The visualization market will increase at a compound annual growth rate (CAGR) of 9.21% from $4.12 billions in 2014 to $6.40 billions by the end of 2019 (Mordor Intelligence, 2014). Visual analytics combines visualization and data analysis. It draws from research practises in information visualization and science visualization by developing visualization technologies and tools (Olshannikova, Ometov, Koucheryavy, & Olsson, 2015, p. 9). But, there are epistemic issues with big data (Floridi, 2012). I want to explore the epistemic issues specific to visualisation.

Virtual Reality is an accessible technology for users and researchers. What can we do with it? How does it affect thinking and reasoning?

Virtual reality research to date has focussed on accuracy with the exclusion of philosophy or psychology of the cognitive impacts of our choices based on what data we represent. E.g. what if we make a crime scene that appears real, yet is built on biased, incomplete evidence. Virtual worlds could have a disproportionate, unjustified impact on a jury’s beliefs.

If a virtual crime scene is constructed from incomplete data, it may have a disproportionate impact on jury beliefs. Virtual reality models need to not just consider perceptual fidelity, but how the choice and representation of visual data affects reasoning, decision-making, belief and memory.

In VR wallaby environment experiment, experts picked out possible wallaby habitats and assigned narrow confidence margins to their estimates. Did the virtual reality environment make them more confident than they should have been?

Evidence suggests that humans are irrational because we systematically make mistakes in reasoning (Daniel Kahneman, 2011; D. Kahneman, Slovic, & Tversky, 1982).

Debias Anchoring
Avoid anchors
Provide multiple and counter anchors
Use different experts who use different anchors

Debias Confirmation bias
Use multiple experts with different points of view about hypotheses,
Challenge probability assessments with counterfactuals,
Probe for evidence for alternative hypotheses,

Data dashboards get made, but who really knows how to read them to make better decisions? Who enjoys inhabiting dashboards? What if information could be made more psychologically engaging?

3D hasn't improved data comprehension and analysis. How do we make a better decision making environment?

Let's create a beautiful virtual reality environment. A place that resonates with the human mind. A place that people want to be in.

The VR environment would be a Pleistocene savannah, within which modern human beings evolved over 1.6 million years. In style it would mimic a sublime Hudson River School painting or African savannah with open spaces of low grasses interspersed with copses of trees. Trees have forks close to the ground to offer places of refuge against attack. There is the presence of water; animal or bird life; diverse greenery, and a path (perhaps a river bank or shoreline) that extends into the distance, inviting you to follow it.

The choice of savannah comes from literature in the philosophy of aesthetics [2] , Vitaly Komar and Alexander Melamid’s research on artistic preferences of people in ten countries [3] and Orians & Heerwagen [4] ’savanna hypothesis’.
"body of psychological scholarship that is much more potent in addressing cross-cultural landscape tastes. It is a wide-ranging literature, some of it statistical (not unlike Komar and Melamid's poll) and also theoretical, offering hypotheses to explain pervasive tastes for natural habitats. Though the ideas behind it are old, it was initiated in its contemporary incarnation in the 1970s by Jay Appleton, particularly in his book The Experience of Landscape. Appleton's ideas were deepened and extended by Roger S. Ulrich, connected to larger issues of cognition and perception by Stephen and Rachel Kaplan, and validated and given a summary expression by Gordon H. Orians and Judith H. Heerwagen [4]. Orians put forward a general account of the kind of ideal landscape that human beings would find intrinsically pleasurable. In his formulation, this landscape has much in common with the savannas and woodlands of East Africa where hominids split off from chimpanzee lineages and much of early human evolution occurred; hence it is called "the Savanna Hypothesis." Briefly, this landscape type includes these elements:

open spaces of low (or mown) grasses interspersed with thickets of bushes and groupings of trees;

the presence of water directly in view, or evidence of water nearby or in the distance;

an opening up in at least one direction to an unimpeded vantage on the horizon;

evidence of animal and bird life; and

a diversity of greenery, including flowering and fruiting plants.

By now, this research is developed enough to be able to say much more about innate landscape preferences. These preferences turn out to be more than just vague, general attractions toward generic scenes: they are notably specific. African savannas are not only the probable scene of a significant portion of human evolution, they are to an extent the habitat meat-eating hominids evolved for: savannas contain more protein per square mile than any other landscape type. Moreover, savannas offer food at close to ground level, unlike rain forests, tropical or temperate, which are more easily navigable by tree-dwelling apes.

Human beings are less attracted to absolutely open, flat grasslands and more toward a moderate degree of hilly undulation, suggesting a desire to attain vantage points for orientation. Verdant savannas are preferred experimentally to savannas in the dry season.

Note, arguments about humans evolving to process information based on particular evolutionary landscapes is contentious. Researchers often accuse evolutionary psychologists of making up ‘just so’ stories (aka Rudyard Kipling) or storytelling [5]. We don’t have to buy into an argument that we have ‘the right landscape’ to use this landscape as a test for ecological decision-making

Can we create a virtual reality savannah and build the data points into the objects, shapes, contours and features of the virtual world?

I propose to start research by testing for anchoring in VR and then attempting to mitigate against anchoring through lures.

For example, suppose we ask participants to guess how many wildebeest exit the main group before the herd changes direction?

We get VR wearers to circle the wildebeest with their controllers in hand. This establishes a quantitative output from the experimental design.

Then, we prime participants with a VR training task. Some subjects would get a small number of creatures to train with, others would get a great deal. We expect the number of animals presented during training would anchor participants response to the Wildebeest task.

If training can cause anchoring. Can we reverse or improve the bias by including visual groups to sway participants in the opposite direction? That is, do the guesses of participants change if the numbers of non-Wildebeest animals in the VR environment during the test vary, e.g. a herd of 15 zebras vs. 3 zebras near the Wildebeest?

...and does it enable us to make better judgements about it.