Uncertainty and Ambiguity: Challenging Layers in Model Construction.

When building models that causally explain observed data and predict future data, quantifiable uncertainty and unquantifiable uncertainty (ambiguity) should be considered. Every decision and action that we take in life is associated with a degree of doubt, whether it be uncertainty or ambiguity: whether we turn right or left at an intersection, what research idea we follow, ...and the thousands of other decisions that we make on a daily basis. In decision making, doubt can manifest itself in a variety of ways: one could have ...doubts about the data itself; doubts about what data is needed; doubts about the available processes and transformations; doubts about the possible models; doubts about the decision criteria; ...or even doubts on one’s own preferences for any of these options. This contribution will reflect on model construction and provide some answers. We propose that uncertainty and ambiguity are factors that must be considered in model construction. Machine learning within health care and medical fields is becoming popular and proving incredibly fruitful in the areas of predicting diseases and analyzing transmission of diseases. A major class of problems in medical science involves the diagnosis of disease, based upon various tests performed upon the patient. The evaluation of data taken from patients and complex decision making are the most important factors in diagnosis. A publicly available database for breast cancer prediction will be used to study ambiguity and uncertainty in model construction.