Research:Characterizing Reader-to-Editor Journey/Encourage contributions
This page summarises the ongoing research effort to design systems and mechanisms to encourage contributions to Wikipedia―especially favoring the transition from reader to active editor.
WikiGnome Toolbox edit
A WikiGnome is a wiki user who makes useful incremental edits without clamoring for attention. WikiGnomes work behind the scenes of a wiki, tying up little loose ends and making things run more smoothly.
Inspired by the Legitimate Peripheral Participation idea , which describes how people could become active community members by engaging in simple but necessary tasks, we explore possible designs to bring the editing experience into the reading experience, blurring the boundary between the two types of participation.
WikiGnome is a browser extension that aims to provide personalized and discreet recommendations on how to improve Wikipedia. Since the goal is to design a tool that can encourage contributions from readers who are not necessarily familiar with the community norms and edit mechanism, we need to identify a set of tasks that have the following characteristics: 1) they are productive and necessary for Wikipedia, 2) do not require deep knowledge of community norms or training, and 3) take around 1-2 minutes.
Some examples of tasks in this category may be typos fixing, evaluating readability or image relevance, voting for image insertion, adding a link, fixing formatting issues, verifying reference reliability, or fixing accessibility.
Image Accessibility edit
To allow visually impaired readers to access the content of images, the web standard is to use the alt-text HTML field. This field should contain a short and informative description of the image's content that is typically read by a screen reader.
Despite the crucial role that the alt-text has in ensuring universal access to Wikipedia knowledge, previous work suggests that only about 6% of images in English Wikipedia have an alt-text description, an issue that is even more severe in other languages .
With WikiGnome, we are investigating how to encourage readers to contribute with small edits to mitigate this issue. The current design is based on a nudging logic that shows a small banner directly where the article needs the attention of a contributor.
This section is a work in progress...
What about AI? Yes, caption generation models improved significantly and can reach a decent degree of accuracy, but they still perform poorly on some tasks. These limitations may be introduced both by the high visual diversity of Wikipedia that makes the task challenging and by intentional filters introduced to the model (GPT4 Vision Preview refuses to name people). For this use case, discussion with blind people and experts on accessibility topics highlighted that "it is better to have an empty alt-text than an imprecise one". This consideration suggests that human validation is still required, although AI can be a valuable tool to support the process.
- Lave, Jean; Wenger, Etienne (1991), Situated Learning: Legitimate Peripheral Participation, Cambridge University Press, ISBN 0-521-42374-0
- Kreiss, Elisa; Srinivasan, Krishna; Piccardi, Tiziano; Hermosillo, Jesus Adolfo; Bennett, Cynthia; Bernstein, Michael S.; Morris, Meredith Ringel; Potts, Christopher (2023). "Characterizing Image Accessibility on Wikipedia across Languages". WikiWorkshop 2023.