Data-driven Biased Decision-making? - Exploring the landscape between dashboards, visualization literacy and decision bias

Abstract: Data quantities and their sources have amplified over the years and so has the trend to employ dashboard-based data visualizations into the hands of a wider audience of end-users. By selecting four of the most common data visualization formats and combining these into a dashboard this thesis quantitively explored the relationship between similarity features of dashboard-based data visualizations, interpretation accuracy and systematic errors in decision-making i.e. decision biases as defined by Kahneman and Tversky (1974). By sampling 87 business practitioners through a double-blind randomized field experiment conducted at a large IT-company in Sweden, the objective of this thesis was to gauge the nature and extent of the relationship between dashboard-based data visualizations, interpretation accuracy and decision biases. The results of the field experiment did not suggest a relationship between similarity features of dashboard-based data visualizations and decision biases. The relationship between peoples’ ability to interpret these data visualizations and decision biases was more nuanced, suggesting no overall bias while a difference between two natural groups with a [i]high[/i] and [i]low[/i] degree of interpretation accuracy could be demonstrated. The discussion highlights the implications of quantitatively analyzing systematic errors or decision biases that may arise in the expanding territory of dashboard-based data visualizations.