Should You Say Please to LLMs?

The question of politeness toward AI has moved from novelty to a testable prompt design choice. The topic surfaced in a lively Hacker News discussion, flagged on Hacker News last week per a recent thread. The thread compiled 12 points and 31 comments, evidencing broad interest but no consensus. This article treats the topic as a practical prompting question, not a moral thesis, and offers concrete steps for experimentation, plus comparisons to established prompt-design patterns.

What It Is / How It Works

Politeness in prompts refers to explicitly instructing an LLM to behave in a courteous or helpful manner, for example by including phrases like “please,” “thank you,” or a directive such as “be extremely polite.” In practice, politeness is a form of instruction-tuning signal. When users embed social norms in prompts, models trained with alignment objectives that value user satisfaction may adjust tone, responsiveness, or willingness to cooperate accordingly. The discussion in the source thread highlights that readers view politeness as a potential UX knob rather than a guarantee of higher factual accuracy or reliability. For context, consider how system prompts and instruction-following cues shape model behavior in documented workflows: prompts can set goals, tone, and constraints beyond raw task instructions. See OpenAI’s chat guidance and design docs for how prompts shape behavior in practice.

• References that frame the broader mechanism: system prompts and instruction-following practices in established docs and blogs.
• Real-world takeaway: politeness is a UX lever, not a proven performance amplifier.

Benchmarks / Specs / Numbers

There are no formal benchmarks quantifying the effect of politeness in prompts across models. The Hacker News thread itself serves as anecdotal evidence, not a controlled study. The thread’s 12 points and 31 comments reflect divergent opinions, not standardized metrics. Practically, expect at most qualitative signals: length of replies, tone alignment with user cues, and variation in explicit acknowledgment or gratitude in responses. In other words, this topic currently lives in user-experience territory rather than strict performance benchmarks. For readers requiring numbers, it’s best treated as a hypothesis to test within your own prompts and datasets.

How to Try It

Pros and Cons

• Pros
  • Improves perceived user experience when interacting with AI.
  • Low-cost experiment; easy to toggle in prompts or templates.
  • Can harmonize interactions across diverse user groups and contexts.
• Cons
  • May introduce bias toward over-politeness or hedging, affecting decisiveness.
  • No guaranteed improvement in factual accuracy or usefulness.
  • Effects are model- and task-dependent; not universally beneficial.

Alternatives and Comparisons

Contextual note: the broader prompt-design ecosystem includes system messages, instruction-following frameworks, and user-facing templates. For readers who want to anchor tone reliably, system prompts and explicit tone controls often outperform ad-hoc politeness cues in long-running sessions. See OpenAI’s guidance on chat prompts and prompt design for more on these mechanisms. For background reading on how prompts shape behavior in practice, consult the linked OpenAI docs and related resources.

Who Should Use This

• Product teams and UX researchers exploring human-AI interaction, where perceived helpfulness matters as much as raw accuracy.
• Developers building consumer-facing chat assistants who want a friendlier tone without sacrificing task fidelity.
• Educators and researchers studying prompt design or human-computer interaction, who can run controlled A/B tests to gather data.
• It may be less advantageous for domains requiring highly terse, risk-averse, or strictly factual outputs, where extra hedging or politeness could degrade signal clarity.

Bottom Line / Verdict

Politeness in prompts is a practical UX knob rather than a proven performance lever. The Hacker News thread shows wide-ranging opinions but no consensus or formal benchmarks. For teams, the prudent path is to treat politeness as a lightweight, testable variant in your prompt design toolkit: implement polite and neutral templates, run small A/B tests, and measure user satisfaction, answer usefulness, and any shifts in risk or bias. In parallel, lean on system prompts and instruction-following patterns to control tone more reliably across sessions. If you’re building a conversational tool, politeness warrants lightweight experimentation; if you’re chasing absolute precision, keep politeness as a secondary signal and prioritize clarity and accountability in your prompts.

CLOSING
Politeness as a design choice belongs to the broader spectrum of prompt engineering—worth trying, worth measuring, and worth comparing against more established controls like system prompts and explicit task constraints. The real value lies in evidence from your own tests, not anecdotes from a single thread.

FURTHER READING

• Hacker News discussion: https://news.ycombinator.com/item?id=48945125
• OpenAI Chat API docs: https://platform.openai.com/docs/guides/chat/introduction
• OpenAI Prompt Design docs: https://platform.openai.com/docs/guides/prompt-design
• OpenAI Instruction Following: https://openai.com/blog/instruction-following
• Human-Computer Interaction — Wikipedia: https://en.wikipedia.org/wiki/Human%E2%80%93computer_interaction
• Politeness — Wikipedia: https://en.wikipedia.org/wiki/Politeness