Persistent context
Memory decay
Memory decay is an AI system’s deliberate weakening or expiry of retained personal context, so older and less consequential material has less influence over time. It is designed forgetting that keeps a system current with who someone is now rather than who they once were.
Also called forgetting in AI systems, graceful forgetting
Total recall can sound appealing and act like a defect. People change their minds, recover from bad periods, abandon preferences, and say things they do not mean. A system giving a two-year-old remark the same weight as this morning’s knows an outdated composite and mistakes it for the present.
Human memory fades, and that fading helps relationships stay current. Old frictions lose force. People get room to become different from who they were. A system that retains everything takes that room away.
The design work lies in deciding what fades and how. Age alone is crude. A fact said once may deserve indefinite retention, while one repeated weekly may not. Decay guided by whether context remains useful is harder to build and closer to the aim.
Why it matters
Without decay, a persistent system can become an archive of someone’s worst days, with no room for them to outgrow their record.
Where TAICU stands
TAICU sees forgetting as designed behavior with requirements of its own, rather than a storage limit to remove. What a system should decline to retain deserves the same serious attention as what it recalls.
Common questions
- Should an AI system intentionally forget something?
- Yes. Designed forgetting keeps a system current with someone’s present life, prevents an old remark from shaping years of later exchanges, and limits what accumulates in a record the person never agreed to.
- How do deletion and memory decay work differently?
- Deletion is a person-requested event that removes a record. Decay is continuous: a record’s influence fades by default, pulling the system toward the present instead of the archive.