In the first study, ChatGPT 3.5 and a group of U.S. adults were asked to complete a comedy writing task. Then, all the answers were presented anonymously to a separate group of adults, who were asked to rate the funniness of each response. They thought that ChatGPT was funnier, with 69.5 percent preferring them while 26.5 percent preferred the human responses and 4 percent found them equally humorous.
In the second study, ChatGPT was shown a selection of satirical headlines from The Onion since Oct. 1, 2023. It was then asked to write 20 new headlines, and a group of USC psychology students were presented with all the headlines and asked to rate how funny they were. Most preferred The Onion’s headlines at 48.8 percent, but ChatGPT’s were not far behind at 36.9 percent, showing the LLM can write at least similarly funny headlines to professional comedy writers.
“Since ChatGPT can’t feel emotions itself, but it tells novel jokes better than the average human, these studies provide evidence that you don’t need to feel the emotions of appreciating a good joke to tell a really good one yourself,” said the study’s lead and corresponding author Drew Gorenz, a doctoral candidate in social psychology at the USC Dornsife College of Letters, Arts and Sciences, and an amateur standup comedian.
