When robots own their mistakes: Study reveals error acknowledgment builds human trust

As robotic systems move into workplaces and homes, a critical question emerges: how do people decide whether to trust—or distrust—machines that fail? A new study from arXiv reveals that the answer hinges less on perfection than on honesty.

The trust experiment#

Researchers conducted a four-phase survey designed to map how trust forms and shifts during human-robot interactions. The work began by gathering demographic profiles and measuring participants' baseline confidence in robots. Early expectations about what robots could do came next, followed by a direct encounter stage that recorded participants' reactions to robot performance, including accuracy and how robots handled mistakes. Finally, participants reflected on whether they'd trust robots again, forgive their errors, or recommend them to others.

Honesty changes everything#

The standout finding: when robots acknowledged their errors or gaps in capability, human trust increased measurably. This suggests that transparency about failure may paradoxically strengthen confidence. People also reported higher willingness to suggest robots for future tasks after direct experience, indicating that personal interaction shapes attitudes more powerfully than assumptions.

What this means for robot design#

The results point toward a design philosophy that prioritizes responsiveness and candor. Rather than striving for an illusion of flawless operation, engineers might better serve users—and boost adoption—by building systems that communicate candidly about what they can and cannot do. Such an approach mirrors human collaboration: we trust colleagues not because they never stumble, but because they acknowledge when they do.