How students reason about visualizations from large professionally collected data sets: A study of students approaching the threshold of data proficiency

This study identifies a population of students who have an intermediate amount of relevant content knowledge and skill for working with data, and characterizes their approach to interpreting a challenging data-based visualization. Thirty-three undergraduate students enrolled in an introductory environmental science course reasoned about salinity data as shown in map and vertical profiles from the Mediterranean while thinking aloud and being eye-tracked. Students reasoned about 2D and 3D interpretations in the context of two hypothesis arrays (a suite of potential interpretations about a set of data). Findings suggest the students have some effective strategies in reading data: They look at cartographic elements, correctly identify the image as a salinity map, and draw inferences from the data. Common looking strategies include scanning along the salinity gradient, comparing areas of interest, and aligning the color bar with the map. Individual differences emerge in the interpretation of the data, with no interpretations being fully aligned with the scientifically normative explanation. Post hoc analyses identify reasoning tasks and spontaneous behaviors related to a construct we refer to as “data expertise,” which is intended to capture the degree of conceptual sophistication and resourcefulness in reasoning about data. A data expertise scale was developed, with scores ranging from zero (weak) to six (strong) that were normally distributed. Our findings suggest that appropriately coordinating data with a model, comparing and contrasting across data representations from different times or places, and extracting 3D structure from 2D representations are associated with data expertise.