Stop Sloppypasta: Don't paste raw LLM output at people  🔗 **Source:** [Stop Sloppypasta: Don't paste raw LLM output at people](https://stopsloppypasta.ai/en/) > [!note]+ Document Notes > tags: #ai-ethics #communication-etiquette #llm-usage > author: [MISSING] > source: stopsloppypasta.ai > reference: https://stopsloppypasta.ai/en/ > date created: 2024-12-19 16:22 > > # 🏷️ Title: Stop Sloppypasta: Don't paste raw LLM output at people > <!-- deep-read:sections start --> ## 🧰 Action Inbox - [H1] Establish team guidelines for AI-assisted communication — review current practices #communication-etiquette - [H2] Create disclosure templates for AI-generated content — standardize transparency #ai-ethics - [H3] Audit recent messages for sloppypasta patterns — improve future interactions #professional-development ## 🧭 Table of Contents - [[#📝 Summary]] - [[#📖 Main Content]] - [[#🗒️ Notes]] - [[#📚 References]] - [[#🗣️ Paraphrase]] - [[#📰 Newsletter]] - [[#✅ Action Items]] - [[#🧪 Hypotheses]] - [[#📊 Methodology]] - [[#❓ Questions]] - [[#🔮 Future Research]] - [[#💡 Implications]] - [[#🧠 Learning Styles Assessment]] - [[#📊 Accessibility Assessment]] - [[#🌍 DEIB Assessment]] - [[#✅ Logical Validity]] - [[#🛡️ DARVO Analysis]] - [[#🎯 Conclusion]] ### 🔗 References - [Stop Sloppypasta](https://stopsloppypasta.ai/en/) - [It's Rude to Show AI Output to People](https://example.com) - [Personal AI Ethics by Simon Willison](https://example.com) ## 📝 Summary - What happened: A communication etiquette guide defining "sloppypasta" as unread, unrefined LLM output pasted at people - Why it matters: LLMs create effort asymmetry where writing becomes free but reading requires verification - Key stakeholders: Knowledge workers using AI tools in professional communication - Risks and tradeoffs: Eroded trust, cognitive debt, and damaged professional relationships vs. productivity gains - Bottom line: Use AI to accelerate your work, not to replace thinking about what you send ## 📖 Main Content The guide introduces "sloppypasta" - verbatim LLM output copy-pasted without reading, refining, or being requested. This practice violates the traditional balance between writing effort and reading effort that governed professional communication. Three main patterns emerge: The Eager Beaver (contributing generic AI text to discussions), The OrAIcle (using ChatGPT responses to answer specific questions), and The Ghostwriter (presenting AI output as personal work without disclosure). The core problem is effort asymmetry. Previously, writing required effort that balanced the reader's effort. LLMs make writing effectively free while reading effort remains constant - and actually increases due to verification requirements. This creates a negative feedback loop where senders forfeit learning opportunities while recipients lose trust. The solution involves five guidelines: Read output before sharing, Verify facts before forwarding, Distill content to essential points, Disclose AI assistance, and Share only when requested. The guide emphasizes that new tools require new manners. ## 🗒️ Notes #### 📖 [Article] — [H1] Sloppypasta definition and core problem - Key idea: Sloppypasta is verbatim LLM output that's unread, unrefined, and unrequested - Evidence: "When someone forwards text they themselves have not considered, they are asking you to do work they chose not to do" #### 📖 [Article] — [H2] The Eager Beaver pattern - Key idea: Well-intentioned contributors share generic AI responses that derail ongoing discussions - Evidence: "It feels helpful to send. It creates work to receive" #### 📖 [Article] — [H2] The OrAIcle pattern - Key idea: People paste ChatGPT responses as answers to specific questions without context - Evidence: "'ChatGPT says' is the enshittified LLM-era equivalent of LMGTFY" #### 📖 [Article] — [H2] The Ghostwriter pattern - Key idea: Presenting AI output as personal work without disclosure borrows credibility dangerously - Evidence: "If the content turns out to be wrong, that credibility is what gets spent" #### 📖 [Article] — [H1] Effort asymmetry problem - Key idea: LLMs break the traditional balance between writing effort and reading effort - Evidence: "Now LLMs make writing 'free' and increase the effort to read due to additional verification burden" #### 📖 [Article] — [H3] Cognitive debt from delegation - Key idea: Shortcutting thinking with LLMs reduces comprehension and retention - Evidence: "Writing is thinking. The writing process forces the author to work through their thoughts" #### 📖 [Article] — [H1] Trust