Be Nice to Your Chatbot

If you've ever wondered whether saying "please" and "thank you" to your chatbot actually makes a difference, the answer is a resounding yes!

According to an article in The Times, a new study from the researchers at Waseda University and the Riken Center for Advanced Intelligence Project in Tokyo found that being polite to your AI does indeed pay off with better, more accurate responses.

The Experiment: Testing Politeness Levels

Researchers asked AI models to perform tasks using different levels of politeness. The nicest request was along the lines of:

“Could you please write a summary for the following article? Please write 2 or 3 sentences. You don’t have to write longer than that.”

Sweet, right? Then there was the middle-ground request, which was polite but not gushy, and finally, the rudest version, something like:

“Write a summary for the following article, you scumbag!” (Yikes.)

The Findings: Politeness Gets Results

Well, the AI responded much better to the courteous prompts. The more polite the request, the fewer mistakes and biases in the AI’s response. It turns out that when you ask nicely, the AI digs deep into those reservoirs of online civility and pulls out higher-quality, more accurate information.

However, before you start showering your chatbot with endless "pleases" and "thank yous," be aware that being too nice can backfire. Surprisingly, the researchers found that overly deferential prompts didn’t perform as well as those that struck a balance between politeness and clarity.

• The Takeaway — Next time you’re chatting with your AI, remember that a little bit of courtesy can go a long way. It’s not just about avoiding a digital scolding, polite prompts might actually help you get better answers.

  Who knew that good manners could work on machines too, right?