A data scientist experimented with using incentives and penalties in ChatGPT prompts to see how they impacted output quality, but found no conclusive results. via minimaxir.com.