New workshop paper: What Are We Measuring? Bonding, Trust, and Human–Robot Relationships

I’m pleased to share that I’ve had a workshop paper accepted at the Human State-Aware Robotics (H-STAR): From Multimodal Data to Human-Adaptive Behavior in HRI workshop at RO-MAN 2026 in Kitakyushu, Japan.

The paper, What Are We Measuring? Bonding, Trust, and the Evaluation of Human–Robot Relationships, takes aim at a habit that has quietly become standard in HRI: using trust, usually captured with a self-report questionnaire, as a stand-in for the quality of a human–robot relationship as a whole. I argue this is a category error. Trust is a cognitive appraisal of whether a robot will perform reliably: forward-looking, quick to update, and fragile, such that a single salient failure can undo it. Social bonding is something else entirely: an affiliative social connection that accumulates slowly through repeated interaction and shows up in largely involuntary, bodily signals. Trust and bonding can track, but they can also come apart.

Once we treat bond and trust as two independent axes rather than one, we get a simple map of relationship states, which I simply call the bond × trust space. We can then break this space down into four regions: avoidance (low bond, low trust), functional (low bond, high trust), dependence (high bond, low trust), and symbiosis (high bond, high trust). I discuss some of the practical and ethical considerations of thinking about bonding and trust in this way.

This is part of a series of larger papers I’m currently developing, so I’d love to talk to anyone thinking about similar problems e.g. measurement, attachment, long-term deployments, or the ethics of bonding with robots and other artificial systems. Whether or not you’ll be at RO-MAN, please do get in touch.