Separation, Composition, or Hybrid? - Comparing Collaborative 3D Object Manipulation Techniques for Handheld Augmented Reality

Augmented Reality (AR) supported collaboration is a popular topic in HCI research. Previous work has shown the benefits of collaborative 3D object manipulation and identified two possibilities: Either separate or compose users' inputs. However, their experimental comparison using handheld AR displays is still missing. We, therefore, conducted an experiment in which we tasked 24 dyads with collaboratively positioning virtual objects in handheld AR using three manipulation techniques: 1) Separation - performing only different manipulation tasks (i. e., translation or rotation) simultaneously, 2) Composition - performing only the same manipulation tasks simultaneously and combining individual inputs using a merge policy, and 3) Hybrid - performing any manipulation tasks simultaneously, enabling dynamic transitions between Separation and Composition. While all techniques were similarly effective, Composition was least efficient, with higher subjective workload and worse user experience. Preferences were polarized between clear work division (Separation) and freedom of action (Hybrid). Based on our findings, we offer research and design implications.