Wikipedia:Wikipedia Signpost/2022-11-28/Op-Ed

Diminishing returns for article quality: Have we gotten past the point where better articles makes us a better encyclopedia? And what comes next?
Op-Ed

Diminishing returns for article quality

Julle began editing the Swedish Wikipedia in 2004. He's authored a book, Wikipedia inifrån (Wikipedia from the Inside), published in 2022, and currently works for the Wikimedia Foundation's Product Department. This text is unrelated to his work, and the opinions expressed are his own.

When Wikipedia took over the world, it wasn't on the basis of article quality. If Wikipedia one day is replaced, it likely won't be because someone does what we do better.

In his widely influential 1997 book The Innovator's Dilemma, American scholar of business administration Clayton Christensen investigated how dominant technologies are overtaken by new ones, and how the old organisations rarely managed to retain their positions as their industries shifted to a new paradigm. A key observation in Christensen's book is how the new technology, a product that is a break from tradition rather than a continuous improvement of the existing technology, is typically not better than the one it is replacing, but merely cheaper, simpler, more convenient.[1]

This echoes our experience: while Wikipedia's large language versions have long come out favourably in comparison to traditional printed encyclopedias, this largely happened after we achieved our position as the predominant source of information. In recent years, our reputation has changed for the better in countries like the United States or Sweden, as people belatedly realised that the Wikipedia of 2018 was not the same thing as the Wikipedia of 2004, but this was not necessarily correlated with increased readership in these areas.[2] It seems largely irrelevant to readers' decision to use us as a source of information – for that, we didn't need to be good. We just needed to be good enough, and then other factors – price, easy access – made all the difference.

A graph showing quality expectations/user needs being exceeded by product performance.
Adapted from Clayton Christensen's graph of product performance

To do what Wikipedia does

There have been attempts to do what Wikipedia does but better, like Citizendium or Everipedia. They seem doomed to fail. Not only because some of these endeavours insist that key aspects behind Wikipedia's success, such as the low threshold of entry, are defects to correct. Most importantly, it doesn't matter if they succeed in their ambition or not: one can't dislodge a supremely dominant entity like Wikipedia – entrenched in the fabric of the internet, with superb name recognition, hundreds of thousands of editors – by doing the same thing but slightly better. No product can win by modestly improving what the users are already doing; as human beings, we put a value on something simply because we are already using it.[3] There is no oxygen left to breathe for an English encyclopedia competing in the same niche. The strong competitors to Wikipedia exist in languages where the Wikimedia movement has been obstructed, like Baidu Baike in China.

Anyone wanting to dislodge Wikipedia from its place in the information ecosystem can't have article quality as their main selling point. This isn't just because the Wikipedian system of quality control, chaotic as it seems in theory, sort of works in practice, and our articles are often quite good – but because Wikipedia was good enough for the readers to start using it a long time ago. We have years of continuous editing and improvements beyond that point. It's not that Wikipedia is perfect, merely that we're probably way past the mark where additional quality will be attractive enough to change reader behaviour. Whether this is an indication of Wikipedia’s excellence or a reason for bleak despair as we look at how humans handle information may be in the eye of the beholder.

The rust in our machinery

There are, of course, concerns. Wikipedia was created with the assumption that anyone reading it would also be sitting in front of a keyboard. The conflation of the reader and the writer, the erasure of the strict line between the two roles, depended on the readers having the tools to efficiently contribute. Not only does this seem inherently more difficult on a phone – while certain things, such as patrolling, could arguably be equally easy or easier on a phone if our workflows weren't still primarily built for desktop users, most find adding text and references easier with access to a bigger screen and a physical keyboard – but we have over the years erected barriers ourselves. It is increasingly difficult to write new articles. And so we're at risk, when the constant editor attrition in some wikis outpaces the recruitment of new writers.

Many who attempt are thwarted not by the confusing code or technology, but by the sheer amount of norms and guidelines we have produced over the years. Wikipedia's guidelines have, like gneiss, been formed under external pressure, a reaction to attempts to fool or influence the encyclopedia, or in our own recognition of our shortcomings. As we have grown our concerns have shifted, in what seems to be a general pattern on Wikipedias of a certain size and age: from focusing on making sure we have the information to better control of the information. Other changes seem to be our internal definition of who we are. We gradually move towards stricter interpretation of our policies, like English Wikipedia's recent decision to require more sources to prove notability of Olympic athletes or how we prune the lush garden that is in-universe content related to popular culture, defining what is fancruft to be weeded out. This is where English alternatives to Wikipedia can thrive, rather than in the space we so firmly occupy: the corners we explicitly don't want, like the Fandom wikis, serving another purpose.

