Picture two robots. One performs flawlessly, never hesitates, never gets anything wrong. The other stumbles sometimes. Misjudges a move. Admits it.
Which one do you trust more?
Most people assume the flawless one. The research says otherwise.
Vulnerability is not weakness. It is a trust signal.
We have known this about humans for a long time. The colleague who admits "I don't know" earns more credibility than the one who bluffs. The leader who says "I got that wrong" builds more loyalty than the one who never does. Vulnerability is uncomfortable to show and remarkably effective once shown, because it signals something a flawless performance never can: that you are being honest, not performing.
What is surprising is that the same effect shows up with robots.
Researchers studying group trust found that a robot exhibiting vulnerable behavior, compared with a neutral robot, produced more group engagement and stronger social signs of trust. People offered it more support when it struggled. They treated its setbacks the way they would treat a teammate's, not a machine's malfunction.
The robot did not become more competent by being vulnerable. It became more trusted.
The empathy study that proves the same point sideways
A separate study put this to the test in a real classroom. Two students worked with a robot tutor on a sustainability game. One version of the robot expressed empathy: it remembered past struggles, adjusted its tone to the group's mood, looked at whoever was speaking. The other version did the exact same tasks, gave the exact same amount of guidance, but skipped all of that. Same competence. Same help. Just less emotional presence.
The result: children working with the empathic robot had far more meaningful conversations about the subject. They argued, disagreed, took positions. The children with the non-empathic robot mostly just tracked their score.
The robot did not teach them more facts. It made them want to engage.
Why this works on us
Here is the part that matters for leaders. We are not wired to trust performance. We are wired to trust legibility — the sense that we understand what something is feeling, thinking, or struggling with.
A robot that reveals a little of its inner state, even artificially, gives us something to read. A robot that performs without revealing anything gives us nothing to hold onto. We fill that gap with suspicion, not admiration.
This is the opposite of how most organisations think about deploying robots. The instinct is to make them look as capable and seamless as possible. Hide the stumbles. Smooth the rough edges. Project pure competence.
The research suggests that instinct is backwards.
What this means for your team
If you are introducing a robot, or any new system, into your workplace, the most trust-building thing you can do is not to hide its limitations. It is to be transparent about them.
Let your team see where the robot struggles and how it adapts. Let it acknowledge uncertainty rather than perform false confidence. Build moments where the system, or the people responsible for it, openly say: this part is still being figured out.
That is not a weaker rollout. It is a stronger one.
We trust the things that let us see them clearly. Flawless and opaque loses to honest and imperfect, every time.
Based on "Empathic Robot for Group Learning: A Field Study" by Alves-Oliveira, Sequeira, Melo, Castellano, and Paiva (ACM Transactions on Human-Robot Interaction, 2019).