Task recommendations for Wikipedia Editors
This page is obsolete. It is being retained for archival purposes. It may document extensions or features that are obsolete and/or no longer supported. Do not rely on the information here being up-to-date. |
This document describes ways we can automatically suggest tasks to Wikipedia editors.
The problem
[edit]Currently, only 15-16% of newly-registered users who make an edit to Wikipedia content go on to reach active editor status (5+ edits) in their first 24 hours.[1] This conversion rate is quite low, especially considering that users have already made it past the difficult task of completing their first edit.
We think that a large part of this is because:
- There is no clear path forward for the user after their first edit, even if they complete onboarding. If they are successful, they have to seek out additional tasks on their own.
- Users are not provided much positive reinforcement for their first edits.
In accordance with our 2014-15 goals, we want to increase the activation rate of new editors on Wikipedia.
Proposed solution
[edit]We think that active recommendations that are personalized to the user's topical interests will dramatically increase engagement among new editors. Additionally, personalized recommendations may be a tool for continuing engagement.
There are basically no active task recommendations made to new editors on Wikipedia. The most helpful form of recommendations comes from more passive or opt-in methods, such as Community Portals, categories, and maintenance templates embedded in articles. Only one (SuggestBot) is really personalized to the user.
Hypotheses and research questions
[edit]Our full research documentation can be viewed at Research:Task recommendations.
User experience
[edit]We will start by providing task suggestions to the user immediately during their first edit session. We plan to do this via two methods: a post-edit recommendation in a modal, and via a flyout similar to MediaWiki's web notifications.
Later, if these suggestions prove to improve editor engagement and productivity, we will expand to include email/web notifications, and potentially a full landing page. Recommendations will first be targeted at first-time editors, but eventually may be delivered to anonymous editors and more experienced registered editors.
Post-edit recommendation
[edit]The first form of recommendation we will expand on the editing confirmation message. This recommendation is based on the current page just edited, and presents one recommendation for every saved edit.
Elements include:
- The edit saved confirmation message of âYour edit was saved.â
- A subtitle of âEdit a similar articleâ
- The lead image (derived from Extension:PageImages), or an image placeholder (as seen above)
- The recommended page title
- A human-readable timestamp of the last edit
- A close icon
Behavior: after every edit to an article, users will see the confirmation plus recommendation. On hover, the entire recommendation area will be underlined and link to the recommended page. Unlike the regular confirmation message, this recommendation will not fade away automatically, in order to give the user time to see the unique recommended page and decide if they are interested. If the user edits a non-article (i.e. outside namespace zero), creates a page, restores a page, or if the recommendations API fails to return a result, then the regular post-edit confirmation message will be shown.
Recommendations flyout
[edit]In addition to the post-edit notification, we'll provide the user with further recommendations in a dedicated flyout. The purpose of the flyout is to provide an easily accessible area to find interesting things to do, where ever the user might be on Wikipedia. Like the post-edit version, the recommendations will be based on the user's last edit, but will provide a greater number of recommendations to choose from.
Elements' include:
- A link, titled âRecommendationsâ and with an associated lightbulb icon, which opens the flyout. This link lives in the personal toolbar, is displayed only for registered users, and is positioned to the left of the username.
- A flyout title and subtitle, which explain the purpose and nature of the flyout
- Three initial recommendation cards, each containing the same information as the post-edit notification described above.
- Pagination, which will provide three pages with three sets of recommendations (nine total recommended articles)
Behavior: when clicked, the âRecommendationsâ link opens the flyout. The flyout should open immediately, and if it needs to be refreshed or wait for recommendations API calls to finish, then a loading spinner should appear. The flyout has no close icon, but can be closed again by clicking either anywhere outside, hitting ESC, or clicking the âRecommendationsâ link again. If the user has no edits to articles which can generate recommendations, then an empty state inside the flyout should be shown which encourages the user to edit to receive recommendations. If the recommendations API returns an error for any reason, then an error state should be shown inside the flyout.
-
Error state for when users have no content edits to base recommendations on
-
Error state for when an unexpected error occurs with the API
Implementation
[edit]For the purposes of testing, we will serve recommendations by extending the GettingStarted extension's  gettingstartedgetpages
 API to support a morelike
task type. Selecting this task type will allow you to return n pages related to a given source page, using CirrusSearch. For now, the API derives its recommendations via the page last edited via the user. (In the future, we may expand recommendations to be based on a user's entire edit history.)
The API filters pages that are not content pages (i.e. namespace zero only), and prevents you from recommending sensitive pages in Wikipedia's Category:Living people to users who are not autoconfirmed. It also requires you to set an excluded page, to avoid recommending the page the user is currently on.
Qualitative testing
[edit]See Usability testing for notes on qualitative feedback in the form of user tests and a survey.
Footnotes
[edit]See also
[edit]- More mockups and screenshots for Wikipedia task suggestions on Commons
- What should I edit, which explored how to present data on new articles gathered by bots
- Micro contributions on mobile
- https://github.com/Wikia/wiki-recommender