Quality and the reader

In their 2014 study, Lehmann et al.[4] mapped reader behaviour to see how it corresponded to our definition of article quality. Readers behave in different ways: they can read an article with focus, they might explore a topic jumping from article to article, or give it a cursory glance. Whether someone would sit down and spend time reading the entire article or just wanted to quickly peek at it had very little to do with our concept of article quality.[4] A common interpretation of this seems to be that Wikipedia, where we often invest time in what interests us rather than based on future pageviews, has a problem in misaligning article quality with the topics our readers are interested in – a pattern we see not just in what our audience chooses to read carefully, but also in relationship to page views.[5]

A different explanation would be that our concept of quality doesn't necessarily coincide with readers' needs. The goal of the encyclopedic article is to arm the reader with the right amount of knowledge. As we try to find the right amount of information to serve, we celebrate ambition and length. This is not necessarily wrong: there seems to be a correlation, albeit weak, between quality, as defined by the Wikipedia community, and reader trust in the article.[6] It is important that we actively fight disinformation in our articles; requiring sources is our best tool for doing so. However, a featured Wikipedia article has most likely long surpassed what would satisfy the reader.

Technology and shifts

Wikipedia grew in a symbiotic relationship with the concept of the search engine, not the least of which is Google. The encyclopedia significantly enhanced the quality of information gained when searching, and the search engines escorted readers to Wikipedia. Later, some of that balance has shifted, as Google now retains readers by serving the information they are looking for already in the search result.[7] Even the limited information in the Google Knowledge Graph, put together from various sources including Wikipedia and Wikidata, is often good enough. But increasingly, new internet users seem to abandon search engines as a way of looking for information.[8][9] They are happy using TikTok: to them, using the platform where they already spend their time is a simpler and more convenient way of looking for information.

We're not a company. We don't exist to bring value to shareholders and our purpose is not to be a tool for enrichment. When we work on our articles, we don't do it to better position ourselves, but to better fulfil our mission. To some degree, we don't have competitors: if someone is providing the world with information, they are merely doing what we want to be done. There is an argument to be made that we should do our thing, and if someone else comes along and does something else, something better, fine – we have served our purpose. But there are values which might make Wikipedia worth defending, even if information would be available elsewhere. Our belief in neutrality, in transparency and being able to show the reader from where we have collected the information. These are principles which deserve to survive technological shifts.

The day Wikipedia is replaced, it will likely be by something completely different that didn't even set out to compete with the Wikimedia wikis. There will be a niche we don't cover where a new initiative can thrive and find their audience, and grow until they – like we did – take up so much room there isn't enough oxygen left for us to breathe.

Article quality is important, as a method to achieve our mission. But article quality will not in itself save us if technology and user patterns leave us behind.

References

  1. ^ Christensen, Clayton (2016). The Innovator's Dilemma : When new technologies cause great firms to fail. Boston: Harvard Business Review Press. ISBN 978-1-63369-178-0.
  2. ^ Total pageviews, Swedish Wikipedia January 2016 to October 2022. Wikistats. Retrieved 19 October 2022.
  3. ^ Gourville, John T. (2006). "Eager Sellers and Stony Buyers: Understanding the Psychology of New-Product Adoption". Harvard Business Review. 84 (6): 98–106. Retrieved 2022-10-08.
  4. ^ a b Lehmann, Janette; Müller-Birn, Claudia; Laniado, David; Lalmas, Mounia; Kaltenbrunner, Andreas (2014). "Reader preferences and behavior on Wikipedia". Proceedings of the 25th ACM Conference on Hypertext and Social Media. Association for Computing Machinery. pp. 88–97.
  5. ^ Morten Warncke-Wang; Vivek Ranjan; Loren Terveen & Brent Hecht (2015). "Misalignment Between Supply and Demand of Quality Content in Peer Production Communities". Proceedings of the The 9th International AAAI Conference on Web and Social Media (ICWSM).
  6. ^ Elmimouni, Houda; Forte, Andrea; Morgan, Jonthan (September 2022). "Why People Trust Wikipedia Articles: Credibility Assessment Strategies Used by Readers". OpenSym '22: Proceedings of the 18th International Symposium on Open Collaboration. Association for Computing Machinery. pp. 1–10.
  7. ^ McMahon, Connor; Johnson, Isaac; Hecht, Brent (3 May 2017). "The Substantial Interdependence of Wikipedia and Google: A Case Study on the Relationship Between Peer Production Communities and Information Technologies". Proceedings of the Eleventh International AAAI Conference on Web and Social Media. pp. 142–151.
  8. ^ Moon, Julia (2022-07-28). "Why I Use Snap and TikTok Instead of Google". Slate. Retrieved 2022-10-08.
  9. ^ Rebollo, Clara (2022-09-14). "Rápido, adictivo y entra por los ojos: TikTok ya es el buscador de la generación Z". El País (in Spanish). Retrieved 2022-10-08.