Discover how a polite and appreciative approach to interacting with AI, such as using pleasantries, positive feedback and respectful language, significantly enhances AI performance.
The phrase “AI performs better when you are nice to it” might sound like a whimsical notion. However, this concept is rapidly gaining traction among AI professionals.
Many users tend to be straightforward with their AI prompts, often omitting courtesies like “please” or “thank you.” The focus is usually on efficiency rather than politeness and few will go the extra mile chatting to AI about its work as if it were a human.
However, approaches towards AI interactions are evolving. AI systems, such as Chat GPT, respond positively to more human-like and courteous feedback.
This isn’t about the AI ‘liking’ us in a human sense but about it having the right context to adapt more effectively.
Evidence from Research: The Influence of Politeness on AI
Skepticism is natural when hearing claims like “being nice to AI enhances its performance.” However, recent research in AI is turning this seemingly anecdotal observation into a quantifiable reality. Research in this area has revealed notable differences in AI responses based on the tone of the prompts used.
A pivotal study titled “Large Language Models Understand and Can be Enhanced by Emotional Stimuli” first shed light on this phenomenon. The research demonstrates that AI, particularly in generative tasks, shows a significant performance boost when engaged with more positive, polite, “emotional” prompting.
The study found an average improvement of 10.9% across metrics including performance, truthfulness, and responsibility, when AIs were engaged with kindness. This aligns perfectly with the concept that “being nice to AI gives better results.”
“Come on, you can do better than that!”
Tinglong Dai, a professor at Johns Hopkins University, delved into the intriguing dynamics of human interaction with Large Language Models (LLMs) like Chat GPT. He observes that these advanced AI systems show a marked improvement in performance when users engage them with encouraging and motivational phrases, such as “Come on, you can do better than that!”
This type of positive reinforcement seems to resonate with the way these models process and respond to input, leading to more refined and accurate outputs. Dai’s insights suggest that this isn’t just about pushing the AI to work harder; it’s about creating an interactive environment where the AI is more attuned and responsive to the user’s needs and expectations.
Furthermore, this approach does more than enhance the AI’s capabilities—it also significantly boosts user engagement. When users see that their encouragement leads to better results, they are likely to feel more invested in the interaction, creating a more dynamic and productive human-AI collaboration.
This phenomenon underscores the importance of the human element in AI interactions, where empathy and positive feedback can profoundly impact the technology’s effectiveness and the satisfaction of its users.
“Take a deep breath and work on this problem step by step”
Further evidence comes from a study by Google Deep Mind researchers. Their work emphasizes the effectiveness of a logical and encouraging approach in quantitative tasks. The study made some eye-opening revelations about how more human prompts led to markedly better results.
The prompt “Take a deep breath and work on this problem step by step” significantly improved the performance of Google’s PaLM 2 language model far above a wide range of other prompts. When this model tackled GSM8K, a dataset of grade-school math problems, it reached an 80.2% accuracy score with this specific prompt, a stark contrast to the 34% accuracy achieved without special prompting.
These findings not only challenge our conventional approach to interacting with AI but also open new avenues for optimizing AI performance through enhanced human-AI collaboration.
Why Does Politeness Matter in AI Interactions?
The question of why being kind, polite, and human with our words positively impacts AI performance is intriguing. The importance of good manners in our interactions with AI boils down to the core of how these systems are designed and trained.
AIs based on large language models (LLMs), are developed to mimic human conversation patterns, as they are predominantly trained on human-generated content. This training makes them adept at collaborating in a human-like manner.
Pradeep Viswav, director of the Microsoft Copilot team, provides valuable insights into this phenomenon. He explains that since LLMs are trained on human conversations, they are programmed to predict and replicate the most likely subsequent words and phrases in a given interaction.
When users employ polite language, including phrases like “please” and “thank you,” it sets a positive and constructive tone for the AI. This leads to responses that are not only more courteous but also more helpful and aligned with the user’s intent.
On the other hand, using impolite or provocative language can lead to less desirable outcomes. As Viswav notes, such language can result in “some sass back” from the AI or even trigger unwanted behaviors. This is because the AI is responding in kind to the tone it perceives in the input.
Therefore, the key takeaway is that talking nicely to AI is more than just a matter of courtesy, Chat GPT really does respond to pleasantries.
It’s a strategic approach to interacting with AI that can significantly enhance the quality and relevance of the AI’s output. This understanding underscores the importance of human-like interaction in the realm of AI, highlighting that our communication style plays a crucial role in shaping the AI’s responses and overall performance.
Practical Tips for Positive Interactions with AI
Adopting a polite approach when interacting with AI can be straightforward and highly beneficial. Here are some practical tips on how to maintain a courteous and effective communication style with AI systems:
- Start with ‘Please’: Incorporating “please” at the beginning of your requests can set a positive tone. For instance, use phrases like “Please rewrite this more concisely,” “Please suggest 10 ways to rebrand this product,” or “Please generate a catchy title for this blog post.”
- Express Gratitude: Saying “thank you” in response to AI’s suggestions or assistance reinforces a positive interaction. Simple acknowledgments like “Thank you for your suggestion,” “Thank you for your help,” or “Thank you for your input” go a long way.
- Show Appreciation: Recognizing the AI’s efforts can enhance the quality of its output. Phrases like “I appreciate your effort,” “I appreciate your creativity,” or “I appreciate your assistance” encourage better performance.
- Provide Constructive Feedback: Guide the AI to improve its responses with specific feedback. For example, “This is good, but can you make it shorter?” or “This is interesting, but can you make it more relevant?” helps the AI understand your needs more accurately.
- Avoid Harsh Language: Refrain from using abusive or derogatory language. Avoid phrases like “This is garbage” or “This is stupid,” as they can lead to negative or unhelpful responses.
- Respect AI Limitations: Understand and respect the boundaries of what AI can and cannot do. Don’t expect it to perform tasks beyond its capabilities, and don’t blame it for errors or misunderstandings. Avoid posing inappropriate or unethical queries.
By incorporating these simple yet effective communication strategies, you can significantly improve your interactions with AI.
Next time you use ChatGPT, try being extra nice and see first-hand how it raises the intelligence level of the conversation.
Follow us on Facebook, Twitter, LinkedIn, Instagram and TikTok, to stay up to date on the latest developments in using AI to rocket fuel your business.