04

Images as a way of spreading disinformation memorably

The format of disinformation is rarely talked about in this context. However, discussion of the format is crucial for assessing how quickly and sustainably disinformation influences people, as well as for effective methods of combatting it – including the use of automated systems. Currently, the discourse often assumes that disinformation is primarily textual in the form of fabricated news.

Recent studies depict a different image: The Brazilian Comprova project – a coalition of newsrooms and journalists – studied the proliferation of information via messenger services prior to the last presidential election. Comprova debunked mis- and disinformation, fact-checked and verified information and analysed message formats. They found a majority of messages included pictures aligned to a political position. Just as popular were official documents taken out of context and screenshots of digital conversations both real and fake. While the data contained few memes, Comprova assumes memes weren‘t taken seriously and thus not submitted for review10, despite the fact that memes especially distribute propaganda through humour11. Propaganda is (unwittingly) proliferated through memes by ridiculing an opponent, often presenting racist, sexist, anti-Semitic or other stereotypes as funny. Another method misquotes people depicted in memes. Videos were the second most common format, with voice messages making up the smallest percentage12. Text recognition software can be used to highlight or even delete inaccurate information in articles or texts. Links to websites allow the recipient to review the website13, and potentially recognise disinformation for what it is. When presented with images, videos or voice messages, this process becomes difficult – if not impossible. Manipulated audio or video recordings, so-called deep fakes, obscure the truth of a message further.

As early as the 1970s and 1980s, photography and TV programmes were used to show that human perception of text and images is fundamentally different. The brain can process images significantly faster than text. Our ability to critically review content is significantly lower for moving and still images. For instance, memes are generally received as humorous, rather than manipulative or politically motivated. Memes played a significant role in the 2016 US presidential election. Anti-Clinton memes were often tested on websites like 4Chan. Those that garnered the best response would be published on Reddit in the pro-Trump forum ‚The_Donald‘. Some were picked up by the Trump campaign and shared via better-known and widely used social media networks14. Because Facebook‘s newsfeed algorithm favours images and videos over links, memes and other images or videos have significantly greater reach through Facebook. Memes, pictures and video can be shared from the newsfeed via Facebook Messenger and as a link or screenshot via other messenger services like WhatsApp. The non-governmental organisation First Draft supported projects that debunked disinformation during the 2017 French and UK elections. They found that still and moving images, including infographics and memes, accounted for the largest percentage of shared disinformation and were the hardest to disprove. However, as opposed to the 2016 US election, no websites posing as news organisations with purely fictitious ‚‘news‘ were found15.

10 Wardle et al. 2019. 11 Cf. Ascott 2020. 12 Cf. Wardle et al. 2019. 13 For example, whether there is a legal notice, a mentioned author and by analysing other articles or information present on the site. 14 More on this subject, among others, from Moore 2018. 15 Cf. Wardle and Derakhshan 2017.