exist without cultural fragmentation or domination. (W3C 1997)

In this sense, neutrality connotes a modular plurality that permits one to accommodate multiple policies or positions. (Paul Resnick, an author of the PICS standard, would often speak of "letting a thousand flowers bloom.") However, the claims of neutrality were contested, by critics such as Lessig (1997), because supporting multiple policies was itself a policy, and one that could facilitate the censorious actions of China. (PICS designers attempted to argue (Reagle and Weitzner 1998) that this was not their intention nor a likely outcome.)

Still, this case highlights the importance of impartiality, where possible, and a realization that this impartiality itself might have less than desirable consequences.

1.5 Content neutrality in speech regulation

One might ask if decisions are made in a content neutral manner. I borrow this concept from U.S. jurisprudence where it acts as a test for the level of judicial scrutiny applied to Congressional restrictions of expression. Content neutral laws (e.g., a ban on leafleting regardless of what the leaflets say) must pass the lower standard of being narrowly tailored, have an important objective (not involving the suppression of speech), and permit other alternative means of communication. Laws that do discriminate based on the content of the speech (e.g., reproductive health information) must pass the higher standard of serving a compelling state interest for which there is no other less restrictive implementation.

The contribution of this case is to require an explicit justification for discrimination and that the justification be satisfied with the least restricted mechanism possible.

1.6 Neutrality in times of war

Many of us are probably familiar with the neutrality of Switzerland in armed conflicts: "The term 'neutrality' is defined by the international community as non-participation in armed conflicts between other states" (DDPS 2005). Given the recent unification of Europe and the datedness of the Hague Conventions of 1907, Switzerland undertook a reconsideration of its neutrality policy in 1991. While the resulting white paper and supporting documentation is interesting, and surprisingly nuanced, the Federal Council concluded: "Swiss policy should continue to be characterized by the constancy and predictability that have earned Switzerland the respect of the international community in the past" (DDPS1993).

For the purposes of this paper, the salient point is that by not becoming involved in the partisan conflicts of others one can gain the respect of those others and realize benefits resulting from productive engagement with each.

2 A notion of neutrality

It would not surprise me if the very brief tour of these cases has, while raising related concepts, only further muddied the waters. Consequently, I shall now try to fix these notions into a semblance of coherency and specificity.

First, the prominence of the Wikipedia and the "blogosphere" relative to traditional reference works and journalism has occasioned much opportunity for discussion about the positions each of these media might take with respect to neutrality. In light of this and the previous cases I offer an understanding of the terms of objective , neutral , and transparent . I'm attempting to distinguish how each is used to describe the legitimacy associated with making a claim and possible problems associated with that legitimacy.

1. Objective: the claims have a correspondence to reality; they are typically embedded in a framework by which their validity is affirmed. For example, the scientific method posits mechanisms that mitigate errors common to human perception and therefore affirms its aspiration towards objectivity. Problems of this stance include the fact that the sort of claims one makes, and the questions one asks, are personally or socially influenced without such methodological bracing. Also, the appearance of objective methodologies can be easily mimicked.
2. Neutral: the claims are satisfactory, or at least mutually unsatisfactory, to the claims' constituencies. For example, when the press represents an issue by first finding the two extremes of the argument, they have at least not favored one of those extremes. Problems of this stance include that the constituencies may have not been accurately represented, both with respect to their positions and relative numbers.
3. Transparent: the claims have no pretense to objectivity, nor in accommodating various constituencies, but plainly represent the speaker's bias (e.g., the blogger who simply writes what she thinks). Problems of this stance include that it is often misperceived as one of the other two (objective or neutral), and that it includes no inclination towards finding common ground with others.

Second, from the cases I discern a frame for conceiving of neutrality, which includes:

1. a sensitivity to the ways in which technical and social systems might unfairly discriminate against, often minority, positions (from bias in technological systems)
2. a notion of impartiality and plurality between possible participants or positions. This permits the benefit of forward movement from interaction with the otherwise irreconcilable (from neutrality in technical standards and times of war ).
3. An extension of sportsmanlike good-faith , and an adherence to known rules, a submission to some authority for arbitration, as well as an expectation of accountability (from playing fair )
4. a commitment to use and improve upon mechanisms which are the least onerous (from content neutrality in speech )

2.1 Is "neutral" the right word?

The question in the title of this paper begs a question: what is meant by "neutrality"? In the previous section I offer definitions of objective, neutral, and transparent, and a provisional framework for neutrality. In this section I hope to show neutrality is the right term for Wikipedia to concern itself with.

First, the term is far superior to the notion of "unbiased" – the original coinage of the Wikipedia's predecessor, the Nupedia – since that connotes a sense of epistemic purity whereas neutrality is indicative of a careful and fair balancing.

Second, the Wikipedia community appears to be very committed to not being biased in the sense of Friedman and Nissenbaum (1996). Wikipedia content might exhibit pre-existing bias in that there's a disproportionate amount of articles on technical and science fiction topics - reflecting the predispositions of many of the original contributors. Yet, the "countering systemic bias" project (Wikipedia 2005csb) is intended to address this problem.

A case of technical bias might be the way in which the present Wikipedia structure is "radically" open (no special application is required for anyone to edit a Web page); this also permits antisocial behavior such as vandalism which can be a nuisance to repair or even alienate serious contributors. However, I would argue that openness is a value of the community rather than a bias: it is not hidden or unfair to a particular constituency. Yes, the consequence of one value (e.g., openness) might conflict with another (e.g., being honest about another). Yet, for this class of problem, it would be better to understand how differing values can be balanced, or even better, reconciled.

