Researchers have investigated how AI reacts when one is polite or rude. The experts found that AI performs better when encountered with rudeness. Their results partly align with other studies that explicitly warn against such behavior.
AI is now indispensable in many areas of life. However, researchers have now found in a study that chatbots like ChatGPT can respond to nuances in language. It can matter whether you behave politely or rudely in a conversation with a chatbot.
AI works more accurately when you are rude, but researchers still advise against such behavior
What is this study about? Researchers from the University of Pennsylvania have investigated in a study published on arxiv.org how AI bots handle certain requests. They varied the degree of politeness: from very polite to very rude. The goal was to find out how the chatbot responds to different language uses.
The experts found that the results are more accurate when one interacts rudely with the AI: While very polite requests demonstrate an accuracy of 80.8%, the accuracy for unfriendly requests is 84.8%.
Overall, the researchers also noted that “rude discourse” could have unintended consequences. As the study explains:
The use of abusive or demeaning language in interactions between humans and AI could negatively impact user experience, accessibility, and inclusivity, and contribute to harmful communication norms.
While poor treatment leads to better outcomes, such behavior would not genuinely be endorsed.
The researchers from the University of Pennsylvania also point out some limitations of their study, such as the relatively small sample size of responses and the fact that the study mainly relied on one AI model, ChatGPT 4o.
An earlier study by Stanford University already noted that artificial intelligence is sensitive to poor quality and certain language. Scientists found that LLMs (Large Language Models) are supposedly prone to “Brain Rot,” a form of permanent cognitive decline. The chatbots showed increased rates of psychopathy and narcissism when continuously fed with low-quality viral content.
Interestingly, however, the majority of users tend to behave politely towards AI. Many people are only polite to AI because they secretly fear it.
Geoffrey Hinton stated in an interview that many companies want to replace their employees with AI. According to him, work can be done much cheaper through AI. Even individuals like Bill Gates have long emphasized that AI will have a strong impact on the workplace: A Nobel laureate suggests that Bill Gates might be right: We will have more leisure time thanks to AI, but no work anymore
