What is it about?
Creative robots need to do unexpected things to be interesting, but the same surprise can look like a malfunction. We call this an “ambiguous failure”: a moment that can reasonably be understood as either a creative choice or an error. In an online study, 149 people watched short videos of Shimon, a jazz-playing robotic marimba player. During pauses, Shimon either stayed completely still or made subtle, breathing-like movements. After unusual or risky notes, it either repeated and resolved the musical idea or continued without doing so. Viewers disagreed more about these moments than about baseline performances. Breathing-like movement during silence increased ratings of how appropriate or correct the robot seemed by about 10.5 points on a 100-point scale. The musical follow-up tested in these short clips did not reliably change immediate judgments.
Featured Image
Photo by Jakub Dziubak on Unsplash
Why is it important?
Robots are often evaluated by whether they behave as expected. In creative settings, however, breaking expectations is part of the job. Our findings show that audiences use a robot’s visible behaviour, not just its sound, to decide whether it is still participating or has stopped working. Small “still active” movements may help a pause seem like listening or planning rather than a freeze. Designers of creative and public-facing robots may therefore need to communicate intent through several channels, including movement, sound and simple explanatory cues, rather than treating every surprise as a fault to eliminate. This lesson is context-specific: in safety-critical settings, genuine malfunctions must remain clearly visible.
Perspectives
We wanted to explore a familiar puzzle from jazz: a pause or a “wrong” note can become meaningful depending on what happens around it. When a robot does the same thing, people may assume it has frozen or failed. Many famous jazz musicians describe this, such as the (perhaps wrongly attributed) Miles Davis quote "It's not the note you play that's the wrong note – it's the note you play afterwards that makes it right or wrong." Our results suggest that visible movement can make a robot easier to understand as an active performer, while subtle musical logic may be harder for a general audience to notice in a short clip. For us, the broader lesson is that creative robots need to show ongoing intent as well as produce interesting material.
Richard Savery
University of Canberra
Read the Original
This page is a summary of: Ambiguous Failures in Robotic Musicianship: The Impact of Embodied Liveness and Musical Context, ACM Transactions on Human-Robot Interaction, July 2026, ACM (Association for Computing Machinery),
DOI: 10.1145/3834778.
You can read the full text:
Contributors
The following have contributed to this page
