User:Wugapodes/Encouraging small edits improves recruitment and retention

When I talk to people about contributing to Wikipedia, many of them believe they need to be willing to write entire articles. A common response is that people feel they don't have time to contribute to Wikipedia. Subject matter experts especially often spend their work days writing about a topic, so the idea of sitting down and doing more of that for free is not exciting. When trying to recruit and retain new contributors, the main blocker I see is that people perceive a big gap between being a reader and being a contributor. This leads to fewer people trying to edit, and for those that do try, the hurdles they encounter only reinforce their perception leading to discouragement and high drop-out rates.

This surprises me because my sense as an editor is that most of the work we need done is not writing! Many potential contributors are surprised to learn that my day-to-day editing comprises small maintenance tasks or flagging issues for editors with more expertise than I have. I find myself explaining that a large portion of my workflow is just making changes as I read the encyclopedia. In my experience, the gap between reader and contributor is actually quite small, but this reality is not the focus of Wikipedia's recruitment and retention efforts. This discrepancy is not benign because it means we are still likely to recruit and retain those who already fit the profile of the average Wikipedian, especially those "employed as a white-collar worker or enrolled as a student rather than being employed as a blue-collar worker" (which is a factor that has its own racial and gender bias, and so will reproduce those biases in addition to the economic bias).

Minor edits; big impact
My theory of change is that by training new editors on small edits they can make while reading, we can increase recruitment and retention. This is already being partially implemented by Specifically the theory is:
 * if we train readers to make small edits while reading then we will recruit more editors (by reducing the perceived gap between reader and contributor)
 * if newly recruited editors start by making small changes then they are less likely to be bitten (by changing the nature of conflicts new editors will have)
 * Editors who start by writing articles are likely to face hurdles that are quite hostile such as CSD and AFD or reverts of work that maybe took hours on the basis of obscure policies. A common place to see this kind of conflict is seen between student editors and long-term contributors at WP:ENB. By contrast, small edits made while reading are not going to be subject to deletion processes, and because they don't require large investments of time or energy, reverts are less of an inconvenience.
 * if newbies face less unhealthy conflict, then they are less likely to become discouraged and leave
 * if newbies stick around, then they are more likely to become long-term contributors as they slowly learn community norms

This theory of change has been partially implemented by the WMF's Growth Team as Newcomer tasks and the results of these interventions support the theory. Newcomer tasks can be unstructured ("suggested edits") or structured ("add a link"), and the features were released in that order. Results from the unstructured workflow support the theory that suggesting smaller edits leads to increased recruitment. Of editors who make an account, 21.6% make an edit (16.1% make an edit which is not reverted), and newcomer tasks increased this to 24.1%, 2.5pts above baseline (20.4% make an edit which is not reverted, 4.3pts above baseline). Results from both structured and unstructured workflows support the theory that newly created editors are less likely to be bitten. In the 2020 experiment baseline revert rate of newcomer edits was 28%, and unstructured newcomer tasks lowered the revert rate to 13% of edits. In the 2021 experiment the revert rate of unstructured newcomer edits was 25% and structured tasks reduced the revert rate to 7.9% of edits. Results from the unstructured analysis provide limited support for the theory that newcomers are more likely to stay if they are not bitten, though more study is necessary. Unstructured tasks increased retention from 3.2% to 3.6% (n.s.), and 11% increase which matches the 11% increase in retention. This leads to the hypothesis that retention is increased as a result of the increase in recruitment. The newcomer tasks have so far been very effective in recruiting those who make accounts.

Not every edit needs to be adding chunks of prose. Wikipedia grows best when lots of people make minor improvements over long periods of time, and editors are more likely to stick around if they can make edits as part of their usual reading routine.

Redlinks help the encyclopedia grow
Many readers think that they shouldn't add red links, but the truth is that red links help the encyclopedia grow. Some editors spend their time creating articles that have the most red links. Others look through red links and replace them with redirects to existing articles or sections on the topic. If a reader sees a phrase that they think would make a good title for an encyclopedia article, we should encourage them to turn it into a link and hit save no matter what color the link turns out to be.

