In the Kinematic Chinese Whispers Proof of Principle (PoP), a chain diffusion paradigm is used to examine how social information is transmitted from one agent to another across a sequential chain. The concept resembles the classic children’s game, except that the information is passed through movement kinematics.
The Chinese Whispers PoP applies an adapted fear-conditioning protocol to investigate the transmission of fear-related information associated with an object that is handed from one agent to the next.
Encoding
Fear encoding from real human movements (L0) was first quantified. In a fear-conditioning paradigm, participants were instructed to reach toward, grasp, lift, and move one of two cubes (e.g., a blue or a yellow cube). In 30% of the trials, one cube was paired with a naturally aversive electrodermal stimulation, causing participants to learn to fear it. Using a single-trial kinematic coding framework, the analysis identified which kinematic features carried information about fear. These analyses highlighted specific features—such as wrist velocity, wrist height, and horizontal trajectory—that reliably encoded social information at the level of individual movements.
Readout
Human readout of this encoded information was assessed by testing participants’ ability to discriminate fear from real videos (L0) and avatar retargetings (L1) of reach-to-grasp movements. L1 animations preserve the relevant kinematic properties within VR environments. Readout analyses revealed substantial individual differences: some observers successfully accessed and interpreted the encoded information, whereas others ignored or misinterpreted the relevant kinematic cues.
Amplification
The project’s cognitive architecture for amplification was then applied to generate L2 avatars. This architecture blends the original movement with reference sensorimotor primitives, selectively amplifying informative kinematic features while maintaining biological plausibility (see De Lellis, F., Coraggio, M., Foster, N. C., Villa, R., Becchio, C., & Di Bernardo, M. (2024). Data-Driven Architecture to Encode Information in the Kinematics of Robots and Artificial Avatars, IEEE Control Systems Letters, doi: 10.1109/LCSYS.2024.3416071). Example stimuli below show the same original movement across the three levels (L0 human video, L1 retargeted avatar, and L2 amplified avatar), illustrating the full transformation pipeline:
Proof-of-principle experiments demonstrate that amplification can enhance access to encoded fear information compared with L0 and L1, even when amplification is not perceptually detected (L1 vs. L2 discrimination at chance level).
Together, these steps establish the foundation for the kinematic Chinese Whispers paradigm, enabling tracking of how fear-related information propagates—and potentially transforms—as movements are passed from one agent to the next. After amplification, the subsequent phase is to examine how fear-related information propagates across a chain of agents. Within this paradigm, the output movement of one agent becomes the input for the next.
