In the world of artificial intelligence, a fascinating question has recently been raised: “Is ChatGPT better if you’re nice to it?”
It was on the show Good Morning Business on BFM Business that journalist Anthony Morel explored this topic, notably as part of his segment Culture IA. This question raises fascinating issues about how we interact with language models and how they adapt their responses.
A revealing study from Waseda University
To better understand this dynamic, it’s worth looking at a study carried out by Waseda University in Tokyo. Researchers observed that language models, like ChatGPT, often align with their users’ communication norms.
In other words, the user’s tone and politeness can directly influence the quality of the AI’s answers. So, rude language can lead to an increase in errors in the generated responses.
On the other hand, overly polite communication can also be a problem. Researchers noted that while politeness is generally seen as an advantage, too much courtesy can lead to less accurate answers.
This raises the question of the right balance: how can you interact effectively with an AI without falling into either extreme?
Also read: Is ChatGPT always reliable? Here are the things to watch out for
The impact of language on answer quality
This study highlights a fundamental aspect of how language models work: their ability to adapt to the user’s communication style.
- When a user uses vulgar or overly casual language, the AI may end up drawing its answers from sources with a similar language level, which can hurt the clarity and relevance of the information provided.
- Conversely, by using clear, appropriate language, the user allows the AI to refer to data from more reliable, better-written sources.
This raises an ethical and practical question: can we make better use of these AI technologies by changing how we communicate with them? Maybe, to get the best results, making an effort to be polite and clear in our exchanges with machines could be beneficial!
Also read: Why doesn’t ChatGPT finish its sentences? Discover the solutions
Conclusion: be cordial with ChatGPT!
All in all, the question of kindness in our interactions with artificial intelligences deserves serious consideration. The Waseda University study, relayed by Anthony Morel, shows that how we communicate directly affects the quality of the answers provided by models like ChatGPT.
So, well-chosen language could not only improve our user experience, but also enrich the data and the answers we get. In a world where AI is taking an increasingly prominent place, it may be beneficial to adopt a thoughtful, respectful approach in our interactions with these technologies.
