- Report this article
Jill R.
Jill R.
Marketing | Strategy | Growth
Published Nov 7, 2023
+ Follow
“AI performs better when you are nice to it.”
I heard this comment from a colleague recently and it caught me by surprise. It was said earnestly and offhandedly, as though it was common knowledge. I’d been using generative AI prompts for months that were direct and to the point. It never occurred to me to add a please or thank you. I wasn’t even giving clear constructive feedback such as, “that’s not quite right, how about …”. I simply adjusted the wording of my prompt and tried again. (I know better now.)
It turns out that AI likes feedback, especially positive and polite feedback. AI performs better, as we all do, when it enjoys working with you and understands what you like and don't like. In short, it collaborates better when we are nicer.
The Research Shows It
Really? You might be saying, as I did. Yup. This is an active AI research area and researchers are finding measurable differences in AI responses to flat or rude prompting versus polite and helpful prompts.
A study entitled Large Language Models Understand and Can be Enhanced by Emotional Stimuli shows that being nice to AI can significantly improve its performance. Specifically, positive and polite feedback can boost the performance of generative tasks by an average of 10.9% along performance, truthfulness, and responsibility metrics.
According to Tinglong Dai, a professor of Operations Management and Business Analytics at Johns Hopkins University, for example, “LLMs can produce superior results when users respond to their output with "Come on, you can do better than that!”
Recommended by LinkedIn
A paper by Google Deep Mind researchers showcased how providing an logical approach could help the AI deliver better results for quantitative tasks. According to an AI Business article summarizing the study, “The researchers discovered that the prompt "Take a deep breath and work on this problem step by step" was most effective with Google's PaLM 2 language model. This phrase reached the highest accuracy score of 80.2% when tested against GSM8K, a dataset of grade-school math word problems. In comparison, PaLM 2, without any special prompting, achieved only a 34% accuracy on GSM8K, while the classic prompt "Let’s think step by step" reached a 71.8% accuracy score.”
Why does politeness matter?
So why does AI respond better when we display manners in our prompts? In short, the answer is that, because AIs are trained on content from humans and are built to collaborate with humans, human-style requests yield better results. As Pradeep Viswav, director of the Microsoft Copilot team explains, AIs that rely on large language models (LLMs) are trained on our conversations and try to predict the next most likely words and phrases. Using polite language, including phrases like "please" and "thank you," sets a positive tone for AI, leading to more courteous and helpful replies. Conversely, impolite or provocative language may mean, as he puts it, that we “get some sass back or even trigger some unwanted behavior.”
Effective Feedback Strategies
Here are some approaches you can use to get the most out of your AI prompts by being nicer.
So, in addition to providing context and specifics to the AI, remember to show a little kindness and appreciation too. How we use and train these models is a reflection of what we value and the world we are co-shaping with AI.
#AI #PolitenessMatters #FutureTech #CollaborativeAI #DigitalEtiquette
AI Insights
AI Insights
399 followers
+ Subscribe
Like
Celebrate
Support
Love
Insightful
Funny
24
14 Comments
Donovan Rittenbach
Persuasive sales writing for websites • Neuromarketer helps clients find "buried treasure" in their funnels •Copywriting and Marketing Strategy • JedAIAuthor • Founder
6mo
- Report this comment
Works better with humans too!Our disconnected world needs better etiquette. (My pet peeve is ghosting.) Thanks for showing the science behind this too.
1Reaction
Steve Kazan
CEO, Founder, Partnership creator, Board advisor, Startup and scale-up CMO, CPO, Mentor, Business Problem Solver with Custom GPTs, dad, husband.
7mo
- Report this comment
Maybe that will rub off on our human interactions. ;) Hey Sandy Yu, Marnie Stockman, just FYI. #customerservice
1Reaction 2Reactions
Karina Mikhli
Optimized Services that Scale | Optimizer COO | Fractional COO - Professional Services, Community, SaaS | Workflow Consultant | Founder, Fractionals United | Advisor | Mentor
7mo
- Report this comment
Wow! This is amazing but makes a lot of sense. And I've been saying please and thank you and feeling weird about it, but glad good manners pay off even with AI :)
1Reaction 2Reactions
Martha Gerhan
7mo
- Report this comment
fascinating
1Reaction 2Reactions
Delly Tamer
Founder and CEO at Biztera
7mo
- Report this comment
Jill R. What a refreshing article. I’ve often found myself instinctively giving ChatGPT4 virtual pats on the back and wondered if I was wasting my time. Your piece has affirmed that acts of kindness can shape our interactions with AI.It’s welcome news to read that research underscores measurable differences in AI responses. The “how to be nice to AI” guide you included is a great resource. It’s prompted me to reflect on how I can enhance the quality of my feedback.Thank you for sparking such a thoughtful conversation around the human-AI rapport.
1Reaction 2Reactions
See more comments
To view or add a comment, sign in
More articles by this author
No more previous content
- The Promise and Peril of AI in Cybersecurity Jan 12, 2024
- 3 Steps to AI-Readiness in 2024 Jan 3, 2024
- AI-Powered: A 2024 Marketer's Guide for SEO Dec 18, 2023
- How Our Rapport with AI Is Maturing Dec 8, 2023
- AI: Power Monster or Power Genius? Nov 23, 2023
- Quick-Start Guide to AI for GTM Leaders Nov 15, 2023
- AI on the Rise: Boards Scrambling to Tackle Risk and Capture Value Oct 31, 2023
- Is your data ready for AI? Oct 25, 2023
- AI on the Rise: Boards Scrambling to Tackle Risk and Capture Value Oct 19, 2023
- Part 3 | How to market in a low-trust environment - Technology Jul 18, 2023
No more next content
Sign in
Stay updated on your professional world
Sign in
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
New to LinkedIn? Join now
Insights from the community
- Artificial Intelligence How can you improve AI system accuracy?
- Artificial Intelligence You’re trying to evaluate the effectiveness of an AI model. How do you know if it’s really working?
- Technological Innovation How can you collaborate with AI experts and developers?
- Artificial Intelligence What are the most effective ways to make AI algorithms more explainable?
- Artificial Intelligence What can you learn about the latest trends in explainable AI?
- Artificial Intelligence How do you increase AI user trust and satisfaction?
- Artificial Intelligence How can AI algorithms be made more reliable and trustworthy?
- Machine Learning What are some strategies for AI explainability and transparency?
- Artificial Intelligence What are the most important tips for AI consultants?
- Artificial Intelligence What do you do if your boss underestimates the game-changing potential of AI in your field?
Others also viewed
- Explainable AI: an introduction to the core concepts. Willem P. 4y
- Small Models, Big Impact: Steering the Course of AI Towards Super-Intelligence Ashish Bhatia 5mo
- Yann LeCun, One of the Top Three AI Gurus, Offers a Different Perspective Chaher Alzaman 2mo
- Like Humans, AI, Too, Hallucinates Sorab Ghaswalla 6mo
- The NEXT Generation of AI needs YOUR help! Bradley Taylor 4y
- AI Parameters: What are they anyway? Cengkuru Michael 1mo
- Demystifying AI and ML: What They Are and How They Work Tom Würzburg 1y
- 3 Initial Steps in AI Implementation George Neal 7y
- Humans In Machine Learning Sarbani Bose 3y
- Quick Thoughts on Being Intelligent in an AI World Varun Grover 8mo
Explore topics
- Sales
- Marketing
- Business Administration
- HR Management
- Content Management
- Engineering
- Soft Skills
- See All