By Miriam Redi
Wikimedia Foundation, the non-profit that runs Wikipedia in collaboration with AfroCROWD and Behance, launched a campaign called Wiki Unseen. The campaign is giving visual representation to biographies of Black, Indigenous, and People of Colour on Wikimedia projects. The Wikimedia Foundation’s recent analysis shows that nearly half of English Wikipedia articles are missing images. These figures vary across language editions. For most Wikipedias, between 40% and 60% of the articles are illustrated. Across languages, shorter articles, where non-textual knowledge can be very helpful to complement the lack of content, are unfortunately the most affected by the lack of images.
But why is it so important to have images on Wikipedia? Research tells us that images can play a very important role in encyclopedic knowledge.
Visual representation can help reduce gender gaps, and break down stereotypes: Exposure to biased media knowledge can reinforce existing demographic gaps and prejudices. Vice versa, interventions aimed at exposing subjects to counter-stereotypical role models can have positive influence in reducing biased judgments. For example, studies showed that exposure to biographies of women in leadership positions reduced prejudices about gender. Wikipedia can play a key role here.
Biographies make up about 30% of Wikipedia articles, and this knowledge is accessible to billions of people. Wikimedia communities are aware of existing knowledge gaps in Wikimedia projects, and are organized to rebalance the presence of textual content for different gender or ethnical groups. Sadly, more needs to be done as more than 80% of biographies on Wikipedia are about men. Moreover, Wikipedia is missing many images, and its biographies, especially women biographies, too. Our recent study shows that in English Wikipedia, only about 40% of biographies are illustrated. Wiki Unseen and other Wikimedia Foundation projects such as WikProject Women in Red in English Wikipedia, or Wikiproject Daughters of Hind in Hindi Wikipedia, are making an effort to get more women’s biographies across languages that are well curated.
Images are engaging for our readers: According to the Wikimedia Foundation recent study on readers’ interaction with images, images are engaging for our readers. We found that, on English Wikipedia, users click on images 1 out of 30 times that they read an article. This might not seem like a huge figure, until it is put into perspective. On English Wikipedia, citations are clicked on only 1 out of 350 times, whilst external links are clicked only for 1 in 110 pageviews. So Wikipedia users tend to interact with images much more frequently than with other interactive parts of the page!
The study showed that images with a higher clickthrough rate are of three types. First, readers click more frequently on images of complex objects, such as for example visual arts, or vehicles. Readers also use images on Wikipedia to visit the world, engaging with photos of landmarks and maps of different countries. Finally, biographies: unlike users of other web platforms, such as social media or image search engines, for Wikipedia readers, images of popular people seem to be much less engaging than other subjects. However, readers do interact with portraits from less popular articles, i.e. when the subject of the biography is not very well known.
Images can help satisfy our informational needs: Cognitive scientists observed that, in most cases, pictures are more memorable than words, and that our visual cortex processes and recognizes images in fractions of a second. But what do we use images for? In our research, we found that images on Wikipedia might have a cognitive function: i.e. they can facilitate the comprehension of textual content by providing additional or clearer information.
We observed that readers engage more often with images in shorter articles, probably to complement the lack of knowledge in articles where traditional textual content is missing. This is particularly important for Wiki Unseen because most of the biographies also do not have a lot of information about the people who have been profiled. Through getting visual representation for these images, the project went a long way in addressing this knowledge gap. This research, learning through Wikipedia, is a first step towards understanding how readers use images on Wikipedia.
Images can be reused across different pages and projects: One of the reasons why Wiki Unseen biographies had lacked visual representation is due to the fact that there were no free licenced images on the internet of these individuals that could be used on Wikipedia. An image added to an article on Wikipedia is not an isolated contribution. It is a gift to the largest source of visual encyclopedic knowledge on the web. Images in the Wikimedia network are freely shared in the broader web and reused by everyone who needs it, for free. A single image can be used, on Wikipedia, as visual support for more than one concept, in more than one language, and beyond Wikipedia, in other Wikimedia projects. For example, a recent study found that images of paintings are widely used to illustrate not only articles about visual arts, but also very popular articles about places, historical events and characters, as alternatives to photographs.
But adding images to Wikipedia can also have a positive domino effect on visual gaps. Algorithms and tools designed to address visual knowledge gaps also use the presence of images in articles as a signal to discover new matches between images and text. For example, our image recommendation algorithm, the new MediaSearch engine for image search, as well as many community initiatives that gamify the addition of images to Wikidata, use the existing image-article links to discover new image matches for natural language queries, unillustrated articles and Wikidata items.
Miriam Redi is a Research Manager at the Wikimedia Foundation and Visiting Research Fellow at King’s College London. Formerly, she worked as a Research Scientist at Yahoo Labs in Barcelona and Nokia Bell Labs in Cambridge. She received her PhD from EURECOM, Sophia Antipolis. She conducts research in social multimedia computing, working on fair, interpretable, multimodal machine learning solutions to improve knowledge equity.