Persistent context
Memory correctability
Memory correctability means an AI system lets a person inspect and change the picture it retains of them. That depends on legibility from the outset. A person cannot correct a memory they cannot see.
Also called editable AI memory, memory legibility
Systems holding personal context fail in a particular way. They may retain an inference from something said once, during a mood or situation that later changed. That inference then shapes later exchanges, leaving the person with a system that feels subtly and repeatedly off.
Correctability recognizes this failure and gives people a remedy. Someone can ask what the system believes it knows, receive an answer they can judge, and change it. The work goes beyond adding an edit button because much of a memory system’s record is not written in sentences a person would recognize.
There is a real design tension. A fully exposed memory can resemble a surveillance report, and reading it can feel unpleasant even when it contains no error. Inspection needs to feel ordinary rather than alarming.
Why it matters
Without correctability, a system’s mistaken inference can stay permanent and invisible. The person must manage a version of themselves they cannot reach, or leave the system entirely.
Where TAICU stands
TAICU treats legibility and correctability as prerequisites for retaining personal context, rather than controls added after the memory system is in use.
Common questions
- Why can’t people delete their AI’s memory?
- Deletion is available in many systems, yet it is blunt: the person loses the continuity that made the system useful. Correctability offers a finer remedy, fixing the wrong material while preserving the right material.
- How can someone tell whether an AI memory is legible?
- A person can ask what the system believes about them, receive an answer they can judge in their own terms rather than as an opaque score or embedding, and change the record.