Together with colleagues of mine, I conduct mix-method experiments, both qualitative and quantitative studies, regarding human-AI interactions. These are the result of interdisciplinary research spanning various fields, including philosophy, experimental psychology, linguistics, and computer science. Below, you can find information and links to whichever studies are currently active.
Strategies for Determining Human- from AI-generated Text
For this study, we trained a Large Language Model (GPT-3, like ChatGPT) on the philosopher Dan Dennett's work and posed it 5 philosophical questions, which we also asked Dan to answer. For each philosophical question, there is one answer by Dan and 4 by the LLM. Your task is to try to determine which one was actually written by Dan. We also ask you to describe how you tried to determine which one was written by an actual philosopher.
Link to study
This study is being conducted by Anna Strasser (DenkWerkstatt Berlin, LMU-Munich), John Dorsch (LMU-Munich), and David Schwitzgebel (Institut Jean Nicod, École Normale Supérieure-PSL).