Relational systems
Relational agent
A relational agent is an AI system that maintains an ongoing, recognizable relationship with a specific person across time: it remembers earlier exchanges, keeps a consistent manner, can take initiative, and is understood by the person as a continuing participant rather than a service that is summoned and dismissed.
Also called socially present agent, persistent agent
The distinguishing property of a relational agent is not intelligence but continuity. It is the same interlocutor tomorrow that it was today. That continuity is what allows shorthand to develop, what makes an unexplained reference land, and what makes an out-of-character response feel like a rupture rather than a bad answer.
Relational agents therefore need capabilities that assistants do not. They need durable context, so they are not re-introduced to the person every session. They need a stable personality, because inconsistency reads as unreliability. They need judgment about initiative, because an agent that can speak first can also intrude. And they need repair, because a relationship that cannot survive a misunderstanding is not much of a relationship.
They also inherit obligations. A system that is present in someone’s ordinary life has more capacity to influence them than one they visit deliberately, and the influence is harder to notice precisely because it is familiar.
Why it matters
Relational agents are being deployed at scale, into messaging apps and homes and daily routines, well ahead of any settled understanding of how they should behave when the relationship goes wrong. The design decisions are being made either way; the question is whether they are made with evidence.
Where TAICU stands
HUG is TAICU’s relational agent. Running one is how the lab gets access to the failure modes that only appear after a system has been part of someone’s life for a while.
Common questions
- What makes an agent "relational" rather than just persistent?
- Persistence is a storage property; relational is a social one. A relational agent uses what it retains to sustain a recognizable relationship: consistent manner, appropriate initiative, and the ability to repair a misunderstanding instead of resetting.
- Should a relational agent be able to message you first?
- Initiative is what makes an agent socially present rather than merely available, but it is also the capability most easily misused. The design question is not whether an agent may speak first but what it owes a person when it does.