How Movement Kinematics Reveal Human Intentions

50 concepts28 papers1 proposal

When an agent executes a physical movement, traditional artificial intelligence and classical planning models often treat it as a discrete, context-free command—such as moving a robotic arm to a coordinate or picking up an object. As David J. Israel et al. formalized in Actions and Movements, a massive conceptual gap exists between raw effector capabilities and the actual achievement of environmental goals. Historically, theories of social perception assumed that physical movements are too generic to signal intent on their own, forcing observers to rely heavily on external situational context to guess why an action is occurring (Atesh Koul et al., Decoding intentions from movement kinematics).

However, a growing body of research demonstrates that human intentions are not completely hidden mental states. Instead, they are physically encoded directly within the subtle geometry, velocity, and timing of the movements themselves—a framework known as kinematic intention encoding. The core intuition is simple: the "how" of an action inherently reveals its "why."

Why It Works: Anticipatory Motor Chaining

Why does an intention leave a physical trace before a goal is reached? The answer lies in how the brain plans complex, multi-step actions. The motor system does not execute actions as isolated, reactive commands; instead, it relies on anticipatory motor planning, where individual movements are "chained" together.

When you reach for an object, your brain is already planning what you will do after you grab it. If your intent is to pour liquid from a bottle, your hand approach, wrist rotation, and maximum finger aperture will adapt to accommodate that future task. If your intent is to pass it to someone else, the kinematics will look entirely different. This phenomenon, termed distal goal kinematic modulation, means that the ultimate end-state of an action sequence systematically shapes the velocity and timing of its very first phases (Caterina Ansuini et al., The visible face of intention: why kinematics matters).

Remarkably, this anticipatory chaining is a fundamental, hardwired feature of human motor control. Caterina Ansuini et al. demonstrated in Anticipatory action planning in blind and sighted individuals that early-blind, late-blind, and sighted individuals display the exact same kinematic adjustments when preparing for subsequent tasks. This proves that anticipatory motor planning operates independently of visual experience or real-time visual feedback, relying instead on a robust, multimodal cortical network. Because these physical adjustments are systematic, the geometry of the movement alone provides sufficient information for an observer to decode an actor's goals without needing broader environmental context (Cristina Becchio et al., Grasping intentions: from thought experiments to empirical evidence).

The Social Amplifier and Readability

Kinematic signatures become significantly more pronounced when actions shift from solitary tasks to social interactions. When an individual coordinates an action with another person—such as passing an object rather than placing it down—the integration of social intent into motor planning amplifies the distinctiveness of the kinematic profile (Andrea Cavallo et al., Identifying the signature of prospective motor control in children with autism). This social-motor modulation is a core developmental milestone; Lisa Ouss et al. showed that by five to six months of age, typically developing infants begin to automatically differentiate their hand kinematics based on whether their play is person-directed or object-directed (Developmental Trajectories of Hand Movements in Typical Infants and Those at Risk of Developmental Disorders: An Observational Study of Kinematics during the First Year of Life).

These social kinematic profiles are rich enough to differentiate between abstract social intentions, such as cooperation versus competition, purely through velocity and trajectory dynamics (Anna Sedda et al., Automatic Human Interaction Understanding: Lessons from a Multidisciplinary Approach). To isolate whether these alignments stem from basic motor drives or high-level social goals, researchers have utilized virtual agent tasks to decouple behavioral synchrony from explicit cooperative instructions (Julia Ayache et al., The « jingle-jangle fallacy » of empathy: Delineating affective, cognitive and motor components of empathy from behavioral synchrony using a virtual agent).

Crucially, the "readability" or legibility of these actions is highly variable:

Computational Decoding and Robotic Applications

How do brains and machines exploit these physical tells? In the human brain, Daniel M. Wolpert et al. proposed a computational model where observers decode intentions by running observed actions through their own internal visuomotor feedback controllers (Mental state inference using visual control parameters). By simulating the movement using our own neural circuits, we map visual kinematics directly to intended targets. This real-time sensorimotor integration is managed by a dedicated fronto-parietal network that extracts social meaning from movement stability and timing (Guillaume Dumas et al., The Human Dynamic Clamp reveals the fronto-parietal network linking real-time social coordination and cognition).

In computer vision and robotics, moving away from treating humans as simple coordinate points has unlocked massive performance gains. Saeed Saadatnejad et al. demonstrated that incorporating full body pose kinematics (like torso rotation or weight shifts) allows trajectory prediction models to anticipate turns and stops long before a person's physical position changes (Social-Pose: Enhancing Trajectory Prediction with Human Body Pose). Similarly, low-level physics-based trajectory properties, such as self-propelled direction changes and repetitive motions, serve as highly effective proxies for biological agency and intent across diverse viewpoints (Ye Luo et al., Actionness-Assisted Recognition of Actions).

These insights enable more intuitive human-robot collaboration across three main domains:

[Observed Kinematics] 
       │
       ├─► Speed & Directness ──► Path Dominance Inference (Bera et al.)
       ├─► Trajectory + Context ─► Proactive Intention Blending (Veselič et al.)
       └─► Careful Velocity ────► Motor Contagion / Shared Safety (Lastrico et al.)
  1. Social Navigation: Aniket Bera et al. showed that a robot can infer a pedestrian's social dominance and likelihood to yield simply by analyzing their speed and path directness (Pedestrian Dominance Modeling for Socially-Aware Robot Navigation).
  2. Proactive Assistance: In physical human-robot interaction, Sebastijan Veselič et al. proved that combining simple trajectory kinematics with environmental context allows a robot to proactively assist users with superhuman accuracy, even when the user changes their mind mid-movement (Human-Robot Interaction With Robust Prediction of Movement Intention Surpasses Manual Control).
  3. Implicit Communication: Communication via kinematics is a two-way street. When a robot arm moves with a "careful" velocity profile, human partners spontaneously mirror that behavior through motor contagion, creating a safer, more intuitive shared workspace (Linda Lastrico et al., If You Are Careful, So Am I! How Robot Communicative Motions Can Influence Human Approach in a Joint Task).

Key Foundational Work

Mental state inference using visual control parameters (2004) — contributed 1 concept elements to this direction.
Actions and Movements (1991) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_kinematic_intention_encoding_35b9b69d,
  title = {kinematic intention encoding},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/kinematic-intention-encoding/rdr_35b9b69d359a4d7b81268478b720b588},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 1, frozen text):

@misc{lacuna_kinematic_intention_encoding_35b9b69d_v1,
  title = {kinematic intention encoding},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/kinematic-intention-encoding/rdr_35b9b69d359a4d7b81268478b720b588/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake