Source-linked AI summary
What Are We Measuring? Bonding, Trust, and the Evaluation of Human-Robot Relationships
Imran Khan
TL;DR
The paper argues that trust-based measures do not adequately capture long-term human–robot relationship quality because trust and social bonding are distinct constructs. It proposes a two-dimensional bond × trust framework with four relationship states and draws consequences for estimation, adaptation, failure interpretation, and ethical oversight.
Problem
Long-term human–robot relationship quality is often assessed through trust measures, although trust does not capture affiliative bonds that can exist without functional reliance.
Method
The paper distinguishes trust from social bonding and models them as independent axes in a two-dimensional bond × trust space.
Results
The framework identifies four relationship configurations: avoidance, functional, dependence, and symbiosis.
Takeaways & Limitations
Social bonding should be estimated separately from trust, used to condition online adaptation, and considered when interpreting robot failures and detecting dysfunctional dependence.
Takeaways & Limitations
The proposed framework is ongoing work and raises unresolved questions about how systems should respond when they detect dysfunctional bonded dependence.
Abstract
from arXiv · showhide
In human-robot interaction, relationship quality is often quantified using self-report measures, particularly related to "trust", such that a robot's trustworthiness comes to serve as an index of how close or "bonded" a human feels to it. I argue that this is a category error: trust and social bonding are distinct constructs, differing in their antecedents, their timescales, their bodily signatures, the human experience they produce, the robot responses they call for, and the ethical concerns they raise. I propose that we view them as independent dimensions, and describe the resulting two-dimensional space of possible relationship states under this view, with four configurations: avoidance, functional, dependence, and symbiosis. I then draw out some consequences for human-state-aware robotics: (1) social bonding is an explicit estimation target distinct from trust, (2) it should condition online adaptation (3) it reframes what a "failure" means, and (4) it raises the possibility of identifying dysfunctional relationships, in which a user remains attached to a robot that no longer merits reliance. This is ongoing work, offered in part to prompt the field to reconsider what it means to evaluate relationship quality in human-robot dyads.
1 Introduction
Long-term HRI often treats self-reported trust as a summary of relationship quality, but short-term trust measures do not capture the dynamics of enduring human–robot relationships.
- Trust is commonly used as an index of relationship quality and to drive adaptive robot behaviour.
- This approach is reasonable for short, task-oriented interactions but falls short for long-term robot deployments.
- Trust measurements do not necessarily reveal what happens between social agents over time.
2 Distinguishing Bond from Trust
Trust and social bonding are distinct relationship constructs: trust appraises likely performance, whereas bonding develops through repeated interaction and is expressed through attachment-related bodily and behavioural responses.
- Trust is a forward-looking, rapidly updating cognitive appraisal of whether a robot will perform reliably.A single salient failure can substantially reduce trust.
- Social bonding develops gradually through proximity maintenance, safe-haven and secure-base use, and distress on separation.
- Bonding is associated with extended physiological changes, including oxytocinergic signalling, reduced amygdala reactivity, and lower cortisol.
- Bonding is expressed through lower arousal, reduced interpersonal distance, and behavioural synchrony, whereas trust reflects what the robot does.
- Trust governs momentary reliance, while an established bond stabilises relationships and can absorb the effect of failures.Bonding physiology can attenuate threat reactivity and cortisol responses to errors.
- A person may trust a competent machine without attachment or remain attached to an unreliable partner, showing that the constructs need not covary.
3 Trust as a (Failed) Proxy for Relationship Quality
HRI commonly substitutes trust measures for social bonding, leaving affiliative relationships and cases where bonding and trust diverge inadequately represented.
- Self-report questionnaires, especially trust instruments, remain a dominant way to assess relationship quality in HRI.
- Studies of attachment or companionship often use trust, competence, or positive affect as proxies for attachment.
- Treating reduced interpersonal distance as part of trust prevents detection of close proximity alongside declining trust.
- Trust instruments provide little to measure when a robot has no task whose reliability can be assessed.
- Bonding proxies such as social buffering, narrowed interpersonal distance, and behavioural synchrony have been proposed, but no longitudinal study had paired them with validated trust measurement.
4 A Two-Axis Model of Relationship State
Treating bond and trust as independent dimensions yields a two-dimensional relationship space with four configurations: avoidance, functional, dependence, and symbiosis.
- Model: The bond × trust space represents each dyad as a point defined by separate bond and trust dimensions.
- Avoidance: Avoidance combines low bond and low trust, representing no formed relationship or a return after early failures.
- Functional: Functional combines low bond and high trust: the robot is trusted as competent without a meaningful affiliative relationship.This configuration suits short task-focused interactions but can deteriorate after serious or repeated failures.
- Dependence: Dependence combines high bond and low trust, including relationships where attachment persists after the robot’s usefulness declines.This configuration can become harmful when vulnerable users continue relying on unreliable robot guidance or assurances.
- Symbiosis: Symbiosis combines high bond and high trust and is described as more robust than either dimension alone.An established bond can buffer occasional failures, although no study had deliberately targeted this state while tracking both dimensions.
5 How Does a Bond x Trust Framework Affect Human-State-Aware Robotics?
Separating bonding from trust changes human-state-aware robotics from single-axis adaptation to relationally informed estimation, adaptation, failure handling, and ethics. The framework makes bonded state explicit and highlights dependence on an unreliable robot as a distinct ethical concern.
- Bonded state as an estimation target: Bonded state should be estimated separately from trust using multimodal, largely involuntary signals such as arousal, interpersonal distance, and synchrony.The target includes whether the human cares about the robot, not only whether they trust its performance.
- Two-axis adaptation: Robot adaptation should use both bond and trust axes because the four relational regions require different responses.In the functional region, robots should facilitate bond formation; in the dependence region, they should make failures and uncertainty legible or escalate to a carer.
- Relational failure: A robot’s failure should be interpreted relative to relational state rather than treated as a fixed decrement requiring one repair policy.The same error may be near-terminal without a bond but survivable within an established bonded relationship.
- Ethical constraint: Dependence is both a measurement blind spot and an ethical problem when users remain attached to robots whose performance has declined.This risk is especially concerning for vulnerable users, whose continued dependence may be unnoticed or not consented to.
6 Conclusion
Trust alone does not adequately capture relationship quality in human–robot interaction. Treating trust and social bonding as separate dimensions yields distinct estimation, adaptation, failure, and ethical implications, motivating new measures of human–robot relationships.
- Trust and social bonding are distinct constructs with different antecedents, timescales, behavioural signatures, and functional roles.
- Human-state-aware robotics should estimate bonding separately from trust and condition online adaptation on the relational state.
- Failure interpretation depends on the relational state in which an error occurs.
- Separating bonding from trust exposes ethical questions about dysfunctional bonded states that may otherwise remain latent.