We investigate the use of subjective views for co-located collab-orative Augmented Reality (AR) environments for 3D scientific visualization, i. e., views that are perceived differently per user, no matter their viewpoints. AR technology allows collaborators to benefit seamlessly from both their visualization and social spaces. However, those same collaborators might have different roles and expertise levels, which might need one visualization rather than another. When using one active screen per user, an AR collaborative system can render the scene with different parameters per user, yet with the same spatial reference. As 3D visualization often relies on multiple data layers, heavy filtering, and several transfer functions, the resulting multi-view representations may no longer all be shown or can lead to conflicts in the users' social space where collaborate with others. Here we propose multiple designs to overcome this issue.