Understanding the efficacy of interactive visualization for decision making for complex systems

Interactive visualization has become a popular approach to support decision makers coping with complex decision problems in the commercial world. However, it is not clear that these interactive visualizations result in better decision making. Human judgment under uncertainty is known to fall victim to a number of biases that result from the heuristics that we employ, yet the design of many interactive visualizations seems to favor aesthetic impact over mitigating biases. We hypothesize that as the complexity of the decision problem increases, improperly designed visualizations will lead decision makers to mistake spurious correlations as causal relationships. In this paper, we discuss findings from the literature that reinforce this concern. We consider the possibility that properly designed aiding could be used to mitigate these effects, and we describe a series of experiments that we are performing to test this hypothesis. Our first experiment suggested that experts were less likely to latch on to an oversimplified causal mechanism than non-experts. Consequently, our second experiment is intended to test an aiding approach designed to encourage non-experts to consider more possibilities.