Rationale: Whether working together to move a table or supporting a child learning to ride a bike, physically connected individuals use the exchange of haptic information to improve motor performance. However, this improvement occurs at the cost of additional effort for the more skilled partner. Objective: Here, we aim to assess whether an asym-metric connection, consisting of a stiffer link to the less skilled partner, could increase performance without additional effort in collaborative tasks. Methods: Through computational modelling, we first evaluated such a hypothesis on simulated human dyads tracking a common target. The approach was then experimentally validated on a three degree-of-freedom tracking task using two commercial robots as individual interfaces. Results: The simulation and experimental results confirm that using an asymmetric connection stiffness can improve joint performance without requiring additional effort from either partner compared to solo performance. Conclusion: This suggests that the training of motor skills with a proficient partner-like a physical therapist assisting a patient or a violin teacher demonstrating bowing techniques-may be enhanced through the use of robot-mediated asymmetric haptic communication. Index Terms-Human-robot interaction, joint action, human motor control Impact Statement-Using an asymmetric connection in a dyad enables the partners to improve their common performance without additional effort.