Let’s put immersive together with analytics. How do these concepts relate?
Immersive analytics is: Being there with others to generalize beyond data known about a system. Thus, we have…
- place or context (there)
- group or team (others)
- purpose or activity (generalizing)
- object or focus (system).
Where is There?
The place could be a large room with a bunch of computers and whiteboards. Sound familiar? The IA community would like to explore the potential of a different type of place – virtual data world – an immersive virtual space has the potential of being an unlimited space where scale and time are malleable. It must be believable but not replicate physical reality. It must be coherent where objects and interactions make sense.
Who are these Others?
First, there is you – a person who enters into the data world for the purpose of understanding a specific system. This person could capture the experience in a PowerPoint slide or video for the staff meeting. Or, the person could invite colleagues in-world to experience the same. The data world can then become more social and collaborative in its experience. For instance, you go to an art museum to study an exhibit and discover some important insights. You can take photos to explain your insights to friends. Or, you can take them to the museum and guide them to similar insights. Afterwards, you have a drink together to discuss your shared experience and resulting insights. That analogy aptly describes being there with others.
How to do this Generalizing?
The objective is to augment human judgment with analytic reasoning (HJ-AR). Sounds complicated, but we all do the same every hour of every day for all the normal stuff of life. With immersive analytics, we are attempting to be more precise and purposeful in this augmentation. Humans are terrible at proper statistics! We get confused and backwards with conditional and low probability events. We need help with large complex systems, upon which our jobs and maybe lives depend.
The current BI paradigm for this HJ-AR augmentation is to display well-designed interactive visualizations and let users find interesting patterns. We have gone well beyond the static pen-and-paper figures. Now, understanding comes more from the interaction than from the static image. A PowerPoint slide does not capture the experience. What if… Take the current interactive visualization tools and bring them in-world with the data. Create a virtual data lab where groups can play together. If insights are discovered, bring others to watch (sense) the recorded interactions, sharing the discovery experience.
There is a deeper problem. We see a pattern of sales data over time and infer that sales for next month will be a certain amount. How did we decide on that amount? What is happening is that we are taking a descriptive analysis and inferring the amount via our human judgment based on visual recognition. This has worked for centuries, but things have gotten too complex for this strategy to continue to be effective. Analytic reasoning needs to assist in going beyond the data that is known, albeit displayed nicely. The critical points in advanced analytics are:
- Sampling: typical data from a multi-terabyte dataset
- Attribute distributions: likely/unlikely, more than means, etc.
- Training versus test data: keeping them separate, overfitting
- Bias versus variance tradeoffs: dealing with false pos/neg
- Model ensembles: consider not just one model, but hundreds
- Deep learning: bringing it into the open
Immersive analytics does not solve the above nasty problems, but it may provide a useful approach to find practical solutions.
What is this System?
The intent is to extend the unit of analysis from the dataset to the system …from matrices of correlations among columns to a network of causal relationships among entities. Current analytics is getting excellent at munching tidy data. However, analytics is good at suggesting causal relationships but poor at validating causality. Here is where human judgment must intervene.
If we are to understand and even manage complex systems (like our corporations and governments), then we have to achieve a healthy symbiosis of human judgment and analytical reasoning.