Do Robots Need Body Language? Comparing Communication Modalities for Legible Motion Intent in Human-Shared Spaces
Expressive motion cues from Boston Dynamics Spot quadruped proved more effective than lights or text.
A team from MIT's Media Lab and CSAIL, led by Jonathan Albert Cohen, published a research paper titled 'Do Robots Need Body Language?' on arXiv. The study directly compared four communication modalities—expressive motion, LED lights, on-screen text, and audio cues—to determine how best for a Boston Dynamics Spot quadruped robot to signal its navigation intent to humans in shared environments like hallways or doorways. The core finding was that implicit, expressive 'body language' motions were not only more accurately interpreted by people but also fostered greater confidence and trust than explicit signals like text or lights.
The research measured three key metrics: prediction accuracy, user confidence, and trust in the robot's safety. Across four common scenarios, expressive motion cues led to a 40% improvement in humans correctly predicting the robot's next action compared to baseline conditions. The study also explored the effects of aligned multimodal cues (e.g., motion and lights pointing the same way) versus conflicting signals, providing initial evidence that natural, motion-based communication may reduce the cognitive burden on humans who currently must adapt to opaque robot behavior. This work shifts the design paradigm from making robots merely functional to making them legible and predictable partners.
- Expressive motion cues improved human prediction accuracy of robot actions by 40% over other signals.
- The study tested a Boston Dynamics Spot robot using four modalities: motion, lights, text, and audio.
- Findings provide concrete evidence for using implicit 'body language' over explicit displays to build trust in shared spaces.
Why It Matters
As robots enter workplaces and homes, this research provides a blueprint for designing intuitive, trustworthy machines that communicate like collaborative partners, not unpredictable tools.