Machine Augmentation
Exploring machine-in-the-loop technologies to enable knowledge equity at scale
Overview
In 2018, the Wikimedia Foundation Product team identified machine augmentation of human activities on the wikis as a key dependency for realizing the Movement's vision of making the sum of the world's knowledge accessible to all the world's people. This section covers work related to evaluating, debiasing, and developing the algorithms that will make knowledge equity achievable at scale. Algorithms have the potential to make content better and more available across language and cultural boundaries, to make contribution and moderation tasks more efficient, to flag article (un)trustworthiness to the reader, and to provide an overall picture of coverage of knowledge across and within wikis.
Studies
-
MinT (Machine in Translation) Research
The goal of this project was to understand how to better leverage MinT to support more readers and contributors in their aims of accessing, interacting with, and contributing to Wikipedia content, as well as knowledge more generally. -
Translatable Pages
Developing a better understanding of the challenges users face when creating and updating translatable pages in order to improve the user experience. -
Section Translation Feedback Survey
Learning from the experiences of editors who used Section Translation during a Bengali Wikipedia article quality improvement competition in 2022. -
Section Translation Post Improvements Testing
Section Translation brings translation support to mobile device editors, and this project provided usability testing after a number of initial tool improvements and at a time when it was becoming available in a greater number of wikis, including Thai Wikipedia. -
Section Translation Entry Points
Improving discoverability of translation tools on Wikipedia. -
Section Translation Usability Testing
Many Wikipedia contributors edit articles via mobile devices. Section Translation brings translation support to mobile device editors, and this project provided usability testing as the tool became available in the first wiki. -
Multilingual Editor Experiences in Small Wikis
Supporting multilingual editors in small Wikipedias who are leveraging translation to contribute across knowledge and content gaps. -
Content Translation Newcomer Survey
A customizable survey that can be used to easily, quickly, and reliably collect feedback from a more diverse pool of Content Translation users. -
Article Section Translation Study
An evaluation of new designs for tools that help editors translate article sections and receive translation support on mobile devices. -
Machine Translation Meets Human Perception
The Machine Translation Meets Human Perception (MTMHP) study developed a protocol for evaluating how readers perceive and evaluate machine-translated content - across three different languages, cultural contexts, and content domains. -
Ethical & Human-Centered AI
This project identfies challenges and emerging opportunities to leverage AI technologies to further the mission of Wikimedia. -
Augmentation
This is a position piece developed by Wikimedia Foundation to frame the potential of machine-in-the-loop technologies in making all the world's knowledge available to all, and to ensure the process to assemble that knowledge is inclusive, balanced and safe.