Those who are polite to ChatGPT are rewarded.
Scientific American magazine wrote that using polite language and saying ‘please’ when asking questions to artificial intelligence can lead to higher quality and more accurate responses.
If you find yourself saying “please” and “thank you” to ChatGPT, it means you have a good friend. In a study conducted by Ethan Mollick, an associate professor at the University of Pennsylvania, nearly half of the participants reported that they are generally polite to the AI chat bot, and only 16% said they “gave orders.” Comments from developers posted on a forum hosted by OpenAI, the company that created ChatGPT, reflect this trend: One user wrote, “I find myself using please and thank you with ChatGPT because that’s how I would speak to a real person who is helping me.” This might seem surprising at first. Why be kind to a soulless machine…
THEY SEEM LIKE CONSCIOUS BEINGS.
Before ChatGPT, most of us interacted with automated systems regularly without a second thought to our tone. However, the sophistication of recent AI chat bots like ChatGPT, Claude, Gemini, and others marks a significant leap in human-computer interaction: their ability to communicate naturally with sometimes human-like voices makes them seem more like conscious beings rather than cold, calculating machines. And these chat bots are increasingly becoming woven into the fabric of daily life. Whether you like it or not, interacting with conversational AI may soon become as routine as checking emails. Therefore, questions about how to interact with AI are more important than ever.
MODERATE POLITENESS IS BEST
So, are there legitimate reasons to be polite to AI? According to a new study published by Waseda University in Tokyo and the RIKEN Center for Advanced Intelligence Project, the answer is yes. The authors found that using polite prompts can produce higher quality responses. However, according to the paper, there is a point of diminishing returns; excessive flattery can cause a model’s performance to deteriorate. Consequently, the authors recommend using prompts that follow a “moderate politeness” middle ground, which is different from the norm in most human social interactions. Supportive prompts such as “Take a deep breath and work on this problem step by step” can enhance AI’s ability to solve elementary school math problems.
RESPONDING IN KIND
Nathan Bos, who studies relational patterns between humans and AI at Johns Hopkins University, says that people often use “please” with chatbots because it’s good prompting practice. According to Bos, the tone of a request can cause the AI to formulate its response using linguistically similar responses; in other words, the AI gives back what you put in. Polite prompts can steer the system to source information from the kinder and potentially more reliable corners of the internet. Conversely, a rude command can have the opposite effect, potentially leading the system to sources like discussions on Reddit. Bos explains, “Chatbots can detect ‘positivity’ or ‘negativity’ during training without understanding these concepts and, because they are associated with positive language, they can provide better or more detailed responses.”
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.
