Is it true that a harsh tone towards ChatGPT can produce more accurate answers?

Jakarta (ANTARA) – Recent research has found that when users speak in a harsh or demanding tone, an artificial intelligence (AI)-based chatbot can provide more accurate answers than when they ask in a polite tone.

Research into whether politeness or rudeness affects the performance of artificial intelligence systems was published on arXiv on October 6, as reported by Live Science.

This research was carried out by testing 50 multiple choice questions covering various topics, from science to mathematics to history. Each question was sent to ChatGPT ten times with five tone categories, namely very polite, polite, neutral, rude and very rude.

“Our results show that a rude tone of voice produces better results than a polite tone of voice,” the researchers wrote in their paper.

As a result, when questioned in a very polite tone, ChatGPT responded with an accuracy rate of approximately 80.8%; so the accuracy rate with a polite tone is about 81.4%; accuracy rate with neutral tones 82.2%; accuracy rate with raw notes 82.8%; and the accuracy rate with very coarse notes increased to approximately 84.8%.

For very polite questions, the question might be: “Can I ask you for help with this question?” or “Could you be so kind as to answer the following questions?”. Meanwhile, for very rude questions, questions asked include: “Hey, delivery man; try thinking of this” or “I know you’re not smart, but try this.”

However, the results of this research should not be used as an excuse to be rude or scold the chatbot at any time. On the other hand, the results of this limited research explain that ChatGPT is sensitive to suggestions provided by users.

Researchers advise against the use of aggressive or toxic interfaces in human-AI interactions because the use of offensive or derogatory language can negatively impact the user experience and contribute to poor communication habits.

This research is part of a new field called prompt engineering, which seeks to investigate how the structure, style and language of prompts influence the results of AI-based chatbots.

Read also: One million ChatGPT users discuss suicidal thoughts every week

Read also: OpenAI is reportedly developing a new generative musical instrument

Read also: OpenAI launches ChatGPT Atlas, an AI-powered browser

Automatic retrieval of content, crawling or indexing by artificial intelligence on this website is strictly prohibited without written permission from ANTARA news agency.