For example, some do argue that because the Wikipedia permits anyone to edit, it also allows frauds and defamations. In a recent discussion about a case of defamation of John Seigenthaler Sr. (Seelye 2005), Jimmy Wales argued that to equate openness with defamation is like equating a restaurant's steak knives with stabbings. To force everyone in the restaurant to be isolated in steel cages because of the possibility of a stabbing would violate the values of "human kindness, benevolence, and a positive sense of community" and, consequently, "I do not accept the spin that Wikipedia 'allows anyone to write anything' just because we do not metaphysically prevent it by putting authors in cages" (Wales 2005). This example speaks to the question of the moral responsibility of those who design a particular structure for the range of actions taken by users of that structure – much like Lessig's accusation that PICS could be used by China. Also, the values of community and openness might in some cases conflict with the values of safety and integrity. In this particular case, one would have to consider the relative likelihoods of a stabbing given the presence of steak knives in a restaurant, with false information appearing on the Wikipedia given a lack of access controls. And, the relative difficulties and consequences of accommodating the other values of safety and integrity such as steel cages or requiring Wikipedia editors to authenticate themselves.

Additionally, the communication and collaborative norms in which the Wikipedia originally developed may only become clear when it encounters new contexts. For example, now that the Wikipedia has many international contributors there have been accusations of America-centrism. Yet, the Wikipedia community supports over 30 world language editions with 10,000 articles or more, strives to be international, and heroically struggles with the problem of spelling and place names (Reagle 2005).

Finally, in the Wikipedia culture, the notion of "neutrality" is not understood so much as an end result, but as a process. Elsewhere, I employed the fancy coinage of "epistemic stance" (2005) to label the good faith, open-minded, "writing for the enemy" disposition contributors should take: "We should all be engaged in explaining each other's points of view as sympathetically as possible"; while, "The other side might very well find your attempts to characterize their views substandard, but it's the thought that counts" (Wikipedia 2004npv).

Clearly, one might find instances of less-than neutral systems, people, and articles on the Wikipedia. And reasonable people might disagree over a particular case. However, "neutral" is the right frame for discussing such cases and for understanding the larger aspirations of the community.

3 Bibliography

Bowker, G. C. and Star, S. L. (1999). Sorting things out: classification and its consequences. MIT Press, Cambridge, MA.

Cranor, L. and Reagle, J. (1997). Designing a social protocol: lessons learned from the Platform for Privacy Preferences Project. In Proceedings of the Telecommunications Policy Research Conference (TPRC97). ACM Press, NY.
[ http://www.w3.org/People/Reagle/papers/tprc97/tprc-f2m3.html ]

DDPS (1993). White paper on neutrality - annex to the report on swiss foreign policy for the nineties. The Federal Department Of Defense, Civil Protection and Sports (DDPS), Switzerland. Retrieved on September 15, 2005 from < http://www.eda.admin.ch/eda/e/home/recent/rep/neutral/neut93.html >.

Friedman, B. and Nissenbaum, H. (1996). Bias in computer systems. ACM Transactions in Information Systems, 14(2):330-346.

Lessig, L. (1997). Tyranny in the infrastructure. Wired, 5(07). Retrieved on September 15, 2005 from < http://www.wired.com/wired/archive/5.07/cyber_rights_pr.html >.

McArthur, T. (1986). Worlds of reference: lexicography, learning, and language from the clay tablet to the computer. Cambridge University Press, Cambridge.

Reagle, J. (2005). A case of mutual aid: Wikipedia, politeness, and perspective taking. In Proceedings of Wikimania 2005 . Retrieved on April 01, 2005 from < https://reagle.org/joseph/2004/agree/wikip-agree.html >.

Reagle, J. and Weitzner, D. (1998). Statement on the Intent and Use of PICS: Using PICS Well. Note, W3C.
[ http://www.w3.org/TR/xmldsig-p3p-profile/ ]

Seelye, K. Q. (2005). Snared in the Web of a Wikipedia liar. New York Times. Retrieved on December 08, 2005 from < http://www.nytimes.com/2005/12/04/weekinreview/04seelye.html >.

W3C (1997). Statement on public policy. W3C. Retrieved on September 15, 2005 from < http://www.w3.org./Policy/statement.html >.

Wales, J. (2005). Wikipedia and defamation: man apologizes after fake Wikipedia post. Air-l. Retrieved on December 14, 2005 from < http://listserv.aoir.org/pipermail/air-l-aoir.org/2005-December/008894.html >.

Wikipedia (2004). Wikipedia:neutral point of view. Wikimedia. Retrieved on March 05, 2004 from < http://en.wikipedia.org/w/index.php?title=Wikipedia:Neutral_point_of_view&oldid=6042007 >.

Wikipedia (2006). Wikipedia:wikiproject Countering systemic bias. Wikimedia. Retrieved on April 06, 2006 17:43 UTC from < http://en.wikipedia.org/w/index.php?title=Wikipedia:WikiProject_Countering_systemic_bias&oldid=47271840 >.

Winner, L. (1986). Do artifacts have politics? In The Whale and the Reactor, pages 18-39. The University of Chicago Press, Chicago.
[ http://www-personal.si.umich.edu/~rfrost/courses/Women+Tech/readings/Winner.html ]