erosion in LLM era - Key idea: LLM hallucination propensity means "trust but verify" becomes "untrusted by default" - Evidence: "LLMs generate the most probable next token... which explains their propensity for hallucination" #### 📖 [Article] — [H2] Authority without expertise problem - Key idea: LLMs write with expert tone regardless of actual knowledge accuracy - Evidence: "LLMs write authoritatively with the tone and confidence of an expert" #### 📖 [Article] — [H1] Five guidelines for better practice - Key idea: Read, Verify, Distill, Disclose, Share only when requested - Evidence: "Use AI to accelerate your work or improve what you send. Don't use it to replace thinking" #### 📖 [Article] — [H3] Link sharing recommendation - Key idea: Share AI output as links rather than inline text to respect recipient choice - Evidence: "A link lets the recipient choose when - and whether - to engage" #### 📖 [Article] — [H2] Disclosure restores trust signals - Key idea: Transparency about AI assistance helps recipients understand what was verified - Evidence: "Disclosure restores the trust signals that sloppypasta destroys" #### 📖 [Article] — [H3] Productivity should not become others' burden - Key idea: AI acceleration benefits should not create work for recipients - Evidence: "Using AI should not make your productivity someone else's burden" ## 📚 References - [Stop Sloppypasta](https://stopsloppypasta.ai/en/) - [It's Rude to Show AI Output to People by Alex Martsinovich](https://example.com) - [Personal AI Ethics by Simon Willison](https://simonwillison.net) - [AI Writing Etiquette Manifesto by Blake Stockton](https://example.com) ## 🗣️ Paraphrase - LLMs have disrupted communication etiquette by making text generation effortless while reading and verification remain burdensome - Three common patterns of problematic AI sharing emerge: eager helpers flooding discussions, oracle-style question answering, and undisclosed ghostwriting - The solution requires reading, verifying, distilling, disclosing AI assistance, and sharing only when appropriate to restore balanced effort and trust ## 📰 Newsletter **New Communication Etiquette for the AI Era** As WordPress contributors increasingly use AI tools for documentation and community discussions, we need new guidelines to maintain productive collaboration. The concept of "sloppypasta" - unread AI output pasted into conversations - highlights a growing challenge. When someone drops ChatGPT's generic response into a GitHub issue or Slack discussion, they're asking others to read and verify content they haven't reviewed themselves. **Practical WordPress Applications:** - Review AI-generated documentation before submitting to DevHub - Verify plugin recommendations from LLMs against current versions - Disclose AI assistance in community forum responses - Link to AI output rather than pasting walls of text in Slack This mirrors our existing code review standards: just as we wouldn't commit unreviewed code, we shouldn't share unread AI content. Resources: [Contributor Handbook](https://make.wordpress.org/handbook/), [Documentation Style Guide](https://make.wordpress.org/docs/), [Community Guidelines](https://make.wordpress.org/community/) ## ✅ Action Items #### DevRel - Developer advocacy: Create conference talks on AI ethics in open source communication - Education: Develop workshops on responsible AI use for developer relations teams - Developer experience: Add AI usage guidelines to documentation style guides - Developer marketing: Audit marketing content for undisclosed AI assistance patterns - Community management: Establish moderation guidelines for AI-generated content in forums #### Community - Maintainers: Develop project-specific AI disclosure templates and review processes - Contributors: Create onboarding materials covering AI etiquette in open source - Sponsors: Establish transparency requirements for AI-assisted grant applications and reports - Ecosystem: Coordinate cross-project standards for AI content disclosure #### WordPress - Core: Integrate AI etiquette guidelines into contributor handbook and code review processes - Community: Add AI communication standards to meetup organizer and WordCamp speaker guidelines - Training: Develop lesson plans on responsible AI use for WordPress educators and trainers - Marketing: Audit campaign content for AI disclosure compliance and establish review workflows - Support: Create knowledge base articles on identifying and handling AI-generated support requests - Documentation: Establish DevHub