Design of a Computer Experiment to Assess Task-Irrelevant Parts in AI Instructions (Working Title)

This study aims to compare reaction times and accuracy of participants who are exposed to 10 different texts containing instructions on daily tasks and must answer 3 decision tasks about the text, with half of the texts being written by humans and the other half being generated by an AI chatbot, Gemini 3 Pro, in this specific case. The context of this study is that texts generated by Chatbots tend to be longer than human-written texts and contain much more irrelevant information (fluff) than needed to complete the task. The length of AI-generated texts can cause cognitive overload, leading to longer reaction times and a change in accuracy. In this within-subjects experiment, it will be tested whether the mean reaction times in AI texts are indeed higher when compared to performance in questions on human texts. The sample will mainly consist of German-native or German-fluent students at the Heilbronn campus of the Technical University of Munich, with a calculated N = 34 for a two-sided t-test with 80% power and a significance level of 0.05. unknown other