The Role of Uncertainty, Awareness, and Trust in Visual Analytics

Abstract

Visual analytics supports humans in generating knowledge from large and often complex datasets. Evidence is collected, collated and cross-linked with our existing knowledge. In the process, a myriad of analytical and visualisation techniques are employed to generate a visual representation of the data. These often introduce their own uncertainties, in addition to the ones inherent in the data, and these propagated and compounded uncertainties can result in impaired decision making. The user's confidence or trust in the results depends on the extent of user's awareness of the underlying uncertainties generated on the system side. This paper unpacks the uncertainties that propagate through visual analytics systems, illustrates how human's perceptual and cognitive biases influence the user's awareness of such uncertainties, and how this affects the user's trust building. The knowledge generation model for visual analytics is used to provide a terminology and framework to discuss the consequences of these aspects in knowledge construction and though examples, machine uncertainty is compared to human trust measures with provenance. Furthermore, guidelines for the design of uncertainty-aware systems are presented that can aid the user in better decision making.

A review of volunteered geographic information quality assessment methods

“…This is crucial because more and more text and image-based VGI are being utilized in various applications. Furthermore, the works of Sacha et al (2016), where they introduce a framework that integrates trust and other various quality indicators in a knowledge generation process within the visual analytics paradigm can be adapted in future research to assess and visually analyze quality of VGI. Their framework allows the user to comprehend the associated quality at each step of knowledge generation, and also express their confidence in the findings and insights gained by externalizing their thoughts.…”

How Data Workers Cope with Uncertainty

“…Focusing on the human analyst, Grolemund and Wickham [9] propose a cognitive model and interpretation of the data analysis process that describes the effects of cognitive bias inherited from sense-making processes, with a goal of improving current data-analysis. Other work has examined relationships between uncertainty and human trust [30] to propose guidelines and challenges aiming to handle uncertainty, and to expose frameworks for uncertainty-aware visual analytics systems [6,25]. By better understanding the role of uncertainty in the analytic process, we can provide data workers with much-needed tools to facilitate reflection over the relationships in a whole data set [1] and to provide support for decision-making in the face of uncertainty [2].…”

Reducing the Analytical Bottleneck for Domain Scientists: Lessons from a Climate Data Visualization Case Study

“…Traditional hypothesis-driven methods and tools are inadequate to handle the complexity of these multiple combinations, where scientific questions often depend on the exploratory analysis process. In scenarios where scientists pursue unknown unknowns, 8 these tools don't support rapid exploration of alternative hypotheses. The analytical bottleneck is caused by three main inadequacies in current analysis tools: the ability of scientists to easily convert their highlevel analysis goals into the visualization interface through interactions, the lack of multiple perspectives into the data as scientists often need to look at different views before reaching their conclusions, and rapid, dynamic exploration of different hypotheses in which the system adapts to the interactions, proactively searches for interesting patterns, and helps scientists in narrowing down their visual search process.…”

“…8 To facilitate a more transparent analysis process, we'll explore open areas of research in high-dimensional data visualization 15 by developing and applying promising techniques such as subspace search 16 and topology-based analysis of climate model data. Subspace clustering methods can be leveraged to suggest interesting variable combinations and let scientists visually search and detect relevant subspaces.…”