Entry #188

With the rate at which Artificial Intelligence models are advancing, media literacy and critical analysis of multiple formats of information are vital skills. The current study sampled 87 university students and sought to measure their ability to detect AI-altered/generated images. The study involved a treatment portion with AI literacy tips and training that allowed for participant engagement with the AI-artifacts in the image. They then completed a post-treatment survey to evaluate for change in scores. The research hypothesized that there would be a significant difference between the participant’s pre-treatment and post-treatment scores. A paired-samples t-test showed that post-treatment survey scores (M = 11.66, SD = 2.25) were significantly higher than pre-treatment scores (M = 9.05, SD = 2.61), t(86) = -10.19, p < .001, 95% CI [-3.12, -2.10]. The training produced a substantial improvement in scores, isolating for individual differences such as demographic variables, demonstrating a large effect size with Cohen’s d = 1.09. These results show a strong ability to teach audiences how to recognize images that have been generated or altered by AI, though the applicability to the generalized population of the specific tips provided in this study is questionable as image-generating AI models will likely phase out many of the included AI-artifacts.