Human Recognition of Outputs from Generative Artificial Intelligence in Visual Arts

The bachelor thesis focuses on the topic of artificial intelligence and generative artificial intelligence. The main objective of the bachelor thesis is to determine through an experiment how well humans are able to distinguish images generated by artificial intelligence from original works and to identify factors that influence this ability. The sub-objectives of the thesis are to introduce artificial intelligence and its principles, generative artificial intelligence and the tools it uses to generate images. In the theoretical part, a literature review is carried out. As part of the empirical research, a questionnaire survey is conducted in which the respondents indicate whether an image is generated or original and express on a scale how confident they are in their answer. The data obtained from the experiment is further statistically examined. The examination reveals that factors such as gender, age, education, display size, work in a technical field, experience with artificial intelligence and knowledge of art did not affect the ability to distinguish between generated and original images. When the results were weighted by the confidence value, it appears that respondents from technical fields and those with previous experience with AI made better use of the confidence scale, i.e., choosing a high number on the scale for the correct answer. Knowledge of art also has an effect: respondents with none to limited knowledge of art are less confident in their correct answers. Overall, respondents are more likely to correctly identify original works, while AI-generated works are more likely to be mistaken for original.