5.2 KiB
How Sharp Speech Stops Quietly
Sojorn does not fight hostility. It contains it.
The Problem
Most platforms suppress hostile content after it has already spread. They rely on:
- Post-hoc moderation (viral damage before removal)
- Suspensions (creating martyrs)
- Shadowbanning (creating conspiracies)
- Algorithmic dampening (opaque and manipulable)
This approach fails because it treats hostility as a moderation problem, not a design problem.
Sojorn's Solution: Structural Containment
Sharp speech stops at three gates, before it travels:
1. Tone Detection at Creation
When a user writes a post or comment, the content passes through tone analysis before being stored.
How it works:
- Pattern matching detects profanity, hostility, and negative absolutism
- Content is classified: Positive, Neutral, Mixed, Negative, Hostile
- Profane or hostile content is rejected immediately with a rewrite suggestion
What the user sees:
"This space works without profanity. Try rephrasing."
"Sharp speech does not travel here. Consider softening your words."
Result: The hostile post never enters the database. No removal, no notification to others, no drama.
2. Content Integrity Score (CIS)
Every post that passes tone detection receives a Content Integrity Score (0-1):
- Positive, friendly language → CIS 0.9
- Neutral, factual language → CIS 0.8
- Mixed sentiment → CIS 0.7
- Negative but non-hostile → CIS 0.5
How CIS affects reach:
- Posts with CIS < 0.7 have limited eligibility in the Sojorn feed
- Posts with CIS < 0.5 are excluded from Trending
- Low-CIS posts still appear in Personal feeds (people you follow)
Result: Sharp speech is published but does not amplify. The author can express it, but it doesn't spread.
3. Harmony Score (Author Trust)
Each user has a private Harmony Score (0-100) that adjusts based on behavior:
Negative signals (lower score):
- Multiple blocks received (pattern, not single incident)
- Reports from high-trust users
- High content rejection rate (repeated rewrite prompts)
- Filing false reports
Positive signals (raise score):
- Sustained positive participation
- Validated reports (helping moderation)
- Time without issues (natural recovery)
Harmony score determines:
- Posting rate limits (New: 3/day, Trusted: 10/day, Established: 25/day)
- Reach multiplier in feed algorithms (Restricted: 0.2x, Established: 1.4x)
- Trending eligibility (must be Trusted or above)
Result: Authors who persistently produce sharp speech have reduced reach. Their words don't disappear—they just don't travel far.
How These Gates Work Together
User writes post
↓
Tone Analysis
↓
├─ Hostile/Profane → REJECTED (rewrite prompt)
├─ Negative → Published with CIS 0.5 (limited reach)
├─ Mixed → Published with CIS 0.7 (moderate reach)
└─ Positive/Neutral → Published with CIS 0.8-0.9 (full reach)
↓
Post appears in Personal feeds (always)
↓
Post eligibility for Sojorn feed (based on CIS + Harmony)
↓
├─ High CIS + High Harmony → Wide reach
├─ Mixed CIS or Low Harmony → Limited reach
└─ Low CIS + Low Harmony → Minimal reach
↓
Post eligibility for Trending (based on CIS + Harmony + Safety)
↓
├─ CIS >= 0.8, Harmony >= 50, No blocks/reports → Eligible
└─ Otherwise → Excluded
What This Means for Users
If you write friendly, thoughtful content:
- Your posts pass tone detection instantly
- They receive high CIS scores
- They reach wide audiences
- They may trend
- Your Harmony score grows over time
If you write sharp, negative content occasionally:
- Some posts may be rejected with rewrite suggestions
- Published posts have reduced reach (CIS penalty)
- You still appear in Personal feeds
- Your Harmony score dips slightly but recovers naturally
If you persistently write hostile content:
- Most posts are rejected at creation
- Published posts have minimal reach
- Your posting rate is throttled
- Your Harmony score drops, further reducing reach
- You are never banned, but your influence diminishes
Why This Works
- No viral damage – Hostile content is stopped before it spreads
- No martyrdom – Authors are not suspended or removed
- No opacity – Users know why reach is limited (CIS, Harmony, tone)
- No gaming – You cannot brigade or spam your way to reach
- Natural fit – People who don't fit this space experience friction, not rejection
Comparison to Traditional Moderation
| Traditional Platforms | Sojorn |
|---|---|
| Content spreads first, removed later | Content stopped before it spreads |
| Bans create martyrs | Reduced reach creates natural exits |
| Shadowbanning is opaque | Reach limits are explained transparently |
| Algorithms amplify outrage | Algorithms deprioritize sharp speech |
| Moderation is reactive | Containment is structural |
The Result
Sharp speech does not travel here.
Not because it's censored. Not because it's hidden. But because the platform is architecturally designed to let it expire quietly.
Clean content flows. Sharp content stops. Friendliness is structural.