"If you've gone straight, now, you must turn left" - Exploring the use of a tangible interface in a collaborative treasure hunt for people with visual impairments
Quentin Chibaudel, Wafa Johal, Bernard Oriola, Marc J-M Macé, Pierre Dillenbourg, Valérie Tartas, Christophe Jouffrais · 2020 · Proceedings of the 22nd International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2020) · doi:10.1145/3373625.3417020
Summary
This paper from a collaboration between IRIT (Toulouse), UNSW Sydney, EPFL, and Université Toulouse investigates how tangible user interfaces — specifically a small graspable robot called the Cellulo — can mediate collaboration between two people with visual impairments in a spatial task. The study is notable for focusing on collaboration between two visually impaired participants rather than the more commonly studied pairing of a sighted and a visually impaired person. The researchers designed a collaborative treasure hunt game where one participant (the guide) explores a 2.5D tactile map while the other (the explorer) navigates a real room containing 13 chairs and hidden clues. The guide must verbally direct the explorer to find five clues in sequence. The Cellulo robot (a hexagonal graspable device weighing 168g, 7.5cm across) represents the explorer's position and orientation on the map. Two conditions were tested: Active (the Cellulo moves autonomously to reflect the explorer's real-time position, operated via a Wizard of Oz method) and Passive (the Cellulo is stationary and used by the guide as a manual tool to represent the explorer's location). Six dyads of people with visual impairments (12 participants total — 8 males, 4 females, mean age 36.5, including 1 early blind, 1 late blind, 5 with early low vision, and 5 with late low vision) completed the treasure hunt in both conditions. The game emerged from brainstorming sessions with special education teachers who were enthusiastic about using robots for collaborative spatial learning.
Key findings
The Active Cellulo condition produced significantly better task performance: all dyads found all five clues (versus only three dyads in Passive), the task took 531 seconds on average versus 955 seconds in Passive (p=0.045), and time variability between clues was much smaller in Active (under 50s standard deviation) compared to Passive (92-190s). However, the Passive condition produced richer collaboration. Using Meier's collaboration grid assessed by three independent annotators, five of the ten collaboration factors were significantly different between conditions. Joint Information Processing (p=3.8×10⁻⁵), Reaching Consensus (p=1.7×10⁻⁶), Reciprocal Information (p=1.5×10⁻⁴), and Motivation for both guide (p=0.01) and explorer (p=3.8×10⁻³) were all significantly better in the Passive condition. Only Communication and Coordination dimensions were unaffected by the condition. This reveals a fundamental tension: the Active Cellulo made the task easier but turned the collaboration into a one-directional "guidance task" where the guide gave instructions and the explorer followed. In the Passive condition, both participants had to actively share spatial information, negotiate understanding, and co-construct knowledge about the space — a more equitable collaboration. Interestingly, explorers in the Passive condition showed a non-significant trend toward building better cognitive maps of the room, possibly because the richer dialogue required deeper spatial reasoning. User experience ratings were equally positive in both conditions (4.6 and 4.2 out of ±5), and all participants reported enjoying the game. Guides with some residual vision tended to abandon the Cellulo when difficulties arose, preferring to use their own hand on the map.
Relevance
This study makes several contributions relevant to accessibility practitioners and designers of collaborative technology for people with visual impairments. First, it demonstrates that spatial games can be successfully designed for pairs of visually impaired players without requiring sighted assistance — an important step toward independent recreational and educational activities. Second, it reveals a crucial design insight: making a task easier through technology does not necessarily improve collaboration or learning. The Active condition's real-time tracking reduced cognitive demand but also reduced the need for partners to communicate, share spatial knowledge, and build mutual understanding. This has implications for any assistive technology design — sometimes the "less helpful" design produces better learning and social outcomes. Third, the study validates tangible robots as effective mediators for spatial collaboration among people with visual impairments, opening possibilities for educational applications in orientation and mobility training, spatial skills development, and collaborative problem-solving. The finding that special education teachers were enthusiastic about incorporating such tools into their practice suggests real potential for classroom adoption. For designers of collaborative accessible games, the paper provides a framework for thinking about how interaction design choices shape the quality and equity of collaboration.
Tags: visual impairments · tangible user interface · collaborative learning · spatial cognition · wayfinding · haptic feedback · robots · serious games · 3D models · non-visual interaction