Novelty Given that LLMs are designed to generate approximately the statistically most plausible sequence of text based on their training data, perhaps they generate less-novel ideas. Novelty is not a goal expressed in the prompt for either humans or Chat-GPT. It is typically not a primary objective in commercial product development efforts, nor does it have commercial value in itself. Still, we are curious about how the novelty of ideas varies between LLM- generated ideas and those generated by humans. We adopt the survey instrument of Shibayama, Yin, and Matsumoto (2021) to assess the novelty of the ideas. mTurk respondents answered this question: Relative to other products you have seen, how novel do you consider the idea for this new product? 1. Not at all novel 2. Slightly novel 3. Moderately novel 4. Very novel 5. Extremely novel We weigh these responses 0, 0.25, 0.50, 0.75, and 1.00 to produce a novelty score for each idea. By this measure, the mean novelty of the human-generated ideas is higher than that of the LLM-generated ideas (p