MoveTogether: Exploring Physical Co-op Gameplay in Mixed-Reality
Pin Chun Lu, Wen-Fan Wang, Che Wei Wang, Ting-Ying Lee, TsaiHsuan Lin, Duo-Jie Hsiao, CheHan Hsieh, YuTing Tseng, Neng-Hao Yu, Mike Y. Chen · 2026 · Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI '26) · doi:10.1145/3772318.3791957
Summary
This CHI 2026 paper introduces MoveTogether, a mixed-reality (MR) co-op gameplay mechanic in which two co-located players jointly operate a single tracked physical prop that is virtually transformed inside a Meta Quest 3 MR scene (a ring, a trampoline, a bubble nozzle). The core claim is that most existing co-op games keep collaboration virtual even when players share a room, using separate controllers and separate avatars, and that adding a shared physical channel changes the coordination, communication and safety profile of co-located play. The authors build a two-axis design space (shared vs separate physical controllers x shared vs separate virtual objects), run a three-hour co-design workshop with ten professional VR/game/industrial/interaction designers that produces ten props and fifty two-player interactions, analyse the workshop via Reflexive Thematic Analysis, and distil four prop design dimensions (grasping mode, size, geometry, material) and eleven interaction primitives (balance, move, squeeze, stretch, rotate, flip for handheld; move, rotate, pull, lift, wave for non-handheld). They then build a MoveTogether MR prototype using a ring-shaped foam prop and run a within-subjects user study (n=16, 8 pre-existing pairs) comparing physical co-op to a virtual co-op baseline across a trampoline scenario (moving, balancing) and a bubble scenario (squeezing, rotating), with subjective Likert measures, video-coded collisions and collaborative cues, and post-scenario interviews. The paper does not study disabled players or assistive-technology use; its stated audience is co-op MR game designers. Participants were 19-28, able-bodied, recruited as pre-existing pairs. For this accessibility-focused review, findings are read for their transfer value to shared-control assistive co-op, analogous to Xbox Adaptive Controller 'co-pilot' patterns in which a disabled player and a partner jointly drive a single in-game avatar.
Key findings
Physical co-op was significantly preferred on collaboration (Trampoline p<.001, Bubble p<.01), sense of achievement and engagement, and participants rated the shared-prop condition as dramatically safer: video coding showed about 91% fewer person-person collisions (from 21 to 2), 53% fewer total collisions (34 to 16) and 93% fewer near-misses (44 to 3). Rather than reducing verbal communication, the haptic channel increased it - task-related verbal cues rose ~2.4x and directional cues ~3.4x - with speech becoming more strategic and anticipatory rather than purely operational. Prop design mattered: radially symmetric, medium-sized (60-120 cm) handheld props with clear grasp points and bilateral or radial symmetry (like the final ring-shaped prop) were preferred because they removed the need to agree on orientation and let partners infer each other's intent from subtle kinesthetic pulls. Mirrored coordination (both players doing the same action) felt predictable and fair; complementary coordination (one pulls, one resists) was more engaging but produced uneven 'leader-follower' dynamics. Critically for the accessibility-transfer angle, the authors explicitly flag that shared props 'can amplify player asymmetries' in height, strength, movement speed or skill - participants described feeling 'dragged around' or 'held back' by a mismatched partner - and they propose dynamic difficulty adjustment (adding virtual resistance on the stronger player's side) and asymmetric mechanical roles (one player as 'engine', the other as 'rudder') as design levers to rebalance the experience without removing the shared physical link.
Relevance
For accessibility practitioners, MoveTogether is useful less as an accessibility study and more as an evidence base for designs that extend the Xbox Adaptive Controller 'co-pilot' pattern - where control of a single avatar is split between a disabled player and a partner or support worker - into MR, VR and physical-prop play. The paper provides empirically grounded vocabulary (grasping mode, size, geometry, material; the eleven interaction primitives) for reasoning about shared controllers, and shows that shared physical coupling can replace lost visual channels when HMDs restrict peripheral vision - an effect likely to matter even more for players with low vision, blindness, or reduced field of view. The collision-reduction result is directly relevant to co-located play involving mobility-aid users, wheelchair users or players with coordination differences, where safety is a first-order design constraint. The authors' explicit treatment of 'mediating player asymmetries' - dynamic difficulty adjustment and asymmetric roles - maps cleanly onto co-op design for mixed-ability pairs: the stronger or more skilled player can be handicapped in software so the shared artefact stays literally and experientially shared. Important limitations for accessibility-focused readers: no disabled participants, small sample of pre-existing able-bodied pairs aged 19-28, a single ring-shaped prop in the user study, and no evaluation of seated play, one-handed play, switch-access pairing or fatigue. The design space is a starting point, not a validated accessibility contribution.
Tags: mixed reality · shared haptic controller · co-located collaboration · tangible interaction · accessible gaming · shared control · co-op gameplay · haptics · game accessibility · assistive technology