Show HN: Respectify – A comment moderator that teaches people to argue better

Overall concept and intended use

• Tool analyzes comments before posting, flags issues (toxicity, dogwhistles, spam, off-topic, fallacies) and suggests rewrites.
• Developers frame it as a “nudge and educate” system, especially for bloggers who want comments but fear toxicity.
• They emphasize configurability: site owners can tune thresholds (e.g., dogwhistles, sexual content) rather than enforcing one global standard.

False positives and dogwhistle overreach

• Many users report absurd flags: “Christmas party” as a Christian dogwhistle, “Of course it is!” as off-topic, “horrible people” as inherently wrong.
• Dogwhistle detection is widely seen as oversensitive and context-blind; it initially mislabels benign statements, especially around religion and race.
• Developers repeatedly acknowledge this and say they “dialed it way down” during the thread.

Perceived political bias and echo-chamber risk

• Multiple tests on UBI, Trump, Obama, and transgender-rights topics suggest stricter treatment of certain viewpoints.
• Example: “Obama sucks” is flagged as racist dogwhistling; “Trump sucks” is not. Some pro-Trump or anti-UBI comments are hard to get approved even when civil.
• Critics argue this bakes in ideological bias, launders particular politics as “respect,” and risks turning communities into echo chambers.

Quality of rewrites and “LLM-speak”

• Suggested revisions are often described as mushy, over-equivo­cating, or meaningless, and sometimes alter the original meaning.
• Users worry about timelines being filled with samey, sanitized corporate/LLM tone, encouraging self-censorship and “algo-speak.”

Limits against bad-faith actors

• Several commenters argue the premise is flawed: real bad-faith actors are often eloquent, strategic, and will simply adapt or use the tool to better mask harassment or propaganda.
• Some see it as enabling sealioning or “laundering” bigotry into polite form.

Alternative ideas and use cases

• Suggestions:
  • Use it as a pre-post self-check or browser plugin rather than gatekeeper.
  • Focus more on logic, evidence, structure, and fallacies than on sentiment.
  • Rank or hide low‑quality/angry content instead of blocking it.
  • Create “discussion arenas” for vetted good‑faith participants.
  • Personal blocklists and user-side filters are proposed as a more agency-preserving alternative.

Philosophical and practical objections

• Many see it as paternalistic, dystopian, or a step toward algorithmic speech control / “social credit.”
• Concerns about normalization of AI moderation, chilling honest speech, and creeping censorship.
• Operational issues noted: slow site, timeouts, unstable outputs, privacy policy gaps, and potential for abuse (e.g., fine-tuning better spam).