Interaction Half-Life
> **The transaction fades. The relationship endures.**
Interaction Half-Life
The transaction fades. The relationship endures.
Every time you exchange help with someone on Karmyq, you strengthen a trust edge between you. But trust isn't a static number — it has a half-life. Connections that go quiet gradually fade. Connections that stay active grow stronger and last longer.
What Is Interaction Half-Life?
Interaction half-life is the time it takes for a trust connection to decay to half its current strength, assuming no new interactions.
By default, a single exchange has a half-life of 30 days. After 30 days of silence, the connection is half as strong. After 90 days, it's below the visibility threshold and will eventually be removed.
But here's the key: each new exchange extends the half-life of the connection. Repeat interactions don't just add weight — they add stability. A relationship with five exchanges has a half-life of over 60 days. One with twenty exchanges has a half-life measured in years.
How Stability Works
Each trust connection has a hidden stability value that starts at 1.0. Every time you interact with someone:
- The interaction is counted and contributes to the raw trust weight
- The stability score grows:
stability = stability × 1.20(20% per interaction, by default) - A higher stability means a longer effective half-life:
effective_half_life = stability × base_half_life
| Interactions | Stability | Effective half-life |
|---|---|---|
| 1 | 1.0 | 30 days |
| 5 | ~2.1 | ~62 days |
| 10 | ~5.2 | ~155 days |
| 20 | ~38 | ~957 days |
A relationship built on twenty exchanges is effectively permanent — it would take nearly a decade of silence to fade away.
What the Fading UI Means
In the Trust Graph
Edges in your trust graph fade in opacity as they decay. A bright, fully opaque edge means a recently active relationship. A faint edge means the connection exists but hasn't been reinforced recently.
The opacity formula: opacity = 0.2 + (current_weight / raw_weight) × 0.8
- A fresh connection (just interacted): full opacity (1.0)
- A connection that's halfway to its half-life: ~0.6 opacity
- A nearly-faded connection: 0.2 opacity (minimum visible)
Reading edge opacity: If you see many faint edges in your trust graph, it means those relationships are going quiet. Completing another exchange will instantly re-brighten the edge.
In the Request Feed
Completed requests that are approaching their 30-day deletion window will appear increasingly faded in the feed. This is a visual signal that the transaction history is being archived — but the trust relationship it created lives on.
What Happens at 30 Days (Request Deletion)
Completed, fully-rated requests are permanently deleted 30 days after completion. This is intentional:
- The transaction is ephemeral. Once rated, it has served its purpose.
- The relationship is durable. The trust edge it created — with its accumulated stability and raw weight — persists and continues to decay on its own biological clock.
What is preserved: the trust edge's raw_weight and stability. What is deleted: the original request text, descriptions, and match records.
This keeps the platform lightweight and focused on active mutual aid, not historical record-keeping.
How Your Community Configures It
Community admins can tune the decay parameters for their community via the admin panel or the decay config API:
- Base half-life (default: 30 days): How quickly a single interaction fades
- Stability growth rate (default: 20%): How much each interaction extends the half-life
- Disappearance threshold (default: 0.5): The weight below which a connection is swept away
Communities with faster social cycles (high-activity neighborhood groups) may prefer shorter half-lives. Communities with slower rhythms (regional networks, professional associations) may prefer longer ones.
Summary
- Trust connections are alive — they decay when dormant, strengthen when active
- Each new interaction extends the connection's half-life (stability effect)
- Fading edges in the trust graph are a signal, not a problem
- Completed requests are deleted after 30 days; the relationship they created is not
- The platform cleans up dead edges automatically via a nightly sweep job