This paper proposes a scenario for the analysis of interaction mediated by AR. Using this scenario (a) we can easily track all objects in space and over time and record who handles each at which moment, (b) we can easily adjust the displayed augmented objects, (c) we can add meta-information next to the objects in the users' visual field, and (d) we can explore truly multimodal interactions, such as allowing users to perceive the soundscape at any location on the plan by interactively mixing the acoustic contributions that the exhibits make. Most importantly, we will be able to control which information will be perceived by which participant, for example, presenting different features of the object to the two participants (small vs. big, silent vs. noisy, etc.), so that we are able to induce potentially problematic situations which will allow us to investigate how participants deal with such non-obvious misinterpretation of the setting.