Loading patterns…
Memory Decay & Forgetting Policies(MDF)
Explicit governance of what an agent drops rather than what it keeps: time-based expiry (TTL), usage and relevance decay (Ebbinghaus-style exponential decay that is boosted each time an item is accessed), staleness detection for facts that are still relevant but no longer true, and hard deletion for right-to-be-forgotten and GDPR Article 17 compliance. Unbounded memory measurably degrades agent performance, so forgetting is a first-class design axis. Distinct from `memory-consolidation` (which compresses and abstracts memories into schemas, the opposite operation) and from `generative-agents-memory` (which uses recency decay only as a retrieval score, not as an eviction or deletion policy).
In 30 seconds
- What
- Removes or deprioritizes memories by time-based expiry, access frequency decay, staleness detection, and budget-driven eviction rather than keeping everything.
- When to use
- Long-running agents accumulating thousands of facts where unbounded memory slows retrieval, or compliance requires hard deletion of user data.
- Watch out
- Aggressive decay can discard facts still needed for context; recovery after accidental deletion is impossible if audit trails are incomplete.
Ask the AI expert about this pattern
Opens the assistant with your question prefilled. You review it before sending.
Memory Decay & Forgetting Policies: Overview
Explicit governance of what an agent drops rather than what it keeps: time-based expiry (TTL), usage and relevance decay (Ebbinghaus-style exponential decay that is boosted each time an item is accessed), staleness detection for facts that are still relevant but no longer true, and hard deletion for right-to-be-forgotten and GDPR Article 17 compliance. Unbounded memory measurably degrades agent performance, so forgetting is a first-class design axis. Distinct from `memory-consolidation` (which compresses and abstracts memories into schemas, the opposite operation) and from `generative-agents-memory` (which uses recency decay only as a retrieval score, not as an eviction or deletion policy).
- TTL expiry for time-bounded facts
- Ebbinghaus-style decay scores boosted by access frequency
- Staleness detection for high-relevance but outdated facts
- Budget-driven eviction and archival when memory exceeds a cap
- Hard deletion with an audit record for right-to-be-forgotten requests
- Separation of decay (soft, reversible) from deletion (hard, permanent)
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.
References
The papers, specifications, and repositories this pattern is based on.
From the engineer behind this catalog
Get your agent architecture reviewed
This page documents one pattern. Your system runs dozens, and most failures live in how they fit together. Have the whole design reviewed against the 288 patterns in this catalog: architecture, reliability, evaluation and cost, every finding mapped to the pattern that fixes it.
€750 instead of €1,500, one week, written report and walkthrough call, until 30 September