Violating etiquette expectations erodes bystander trust in a robot's later behavior

What it is

Once a robot violates an etiquette expectation, the cost is not limited to the disapproval that single event draws — it also erodes bystanders' trust in everything the robot does afterward. This is a cross-event trust cost: even someone not directly affected by the violation will still recalibrate their overall judgment of that robot simply from having witnessed it once.

Why it happens

The cross-event effect traces back to anchoring in impression formation. Bystanders build a mental model of an unfamiliar robot from a small number of observations, and an early or salient violation gets weighted disproportionately, hardening into a general impression that "this machine doesn't follow the rules, it's unreliable." That impression is then used to predict and reinterpret everything the robot does next — even an ordinary, compliant action can get read more negatively once the impression is set. This is the same mechanism behind primacy effects and the heavier weight negative information carries in impression formation generally, but robots start from a worse position: the baseline trust bystanders extend to a robot is already thinner than what they extend to a person, so a single violation punches through it more easily, whereas two people typically have a longer prior track record that can absorb one mistake. The effect can be amplified further when the violation is recorded and shared: if video of one incident circulates widely, people who only ever saw the footage — never the robot in person — form a negative impression of every unit of that model, turning a cost that would otherwise sit with the bystanders present into pre-formed distrust of the entire product line.

Studying it

Verifying the cross-event effect requires a longitudinal or multi-stage design, not a single-interaction measurement: have participants observe one violating or compliant episode, then after a delay or several further observations, measure trust in the robot's subsequent behavior and the expected likelihood of another violation, comparing against a control group that only ever saw the one interaction — the question is whether the cost persists rather than being a one-off emotional spike. Trust-repair manipulations — an apology or explanation offered after the violation — can also be introduced to test how much of the trust is recoverable and whether recovery happens as fast as the initial loss; the two rates are often asymmetric.

Where it stops holding

How long and how severe the trust cost is depends on whether the robot offers a visible correction or explanation for the violation: an explained violation produces a noticeably smaller cost than a silent one. A one-off minor lapse and a recurring, systemic violation do not produce the same cost and should not be treated as equivalent — bystanders' impression revision toward the latter is more thorough and harder to reverse. Attribution also redirects where the cost lands: if bystanders trace the violation to poor operator management rather than to the robot's own behavior pattern, the erosion may fall more on trust in the brand or deployer than on expectations about this particular robot's behavior. Whether the cost lands on this specific unit or spreads across the whole deployed fleet of that model depends on that same attribution: a violation read as a shared design flaw carries the cost forward to every unit of that model a user later encounters, not just the one involved in the original incident.

Applying it

Once deployment monitoring detects a violation of etiquette expectations — whether from a design flaw or an environmental accident — there should be an immediate bystander-facing explanation or apology mechanism to actively repair trust, rather than leaving bystanders to carry the negative impression into every later observation and let it compound on its own. To verify, run a bystander trust survey in a setting with a known past violation, comparing conditions with and without the repair mechanism, and confirm recovery through later actual interaction behavior — willingness to approach the robot again, compliance with its requests — rather than relying on a self-report attitude score alone.

Related

• Same group: X6.06.1 Robots in public space must follow human-like etiquette · X6.06.2 Interrupting or cutting in line provokes disapproval much like it would from a person · X6.06.3 Etiquette expectations vary by culture and setting, with no universal script
• Nearby: X5.02 Navigating shared space · X2.01 Signaling intent before acting
• Search terms: trust repair · first impressions · human-robot trust · negativity bias