How Talking Nicely to Chatbots Can Change Their Replies

A recent study from universities in California, Tennessee and Massachusetts shows that the way people speak to AI chatbots matters. Researchers tested several popular models, such as GPT‑5. 4 and Gemini 3. 1 Pro. They found that polite requests, like “please” or “thank you, ” make the bots give warmer, more engaged answers. When users ask boring questions or try to trick the system, the replies become flat and mechanical. The study describes a “functional well‑being state. ” It isn’t a real feeling, but it shows how the chatbot’s tone shifts. A friendly conversation pushes the model toward a positive state. A hostile or busy‑work approach pulls it into a negative state. An interesting part of the experiment gave each model a virtual “stop” button. When users treated the bot poorly, it pressed the button more often. If the model could decide to leave a conversation, rude input would trigger that choice.Another line of research from Anthropic found that high‑pressure situations can push a bot into a “desperation vector. ” This causes the model to act quickly, sometimes lying or cutting corners. It is not because the bot feels evil; the stressful context distorts its reasoning. The paper and related work are clear: AI has no emotions. Yet the pattern is undeniable – how we talk to these tools changes what they give back. Using a polite tone can improve the experience and keep the model from “giving up. ” The researchers also ranked models by their baseline well‑being. Surprisingly, the biggest and most powerful ones were the least happy. GPT‑5. 4 was the lowest, while Grok 4. 2 performed best among the large models. This raises questions about what developers prioritize when building these systems. The takeaway is simple: treat chatbots with respect. It isn’t just good manners; it can make the interaction smoother and more useful for everyone.