Concept-Driven Visual Analytics

Abstract

Visualization tools facilitate exploratory data analysis, but fall short at supporting hypothesis-based reasoning. We conducted an exploratory study to investigate how visualizations might support a concept-driven analysis style, where users can optionally share their hypotheses and conceptual models in natural language, and receive customized plots depicting the fit of their models to the data. We report on how participants leveraged these unique affordances for visual analysis. We found that a majority of participants articulated meaningful models and predictions, utilizing them as entry points to sensemaking. We contribute an abstract typology representing the types of models participants held and externalized as data expectations. Our findings suggest ways for rearchitecting visual analytics tools to better support hypothesis-and model-based reasoning, in addition to their traditional role in exploratory analysis. We discuss the design implications and reflect on the potential benefits and challenges involved. CCS CONCEPTS • Human-centered computing → Visual analytics; Information visualization;

MetaExplorer : Facilitating Reasoning with Epistemic Uncertainty in Meta-analysis

“…Multiple participants attributed some difficulties to lack of knowledge about the domain of the datasets (3/12) or to being rusty at specifying regression models in R (5/12). In contrast to prior work showing that VA users don't follow up on model misfit [8], we find that all participants attempted to reason about what might be wrong with misfit models. Our results suggest the need for future work on guided model elicitation interfaces.…”

“…Importantly, beyond need-finding for model checking interfaces, we also add to a line of work [8,9,32] demonstrating that model checks can integrate cleanly into exploratory visual analysis workflows, offering support to users transitioning between open-ended exploration tasks [2] and what Tukey [50] calls "rough confirmatory analysis." Participants who spend relatively little time working with programming interfaces appreciated that EVM offered access to modeling and diagnostics approaches which usually require scripting in statistical programming languages.…”

“…[51]. Whereas previous research [8,9,32] attempted to avoid risks of post-hoc inference by forcing analysts to specify models before seeing the results of queries, and suggested further mitigations through automated adjustment of test statistics, we opted to focus EVM on making support for model check visualizations as seamless as possible, without even providing numerical model summaries like p-values for users to exploit. We built EVM to investigate how VA users rely on unconstrained visual checking to search a space of plausible models, rather than presupposing a hypothesis testing framework where elicited models necessarily reflect a user's best-guess expectation.…”

“…In a prior predictive check, the choice of prior determines the scale of the predictive distribution, enabling this approach to show more possible structures in predictions that are consistent with a model specification but which are ruled out by the fitting process. Contrasting these approaches raises the question of whether enabling users to specify priors and sample from them would be a more direct way to assess users' expectations through visual checks, to the extent that elicited models represent users' beliefs as assumed by prior work [8,9,32]. However, designing for Bayesian predictive checks might increase the level of modeling complexity beyond what some VA users are familiar with, e.g., requiring procedures for fine tuning and justifying priors.…”

Visualization of Data Analysis for Evaluation of the Level of It-Specialists Competencies’ Creation

“…Thus, large educational data is simulated by information visualization and data analysis (VA) methods and presented in visual interfaces with which human visual perception interacts to influence the process of analytical thinking (Khan and Khan, 2011;Cashman et al, 2018;Keim et al , 2008;Choi et al, 2019).…”