Imagine, Interact: Eliciting Accessible Interactions from Users with Motor Impairments via Imagined Input Devices
Radu-Daniel Vatavu, Ovidiu-Ciprian Ungurean · 2026 · Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI '26) · doi:10.1145/3772318.3790437
Summary
This paper reports an end-user gesture elicitation study with eleven participants with upper-body motor impairments - including spinal cord injury, spina bifida, multiple sclerosis, cerebral palsy, Parkinson's, and traumatic brain injury - who were asked to imagine input devices (with no physical form) and the gestures needed to operate them for twenty common smart-home and content tasks. The study sits at the intersection of two research threads that have historically run in parallel: accessible gesture-based input for people with motor impairments, and "imaginary" or body-referenced interfaces that exist only in users' minds. The authors argue that physical input devices impose constraints - mass, shape, grip requirements, button force, temporal availability - that are often inaccessible, and that imagined devices offer a way to bypass these constraints so that interaction form and modality emerge from the user's own body, abilities, and mental models rather than from hardware. Using Villarreal-Narvaez et al.'s formalization of the elicitation method, participants proposed a device and gesture for each of twenty referents grouped into Devices, Content, Actions, and Navigation categories. Sessions were video-recorded and coded across six measures: Device-Concept (embodied, body-supported, wheelchair-supported, mid-air), Device-Archetype, Supporter body part, Effector body part, Handedness, and Gesture-Type (touch, press, swipe, mid-air), with inter-rater reliability averaging Cohen's kappa = 0.830. The authors interpret their findings through ability-based design (Wobbrock et al.) and ability-mediating design (Vatavu), offering design recommendations and a roadmap for "imagination-powered accessible computing."
Key findings
Four findings stand out. First, 80% of imagined devices were embodied - conceptualized as resting directly on a body part, most often an open palm or fist of the hand (66.8% of Supporter assignments) - while only 13.2% floated in mid-air and just 0.9% were wheelchair-supported. Participants overwhelmingly preferred to avoid holding devices. Second, ten device archetypes emerged, dominated by smartphones (36.4%) and remote controls (27.3%), with joysticks, tablets, keyboards, photo cameras, selfie sticks, smartwatches, push buttons, and smartglasses filling out the set. Third, while agreement on the conceptualization of devices was moderate (AR = 0.345), agreement on the gestures to operate them was extremely low (AR = 0.069), significantly below the 0.100 threshold for "low agreement" - meaning that although people imagine similar objects, they enact them with highly individual gestures shaped by their specific motor profile. Gestures were balanced across touch (38.2%), press (28.2%), mid-air (24.1%), and swipe (9.5%); 58.6% were bimanual, and the index finger (40.9%) and thumb (39.1%) were the dominant effectors. Fourth, motor impairments systematically shaped preferences: participants with rapid fatigue or low strength favored embodied small-scale gestures, those with spasticity proposed unique accommodations (wheelchair-supported joysticks, thigh-mounted keyboards, eye-gaze shortcuts), and participants with poor coordination avoided body-supported designs.
Relevance
For accessibility designers, this paper reframes what an "input device" can be. Rather than adapting keyboards, mice, or touchscreens to users with motor impairments after the fact, imagined devices are used as design probes that surface what each user can comfortably do and expects to do. This is a concrete methodology that practitioners building gesture-sensing systems, AR/VR interfaces, smart-home controls, or assistive technology can apply directly. The finding that device conceptualization converges while gesture articulation diverges is particularly actionable: it argues against shipping one fixed gesture set and for personalized or ability-adapted recognition, likely backed by zero-shot or few-shot learning. The authors also map their findings cleanly onto ability-based and ability-mediating design frameworks, which gives teams a shared vocabulary for translating the findings into design reviews. Limitations are significant: the sample is small (N=11), drawn from a single country (Romania), focuses on upper-body motor impairments only, and does not evaluate whether imagined-device interactions can actually be recognized in deployed systems. Future work integrating commodity cameras, wearables, and gaze input is needed before these recommendations can be operationalized at scale.
Tags: motor impairment · gesture input · end-user elicitation · imagined devices · ability-based design · accessible computing · assistive technology