The rapid spread of inauthentic images online, known as deepfakes, raises the question of whether internet users can recognize such content. Therefore, the aim of this master's thesis is to examine how internet skills and algorithmic literacy are related to users' ability to identify deepfake images created with artificial intelligence. We addressed two research questions: the first concerned the relationship between different types of internet skills and the ability to recognize deepfake images, and the second focused on the connection between this ability and users' awareness and knowledge of algorithms. We conducted an empirical study collecting data with a web survey on a convenience sample (n = 381) of internet users in Slovenia. The questionnaire included self-assessment scales for internet skills and algorithmic literacy, as well as a newly developed test-based scale asking respondents to indicate which of the two images in 12 pairs was a deepfake photo. The results showed that respondents were likely to recognize more obvious deepfakes, while they were, on average, less successful in identifying highly realistic ones. In the easier cases, older respondents performed worse, whereas in the more difficult cases, higher creative internet skills and higher algorithms awareness led to greater success. These findings confirm that advanced internet skills, particularly creative skills, and awareness about online algorithms significantly contribute to more effective recognition of harder-to-detect deepfakes.
|