For some major topics, Wikipedia has very little information and our depth of coverage is shallow due to our systemic biases. This leads to a secondary problem: when articles get created they may be orphaned with few or no incoming links. This makes articles hard to find, and going through existing articles to see where the new article can be linked is time consuming and difficult. If someone links to an article that does not exist yet, the link stays there when that article does get created, and articles can be integrated into the encyclopedia before they are even written.

Redlinks also serve as proposals for new articles. If editors agree, they can make a stub that others can build on further. If editors disagree, they can remove the red link. Most-wanted articles lists non-existent pages by the number of incoming red links, and some editors look through that page for ideas on articles to write. By adding red links it signals to these editors that the article is desired and makes it more likely to be created, and once it is created it is already integrated into the encyclopedia through incoming links minimizing the work needed to de-orphan new articles.

Ask questions
Wikipedia has a number of templates that let readers ask questions about the article they're reading, right in the article. Readers are very familiar with citation needed (for example, xkcd:285), but the wider series of inline cleanup templates are less well known and generally less used. By leveraging familiarity with the citation needed template, potential editors can be introduced to templates like How many?, according to whom?, and Like what?. These inline question templates can be framed as ways for readers to give feedback on article content without needing to research and write the content on their own. and these categoriesIt invites them to add the answer if they know it. Some editors spend their time looking through these questions and trying to answer them. By asking questions, you help improve the article and inform other readers about its current quality.

Answer questions
These inline cleanup templates place articles into maintenance categories, and these categories can be used as entry points for recruitment events such as an "answer-a-thon" (c.f. edit-a-thon). Potential editors can be directed to categories like Category:All Wikipedia articles needing clarification with the goal of clarifying some statements in those articles and removing the tag. The task is well defined and modular, so participants do not need to dedicate large blocks of time to research and writing which can be a barrier to those with low investment or limited time.

Alt text descriptions
Recent research by Elisa Kreiss and colleagues (2023; forthcoming) has used Wikipedia's uniquely large corpus of combined images, article-contexts, captions, and alt text descriptions to develop a computational method for evaluating and generating alt text based on the image and the context in which the image is used. The main innovation is that the model takes into account the context-sensitivity of alt text (see our MOS section). Alt text is crucial for ensuring the encyclopedia is accessible to readers using accessibility technology, but not everyone uses it and even then it can be difficult to write from scratch. An implementation of Kreiss' model could be integrated into the existing newcomer tasks workflow. This would lower the barrier to writing good alt text for images that don't have it, similar to the suggested copyedits model and add a link model, and it would provide a way to crowdsource review of existing alt text.

Readers, not just account holders
While the newcomer tasks provide support for the theory of change I lay out above, it's worth noting we target slightly different populations. The newcomer tasks target community members who have already made an account, but my theory targets recruitment of readers who likely do not have an account and are not interested in editing. A key insight of the Growth Team's intervention is "newcomers arrive with something specific in mind they're trying to accomplish, like add a specific photo to a certain article. We don't want to get in the way of them accomplishing their goal." Similarly, readers are generally reading the encyclopedia with a particular purpose in mind, and while we don't want to get in the way of their information goal, there is still an opportunity for intervention. The theory of change I lay out lends itself to both technological and social interventions. Some people do browse the encyclopedia without a clear goal using methods like Special:RandomPage. These random pages also tend to be poorly integrated with few incoming or outgoing links. These browsers may be an ideal population to target with interventions like an "add a link" fly out or interstitial. On a social level, a shift towards training on small edits can be achieved by changing the framing of recruitment programs. Instead of edit-a-thons, programs like tag-a-thons (for adding or addressing inline templates or problem templates) or link-a-thons (for clearing out categories like Category:Orphaned articles) which emphasize small, quick tasks that participants can simply drop in and out of lowering the perceived barrier to entry of the program. As more of these programs occur, new documentation can be written to leverage the impact for those who don't attend recruitment programming. Ultimately, the goal is to close the gap between readers and editors, and targeting readers who have made accounts targets only part of that population.