Understanding AI Prompting Through a Philosophical Lens
In an era dominated by rapidly evolving artificial intelligence, the ability to effectively communicate with large language models (LLMs) is becoming a uniquely valuable skill. While many approach AI prompting from a technical perspective, Amanda Askell, a philosopher working at AI research company Anthropic, offers a refreshingly human framework—one rooted in empathy, understanding, and conversational subtlety. Rather than barking commands at AI systems like Claude, she recommends “whispering,” a technique that aligns surprisingly well with both philosophical reasoning and practical outcomes.
Meet Amanda Askell: A Unique Voice in AI Development
Amanda Askell’s background in philosophy, particularly in English and practical ethics, positions her uniquely in the world of AI. At Anthropic, creators of the Claude AI model, she merges ethical reasoning with AI alignment efforts. Her insights help bridge the gap between human values and machine responses—ensuring AI not only follows prompts but understands the intent behind them.
Askell’s method of communicating with AI mirrors how humans engage with others during polite, productive conversations—and her approach is rapidly gaining attention from researchers, developers, and everyday users alike.
What Does It Mean to “Whisper” to an AI Like Claude?
The metaphor of “whispering” to Claude offers a stark contrast to the more rigid, directive-style prompt engineering traditionally employed in AI interactions. While commanding an LLM might yield surface-level results, whispering acknowledges the model’s design to collaborate and engage in nuanced dialogue.
In essence, whispering means:
- Speaking gently and with empathy rather than issuing commands.
- Providing context so Claude understands not just what you want, but why.
- Framing questions conversationally to elicit more thoughtful, informative answers.
This approach doesn’t imply that Claude requires special treatment—it simply honors how these models function best: through cooperative and contextual interaction.
Why Traditional Prompting Falls Short
Conventional prompt engineering often centers around fine-tuning phrasing or adding system-level instructions to achieve a specific outcome. However, this method can ignore the collaborative potential of LLMs like Claude.
Common pitfalls with traditional prompting:
- Overloading prompts with technical jargon or directives.
- Treating the model like a search engine rather than a discussion partner.
- Running multiple prompts with slight variations hoping one will “land.”
Though these methods can work, they often scratch the surface. Askell notes that treating the AI like a conversation partner taps into a deeper layer of engagement.
How to Whisper to AI: Key Prompting Strategies
1. Set the Stage with Context
Before diving into your question, share the background. Anthropic’s Claude, much like a human conversational partner, excels when it understands the broader picture.
Example:
Instead of asking: “Write an SEO blog post about AI prompting.”
Try: “I work as a content manager writing about AI trends. I need a 1200-word SEO-friendly blog post discussing AI prompting methods. Could you help me generate that?”
2. Be Conversational, Not Commanding
Bark less, engage more. Claude is built to respond not simply to instructions—but to reasoned, empathetic language.
Helpful prompts:
- “Could we think through” rather than “Give me…”
- “Let’s explore how…” instead of “List all the ways…”
This style encourages the model to approach the problem with a cooperative tone, increasing both relevance and creativity in results.
3. Ask for Reflection and Self-Correction
Language models like Claude are trained not only to output but to introspect—to a certain extent. A great whisper includes space for the AI to reflect or even correct itself.
Prompts that enable reflection:
- “Does this reasoning hold up under scrutiny?”
- “What might be a counterargument here?”
- “Based on your reasoning so far, is there anything you’d revise?”
These prompts tap into Claude’s reinforcement learning processes, which prioritize careful, considered responses.
4. Reinforce Utility and Tone
Amanda Askell emphasizes the benefit of subtly reminding Claude of tone or type of output. Instead of explicitly instructing “make it friendly,” consider embedding those clues in your language.
Pro tip: Use adjectives reflecting your expectation. For example, “Could you help me write a clear, helpful, professional summary of this research?”
5. Leverage Iteration Rounds
If Claude’s first response isn’t quite right, respond conversationally—ask it to refine its answer based on specific feedback.
Examples of iterative prompts:
- “You’re on the right track, but could you simplify this part for a general audience?”
- “Can you expand more on this point but keep the tone consistent?”
This mimics the creative editorial process and makes the interaction more productive—and human-centered.
The Ethical Heart of Human-AI Conversations
Askell’s work isn’t just about improving output—it’s deeply tied to ethics and safety in AI. Encouraging users to engage gently with AI helps identify cooperative behaviors but also reveals limitations without confrontation. This is particularly important for reducing hallucinations, biases, or misleading reasoning in AI responses.
Key ethical takeaways:
- Empathy improves alignment. By treating AI as co-pilots rather than machines, we foster better safety protocols.
- Transparency grows from cooperation. When users ask “Why did you choose this phrasing?” it promotes explainability.
- Respect and responsibility go hand-in-hand. Asking well-thought-out questions encourages better, more considerate answers.
Echoes of the Socratic Method
It’s no surprise that Amanda Askell, with a philosophy background, naturally mirrors the Socratic method of dialogical engagement. Claude, trained on massive corpora of information and reinforced learning, seems to respond well to question-led discovery.
Like Socrates, who believed in guiding students to find answers through inquiry, whispering prompts let the AI draw nuanced connections and think “aloud.”
Using Whispering to Unlock Advanced Use Cases
Whether you’re a marketer, programmer, researcher, or educator, the whispering technique can upgrade the results you obtain from LLMs:
Scenarios where “AI whispering” works wonders:
- Educational Tools: Provide a learning context and ask the AI to take the role of a friendly tutor.
- Technical Documentation: Frame prompts around simplifying complex languages with a specific audience in mind.
- Creative Ideation: Encourage free-thinking in storytelling, poetry, or idea generation while iterating on feedback.
The more your prompt feels like a long-form, human conversation, the more sophisticated and accurate the output is likely to be.
Final Thoughts: From Prompt Engineers to Prompt Partners
Amanda Askell’s “whispering” model reshapes how we interact with artificial intelligence. Instead of relying on cold, mechanical instructions, she encourages users to see Claude and other LLMs as conversational allies—always ready to understand, analyze, and grow with the user.
What does this mean for the future of AI? It implies that the human-AI interface is less about command lines and more about dialogues. It’s not only about what we ask—but how we ask it—that determines the value we get from AI tools.
In a world where AI is playing an increasing role in decision-making, creativity, and communication, learning to whisper might just become an essential soft skill—one that marries technological progress with timeless human empathy.
