VIRTUAL EXPERIENCE AS A PROXY FOR NUMERACY IN PROBABILITY COMMUNICATION

Purpose: Non-numerate decision makers need to know probabilities defining their decisions under uncertainty. As a substitute for numeric, verbal and graphic probability representations which are more effective for the numerate than the non-numerate, probability is directly experienced and unconsciously constructed through a stream of user-controlled, visually mediated, binomial trials. Secondarily, the purpose is to investigate the differences in properties of experiential probability communication between the numerate and non-numerate components of the general patient population.

Method: We have developed a novel patient oriented decision aid in the form of interactive animated decision trees. The simplest of these trees are single-stage binomial chance trees which are used for the purpose of communicating know probabilities to subjects. The subjects are given control of the interface so that they may use it as long as necessary, stopping if mentally fatigued, and restarting when desired. The interface has been implemented in both an HTML-hosted Java Swing applet and an HTML5/Javascript interface. Interstep temporal varies between 250ms and 350ms. Duration of path indication in the ‘on’ state varies between 200ms and 300ms. Different subjects may experience identical sets of probability values to be communicated. For purposes of studying this decision aid, subjects are permitted  to repeat their probability communication exercise. Each time a different pseudo-random number is calculated. Preprogrammed pseudo-random number generators guarantee sequences of unequal sequential values so that subjects cannot report the prior value as theirs. Subjects effectively become actors in an interactive, visual, Monte Carlo simulation where its direct experience is the intended method of communicating the probability values. Investigators were concerned that subjects might have been counting, rather than experiencing, success and failure outcomes and calculating, rather than intuiting, the probabilities reported. To minimize this possibility, short times between successive trials were defined. In addition, subjects were asked if they counted.

Result: Subjects (N = 14) have performed 32 trials. Correlation between input probability to communicate, and probability reported by subject is 0.92, an encouraging value. No users reported counting numbers of success and failures of trials.

Conclusion: Further work needed to determine differences between numerate and non-numerate performance. Trials are needed to determine if subjects over or underreport low and high probabilities. It is not known if experiential probability communication yields consistent results for pairs of experienced complementary values.