Help AlgorithmWatch map Instagram's political impact

Instagram is one of the world's biggest and fastest-growing online platforms. As such, it's an attractive medium for politicians and political parties, especially those hoping to reach a young audience. Like other countries, the Netherlands is seeing a surge in the number of political Instagram accounts. And, while the 2016 Cambridge Analytica scandal has spotlighted the problem of misinformation on Facebook, a report by New York University suggests that Instagram is potentially a bigger source of (political) misinformation during elections than Facebook, YouTube or Twitter.

"Nevertheless, much less research has been done into Instagram than into other platforms," says Nicolas Kayser-Bril of AlgorithmWatch. His internationally active non-profit organisation studies the working of algorithms and their impact on society. AlgorithmWatch has recently started a project called Towards a Monitoring of Instagram (TAMI) to gather information about Instagram's political impact and demonstrate the possibility of data-driven research into the platform.

Instagram presents data collection challenges

According to Kayser-Bril, there is a technical background to the lack of research into Instagram. "Of the main social media platforms, Twitter offers by far the most scope for research by means of APIs [Application Programming Interfaces, technical interfaces for communication between applications, ed.]," he explains. "YouTube and Facebook are also reasonably accessible to researchers, although Facebook has introduced certain restrictions since the Cambridge Analytica scandal." Because access to data is easier with Twitter, YouTube and Facebook, nearly nine out of ten academic studies into social media concern themselves with one of those platforms. As a result, much less is known about how other platforms work. "Instagram is geared primarily to mobile use, which makes it harder to gather information," says Kayser-Bril. "There is a desktop version, but it doesn't have all the features of the mobile app, and we don't know whether the algorithm works the same way." Also, the focus of Instagram has shifted from the Newsfeed to Stories, where posts remain visible only for twenty-four hours. "Gathering data from Stories is harder, because they are only visible to logged-in users," says Kayser-Bril.

Data donation via a plugin

Instagram doesn't provide any APIs for research and doesn't respond to requests, either for information about its algorithm, or for interviews with its developers. AlgorithmWatch has therefore adopted an alternative approach to finding out about the Instagram algorithm. They have developed a plugin for the Chrome and Firefox browsers, which Instagram users can install in order to donate data for research. "Plugin users are asked to follow three particular politicians' accounts," continues Kayser-Bril. "The plugin then regularly checks the user's Instagram Newsfeed. If any posts by the three accounts appear in the feed, details are forwarded to the database." Kayser-Bril acknowledges that the approach does have its limitations: the plugin gathers information only from the desktop version of Instagram and doesn't yet collect Stories. "Nevertheless, it does enable us to systematically collect data and get an idea of what's happening on Instagram," he adds.

Earlier study showed that bare skin scores on Instagram

AlgorithmWatch has previously used the plugin to demonstrate that Instagram gives prominence to photos where bare skin is visible. "A picture of a woman in a bikini was 54 per cent more likely to appear in our volunteers' newsfeeds than a picture of the same woman fully clothed," says Kayser-Bril. "A man's bare-chested photo was 28 per cent more likely to show up than a photo of the man wearing clothes. By contrast, photos showing food and landscapes had 60 per cent less chance of being featured." The study was based entirely on posts by professional content creators, who tend to be highly reliant on Instagram for their earnings. AlgorithmWatch is now planning to use the same approach to map Instagram's political impact. "We want to test a number of hypotheses," says Kayser-Bril. "Does the algorithm reward content for being more extreme, for example? Or is Instagram becoming more news-oriented by bumping up photos that have long captions? And are politicians trying to secure follower loyalty by sharing photos from their private lives on Instagram?" AlgorithmWatch has teamed up with media partners NOS, Pointer and de Groene Amsterdammer to recruit people prepared to use the plugin for a few weeks or months. The aim is to shed light on Instagram's political influence in the run-up to the Dutch elections in March 2021.

Support from SIDN Fund

SIDN Fund and Adessium Foundation awarded the TAMI project a grant following a joint call linked to the theme of Public Values in a Digital Society. Explaining why the two organisations wanted to back the initiative, SIDN Fund's Programme Manager Mieke van Heesewijk says, "Before you can develop responses to microtargeting or online influence, you need to know how online platforms' algorithms work. And AlgorithmWatch has come up with a very good way of studying them. We're also very pleased about the media partnership with NOS, Pointer and de Groene Amsterdammer, which opens the way for clarifying Instagram's political influence ahead of our spring elections." Fancy installing the plugin? It's available to download from the AlgorithmWatch website.