In this paper, the author explored a interaction technique that can visualize multivariable dataset in VR environment, that he claim can bring rich and immersive experience. In the literature review section, the author claimed that several research support the view that projecting 3D data visualisations onto a 2D screen suffers from inherent issues due to occlusions, perspective distortion and interaction problems using a 2D input device. While some research claim that 3D representation does not necessarily bring huge benefits than 2D, which is also my primary concern with this paper that, the author did not convince me why 3D in VR is better. The author developed a system called ImAxes, an interactive multi-dimensional visualisation tool that relies on the arrangement of data axes in space. Specifically, based on the 3D visualization and 6DOF interaction techniques that VR system offers, standard visualisations can be created for understanding multidimensional data: scatterplots, SPLOMs, PCPs and their three-dimensional variant. ImAx is primarily built upon a simple grammar which formalises the rules for constructing visualisations in the VRE using axes’ spatial placement. The author presented a study, in which a data scientist would like to explore the relationship of wine differences with its 12 attributes, the author described the entire process how he make use of the ImAxes system and did not present a quantitative study.
Although this work looks cool visually, I doubt the actual usefulness of this technique. More specifically, I do not see a evaluation of achieving the same goal using traditional data visualization packages, thus I cannot see the benefits of using VR in this case. One of his center claim is that projecting 3D visualization onto 2D screen will cause loss due to occlusions and perspective distortions. He made the point by presenting a case in the user study, in which he twists the PCP to resolve occlusion between the pH values spread of the two wine types, and two components that used to be overlapping on 2D plots can be separated like two layers of paper in 3D space. I agree that using VR and hand tracker enables a more intuitive way of combining different datas and change view perspectives, but that’s it, what else benefit does it provide in the data processing? Using this system usually requires expertise in data science, and one can assume such user are capable of various theories as well as analytics tools, this visualization is a good technique for presenting, but does not necessarily bring more benefits in the analysis process. The author also admitted in the discussion section that, it is necessary to conduct a comprehensive evaluation of ImAxes but this validation must go beyond a Likert UX usability validation or base-line comparison. Maybe I’m being over critique on this, but seeing this study not from data science perspective, but for more HCI perspective, this study definitely propose a interesting field to be addressed extensively. (But I still don’t get it why towards the end the author still mentioned visualization in mixed reality)