Analyzing big data for actionable insights is one of today’s critical challenges in computing. Neither human analysts nor automated algorithms alone have the ability to answer subtle, deep questions from very large abstract datasets. Visualization of large datasets in virtual reality (VR) or augmented reality (AR) can help analysts spot trends and outliers, but visualization by itself is not sufficient either. In this talk, I will review research on the benefits of immersion for spatial data understanding and effective interaction methods for AR/VR. I’ll contrast immersive analytics with traditional visualization, and propose the concept of Immersive Space to Think (IST). In IST, human analysts use immersive virtual reality (VR) to view and organize data items such as documents. The immersive 3D environment becomes a cognitive affordance with nearly unlimited space so that the analysis process can be externalized. At the same time, analytic algorithms and data models support analysis by displaying information such as related data items, similarity or credibility estimates, and text summaries. Finally, IST understands the semantics implied by observations of users and their interactions, and uses these to modify and guide the underlying algorithms. Based on the use case of analyzing the credibility of news, I’ll describe a broad research agenda to learn how to design effective IST systems and to gather evidence of the benefits of the IST approach.