Kvalitativní srovnání uvažování studentů FIS VŠE při rozpoznávání fotografií upravených umělou inteligencí

This bachelor’s thesis examines the perception and recognition of photographic edits made by humans and artificial intelligence. The aim of the thesis is to compare how different groups of respondents view these edits and how they evaluate the quality and authenticity of the resulting photographs. The research focuses on comparing students majoring in Multimedia in Economic Practice at the Faculty of Informatics and Statistics of the University of Economics in Prague with students from other programs at the same faculty who lack specialized knowledge in the field of photography. This study is based on an overview of the principles of digital photo editing, specifically exposure adjustment, color correction, selective editing, and compositional corrections. It also draws on the potential applications of artificial intelligence in visual editing and insights from the field of visual perception. These insights include Gestalt principles, perceptual segregation, and the perception of a scene’s essence. These foundations are further developed through qualitative research conducted in the form of semi-structured interviews. As part of the experimental section, respondents compared pairs of photographs edited by the author of the thesis and by artificial intelligence, and subsequently described their strategies, ways of thinking, and criteria on the basis of which they evaluated the edits. During the interviews, respondents compared a total of 30 pairs of images divided into three categories: portraits, architecture, and nature. Eight respondents participated in the study, four in each group, with an average interview duration of 24.5 minutes. During the subsequent thematic analysis, 12 main codes were identified, which were further divided into thematic areas. Four of these were identified: the effectiveness and practical use of AI, the role of human intent and experience, the influence of the type of photograph on the recognition of edits, and the evaluation of the naturalness and quality of edits. The research results show that differences between groups of respondents are primarily reflected in their reasoning and the strategies they use to detect edits, rather than in their overall success in detecting edits made by artificial intelligence. Respondents with knowledge of photography thought in a more structured way and relied more on the technical aspects of the image, while respondents without such knowledge made decisions primarily intuitively based on their overall impression. Edits made by artificial intelligence were rated as technically high-quality and often convincing, but human edits were perceived as more meaningful, particularly regarding the author’s intent and the naturalness of the result.