Context Drift

The person you were last year is not the person reading this.

Most of the big circumstances of your life have probably shifted at least a little since then, and some of them may have shifted enormously without you fully noticing. Your job may look different. Your health may. Your relationships, your living situation, your finances, your worries, the thing you care most about. Some of these changes were planned and some of them happened to you, but all of them quietly altered the person you are when you show up to any new situation. This seems obvious when you say it out loud, and yet almost every system that is supposed to help you, whether it is a healthcare profile, a learning platform, a work performance review, or a government form, tends to treat the version of you it met first as the version that still exists. It was accurate once. That does not mean it is accurate now.

This matters particularly in anything that adapts to you over time, because an adaptation based on outdated information is worse than no adaptation at all. It looks personalised, so you trust it, but it is personalised to someone who is no longer there. A well designed system should be watching for drift, noticing when what it thinks it knows about you has stopped matching who you are, and treating a stale profile as a warning rather than a fact. Most systems do not do this, because checking takes effort and assuming saves time. It is one of the quietest failures in personalisation, and it is everywhere. Is there a place in your life where you are still being treated as the person you used to be?

Last week we looked at what would happen if every AI shipped with an honest description. This week we look at how easily a profile becomes outdated and at how rarely systems notice.