Overcoming biases to improve search sufficiency and decision accuracy: the effects of data visualization, instructions and order
T h e U n i v e r s i t y o f U t a h G r a d u a t e S c h o o l STATEMENT OF DISSERTATION APPROVAL The dissertation of Heidi S. Kramer has been approved by the following supervisory committee members: Frank A Drews , Chair ABSTRACT The present study explored data presentation and human cognition with the objective of improving electronic Decision Support Systems (DSS). Computers have been used as tools for decision support for over 60 years, with the intent to supplement or replace human
... replace human cognition. However, electronic computing has failed to reliably replace human cognition in complex domains. The suboptimal properties of the data and complexities of the domain often require human interpretation and intervention. Human interpretation relies on experience, values, intuition, insight and learning; which can lead to shortcuts or heuristics. Heuristics in the correct context can be economical and effective in solving many problems. When heuristics fail the results are labeled as cognitive biases or errors. Biases all share the elements of structuring incorrect or inappropriate models or hypotheses and/or insufficient consideration of the data. Most biases can be linked to confirmation bias -which is manifested by searches for and consideration of only confirming data. De-biasing techniques share the concept of shifting cognitive processing from an automatic associative mode to a more deliberate, conscious rule-based mode. This study used a modified Wason 2-4-6 task that combined methods of, 1) increased salience through data visualization with 2) appealing to the rulebased system through task instructions. The results indicate that neither increased salience nor instructions ensure increased search sufficiency, efficiency or decision accuracy. However, this study provides insight into the perceived value of evidence and iv four potential limitations related to self-directed searches: 1) The selection of necessary disconfirming evidence cannot be assumed, regardless of the perceived value of disconfirming evidence. 2) The selection of sufficient evidence does not ensure accuracy; however, 3) insufficient selection of disconfirming evidence results in lower accuracy. 4) Ambiguous evidence is considered more valuable than potentially disconfirming evidence. Implications for the design of decision support systems are presented along with limitations and directions for future research.