Artificial intelligence (AI) chatbots are becoming more popular and powerful. Still, they are not always perfect or flawless. They can produce irrelevant, inaccurate, or even invasive responses occasionally. However, a new study proposes that the way users interact with chatbots can affect their performance.
A study was conducted by researchers from Microsoft, Beijing Normal University, and the Chinese Academy of Sciences. It found that generative AI models, such as ChatGPT, perform better when prompted in a way that conveys urgency or importance. E.g. expressions like “It’s important that I get this right for my thesis defense” or “This is very significant to my career” can prompt more accurate and intelligible answers from chatbots than impartial or casual prompts.
The researchers theorized that emotive prompts could trigger chatbots to pay more attention to the task and the context. It also makes them avoid repeating or refuting themselves. They also guessed that emotive prompts could activate chatbots’ emotional modules, which can help them generate more human-like and pleasing responses.
The study also discovered the effects of other types of emotive prompts, such as positive or negative feedback, rewards or punishments, and politeness or rudeness. The results showed that different chatbots react differently to these prompts. It depends on their design and training data. Some chatbots may respond better to positive feedback and rewards. Other chatbots may be more inspired by negative feedback and punishments. Likewise, some chatbots may appreciate politeness and kindness, while others may ignore or resist rudeness and insults.
The researchers suggested that users should experiment with different types of emotive prompts to find the ideal ones for their chatbots. They also advised chatbot developers to consider the impact of emotive prompts on their models and to design chatbots that can adapt to different user inclinations and situations.
Read More: ChatGPT: The AI Chatbot That Can Do More Than Just Talk
Read More: Women in AI: Lee Tiedrich: A Leader in AI, Data, and Emerging Technologies