standards requiring disclosure of AI assistance in tutorials and guides - Meta: Implement .org platform features supporting AI content labeling and verification ## 🧪 Hypotheses - Teams with explicit AI communication guidelines will show higher trust levels in internal surveys - Disclosed AI-assisted content receives more constructive feedback than undisclosed content - Organizations implementing verification requirements for AI output will reduce misinformation incidents - Distillation practices for AI content will improve both sender comprehension and recipient satisfaction ### 📊 Methodology - Survey design comparing trust metrics before and after guideline implementation - Content analysis measuring feedback quality on disclosed vs undisclosed AI assistance - Incident tracking across organizations with different AI verification policies ### ❓ Questions - How do we balance AI productivity gains with communication authenticity in open source projects? - What technical tools could help automate AI content disclosure and verification? - How should AI-generated content be handled in formal governance and decision-making processes? ### 🔮 Future Research - Long-term effects of AI communication patterns on team cohesion and project sustainability - Cross-cultural differences in AI communication etiquette expectations - Technical infrastructure for trustworthy AI content verification in collaborative platforms ### 💡 Implications - Practice: Requires establishing team guidelines, disclosure templates, and review processes for AI-assisted communication - Policy: May necessitate updates to professional communication standards and open source governance frameworks - Security: Highlights need for verification processes to prevent misinformation from LLM hallucinations - Education: Demands new training on AI ethics and communication etiquette for knowledge workers - Funding: Could impact resource allocation toward AI literacy and communication infrastructure ### 🧠 Learning Styles Assessment Reading level estimate: 12th grade (complex vocabulary, technical concepts) #### 👀 Visual learners - Present: Clear examples with formatting, emoji usage for pattern recognition - Improve: Flowcharts showing decision trees for AI content sharing, visual comparison tables #### 🔊 Aural learners - Present: Quotable phrases and memorable terminology like "sloppypasta" - Improve: Audio version with emphasis on tone differences between human and AI writing #### 🗣️ Verbal learners - Present: Strong definitions, analogies (Boy Who Cried Wolf), clear terminology - Improve: More workplace conversation examples, role-playing scenarios #### 🧩 Kinesthetic learners - Present: Actionable five-step guidelines, concrete examples with before/after - Improve: Interactive checklist, hands-on exercises for reviewing AI output #### 🧠 Logical learners - Present: Clear cause-and-effect relationships, systematic breakdown of problems - Improve: Decision matrices for when to use AI, quantified effort calculations #### 🧑‍🤝‍🧑 Social learners - Present: Workplace conversation examples, team dynamics focus - Improve: Group discussion prompts, collaborative guideline development exercises #### 🧘‍♀️ Solitary learners - Present: Self-assessment opportunities, reflection on personal practices - Improve: Personal AI usage audit worksheets, private journaling prompts #### 🌿 Naturalistic learners - Present: Ecosystem thinking about communication environments - Improve: Analogies to information ecosystems, sustainable communication practices ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Clear heading structure, descriptive link text, emoji with semantic meaning - Improve: Add aria-labels for complex formatted examples, ensure example structure is clear #### Keyboard navigation - Present: Standard web navigation patterns - Improve: Skip links for long examples, ensure interactive elements are keyboard accessible #### Contrast and readability - Present: Good text contrast, clear typography - Improve: Verify emoji accessibility, ensure code examples meet contrast requirements #### Text resizing and reflow - Present: Responsive design supporting zoom - Improve: Test complex examples at 200% zoom for readability #### Media and captions - Present: Text-based content with clear structure - Improve: Add captions if video examples are created, provide alt text for any diagrams #### Forms and interactions - Present: Simple website structure - Improve: If interactive elements added, ensure proper labeling and error handling ### 🌍 DEIB Assessment #### Cultural breadth - Present: Universal workplace communication challenges - Improve: Examples from diverse cultural contexts, acknowledgment of varying communication norms #### Inclusive language - Present: Gender-neutral examples, professional tone - Improve: Consider non-native English speakers, provide glossary for technical terms #### Neurodiversity support - Present: Clear structure, consistent formatting, chunked information - Improve: More frequent headings, executive summary for each section #### Representation - Present: Workplace-focused examples without identity assumptions - Improve: Diverse communication style examples, acknowledgment of different professional contexts #### Psychological safety - Present: Non-judgmental tone, practical guidance rather than criticism - Improve: Emphasize learning and improvement over blame, provide face-saving alternatives ### ✅ Logical Validity #### Strong reasoning - Deductive strength: Clear definition-to-example progression shows strong logical structure - Inductive strength: Multiple examples building to general principles about AI communication - Causal reasoning: Solid effort asymmetry argument connecting LLM capabilities to communication problems - Analogical strength: Effective use of "Boy Who Cried Wolf" and historical communication comparisons #### Fallacies - False dichotomy: Slightly overstates the binary between "good" and "bad" AI use without acknowledging nuanced middle ground - Appeal to authority: Heavy reliance on expert quotes without additional evidence, though quotes are well-chosen and relevant #### Tone of voice - Overall tone: Professional but accessible, authoritative without being preachy, uses humor appropriately ("enshittified LLM-era") ### 🛡️ DARVO Analysis - Deny: No denial patterns detected - acknowledges legitimate productivity benefits of AI tools - Attack: No attacks on AI users - focuses on practices rather than people, emphasizes good intentions - Reverse victim/offender: No reversal patterns - maintains clear focus on impact on recipients - Emotional cues and rhetoric: Uses measured language, avoids inflammatory terms, focuses on practical solutions ## 🎯 Conclusion The rise of LLMs has fundamentally disrupted professional communication by breaking the traditional balance between writing and reading effort. "Sloppypasta" represents a breakdown in communication etiquette that requires new social norms and practices. The practical next steps involve implementing the five-guideline framework: reading AI output before sharing, verifying facts, distilling content, disclosing assistance, and sharing only when requested. Teams and organizations need explicit policies around AI-assisted communication. This connects to broader questions about maintaining human authenticity and trust in an AI-augmented workplace, with implications for open source governance, professional development, and collaborative knowledge work. ## 🧱 Refactor Appendix ### 📚 Literature Split Pad ### [H1] LLMs create effort asymmetry in professional communication > [!quote] Stop Sloppypasta: Don't paste raw LLM output at people by [MISSING] via stopsloppypasta.ai. Published [MISSING]. Accessed 2024-12-19. > "Now LLMs make writing 'free' and increase the effort to read due to additional verification burden." Foundation for understanding why AI-assisted communication feels problematic and requires new etiquette frameworks. [view highlight](https://stopsloppypasta.ai/en/#H1) #xref #articleseed ID:rw_001 ### [H2] Writing delegation to AI reduces comprehension and learning > [!quote] Stop Sloppypasta: Don't paste raw LLM output at people > "Writing is thinking. The writing process forces the author to work through their thoughts." Core argument for why shortcuts in communication preparation harm both sender and recipient relationships. [view highlight]( <!-- deep-read:sections end --> 📄 **Full text:** ![[Full Document Contents/Articles/Stop Sloppypasta Don't paste raw LLM output at people|Stop Sloppypasta: Don't paste raw LLM output at people, stopsloppypasta.ai]] ## Highlights - A conversation participant wants to contribute to the topic at hand, so they ask a chatbot and share whatever comes back. The intention is good - they genuinely want to help - but the wall of generic AI text they contributed blocks the discussion already underway. Now other participants have to scroll past it to continue, or stop to read and validate it. ([View Highlight](https://read.readwise.io/read/01kky5f0tqbcfz7nszwpk2rqgn)) - Tags: 👻 ai highlighted