Real or AI? – Internets nya “real or cake”. Unga vuxnas uppfattningar om sin egen och andras förmåga att identifiera AI-genererat innehåll och förtroende för sociala medier

Abstract

This study explores how young adults in Sweden perceive their own and others' ability to identify AI-generated content on social media, as well as their trust in social media in general. To accomplish this, a qualitative method was chosen in the form of semi-structured interviews with 12 young adults aged 21–24, with an equal distribution between women and men. A theoretical framework consisting of three separate theories is used, each primarily connected to one research question: digital literacy theory (Buckingham, 2015), the third-person ef ect (Davison, 1983), and online information credibility (Metzger & Flanagin, 2013). The study addresses the following research questions:

How do young adults perceive their own ability to identify AI-generated content on social media? This question focuses on examining young adults' subjective experiences of digital competence, particularly how they perceive their own ability to determine what is generated by AI. It aims to capture their experiences and reasoning related to this competence. How do young adults perceive others' ability to identify AI-generated content on social media? The question aims to examine how young adults perceive others' ability to identify AI-generated content on social media, and how they describe the relationship between their own abilities and those of others. How do young adults perceive that their trust in content on social media is af ected by the presence of AI-generated content? This question focuses on examining how young adults themselves perceive that the presence of AI-generated content affects their trust in social media content. The question is open to capture different types of experiences and possible changes in trust. The results show varying levels of confidence in participants’ own ability to identify AI-generated content. While some feel confident, others describe their skills as limited to detecting "obvious" AI, with more sophisticated content is harder to identify. Strategies used by the interviewees to identify AI-generated content include visual assessments (searching for unnatural features), contextual strategies (examining the intent of the sender), and social verification (reading the comment section to observe others’ reactions). The study also shows that even interviewees who feel confident in their abilities experience an underlying doubt due to the difficulty of determining whether content is AI-generated or not. The third-person effect is confirmed by the study, which implies that the interviewees generally perceive their own ability, as well as the ability of people with similar attributes, to identify AI higher than the public. The interviewees believe that people in general overestimate their source criticism and ability to identify AI-generated content, and that the public is more prone to be deceived by AI. Elements and attributes such as young age and education are highly cited by the interviewees as crucial for the ability to identify AI. For the final research question, the results show a decreased trust in social media in general due to the presence of AI-generated content. The knowledge that AI exists in social media content creates a fundamental attitude where the interviewees generally assume that content can be manipulated, that also causes distrust toward genuine content. The trust varies depending on platform, where TikTok is associated with low trust due to its fast pace and usage of filters, while Instagram and Facebook are perceived as slightly more trustworthy due to an increased simplicity of checking the sender’s profile. In conclusion, the rapid technological development has made it increasingly difficult for young adults to identify AI-generated content on social media, causing uncertainty about what is authentic. This uncertainty contributes to declining trust in social media, as awareness of AI causes skepticism rather than confidence. The study also confirms the third-person effect, with the interviewees perceiving themselves and people similar to them as more critical and more capable of identifying AI than others. Overall, the study shows that young adults view themselves and their peers as capable of identifying AI-generated content, while expressing low trust in social media.