Root cause analysis is a common data analysis task. While question-answering systems enable people to easily articulate a why question (e.g., why students in Massachusetts have high ACT Math scores on average) and obtain an answer, these systems often produce questionable causal claims. To investigate how such claims might mislead users, we conducted two crowdsourced experiments to study the impact of showing different information on user perceptions of a question-answering system. We found that in a system that occasionally provided unreasonable responses, showing a scatterplot increased the plausibility of unreasonable causal claims. Also, simply warning participants that correlation is not causation seemed to lead participants to accept reasonable causal claims more cautiously. We observed a strong tendency among participants to associate correlation with causation. Yet, the warning appeared to reduce the tendency. Grounded in the findings, we propose ways to reduce the illusion of causality when using question-answering systems.

While decision makers have begun to employ machine learning, machine learning models may make predictions that bias against certain demographic groups. Semi-automated bias detection tools often present reports of automatically-detected biases using a recommendation list or visual cues. However, there is a lack of guidance concerning which presentation style to use in what scenarios. We conducted a small lab study with 16 participants to investigate how presentation style might affect user behaviors in reviewing bias reports. Participants used both a prototype with a recommendation list and a prototype with visual cues for bias detection. We found that participants often wanted to investigate the performance measures that were not automatically detected as biases. Yet, when using the prototype with a recommendation list, they tended to give less consideration to such measures. Grounded in the findings, we propose information load and comprehensiveness as two axes for characterizing bias detection tasks and illustrate how the two axes could be adopted to reason about when to use a recommendation list or visual cues.

Asking pairwise comparison questions is common. Yet, we often find ourselves comparing apples and oranges --- the two entities of interest are not readily comparable. To understand how technologies can extend our capabilities to conduct pairwise comparisons during data analysis, we analyzed pairwise comparison questions collected from crowd workers and propose a taxonomy of pairwise comparisons. We demonstrate how the taxonomy can be adopted by incorporating pairwise comparison capabilities into Duo, a spreadsheet application that supports comparing two groups of records in a data table. Duo decomposes a pairwise comparison question into rules and showcases sloppy rules, a query technique for specifying pairwise comparisons. We conducted a user study comparing sloppy rules and natural language. The findings suggest that for easier pairwise comparison tasks, the two techniques are comparable in efficiency and preference and that for more difficult pairwise comparison tasks, sloppy rules allow faster specification and are more preferable.

Data analysis novices often encounter barriers in executing low-level operations for pairwise comparisons. They may also run into barriers in interpreting the artifacts (e.g., visualizations) created as a result of the operations. We developed Duet, a visual analysis system designed to help data analysis novices conduct pairwise comparisons by addressing execution and interpretation barriers. To reduce the barriers in executing low-level operations during pairwise comparison, Duet employs minimal specification: when one object group (i.e. a group of records in a data table) is specified, Duet recommends object groups that are similar to or different from the specified one; when two object groups are specified, Duet recommends similar and different attributes between them. To lower the barriers in interpreting its recommendations, Duet explains the recommended groups and attributes using both visualizations and textual descriptions. We conducted a qualitative evaluation with eight participants to understand the effectiveness of Duet. The results suggest that minimal specification is easy to use and Duet's explanations are helpful for interpreting the recommendations despite some usability issues.

Exploring event sequences by defining queries alone or by using mining algorithms alone is often not sufficient to support analysis. Analysts often interweave querying and mining in a recursive manner during event sequence analysis: sequences extracted as query results are used for mining patterns, patterns generated are incorporated into a new query for segmenting the sequences, and the resulting segments are mined or queried again. To support flexible analysis, we propose a framework that describes the process of interwoven querying and mining. Based on this framework, we developed MAQUI, a Mining And Querying User Interface that enables recursive event sequence exploration. To understand the efficacy of MAQUI, we conducted two case studies with domain experts. The findings suggest that the capability of interweaving querying and mining helps the participants articulate their questions and gain novel insights from their data.

Centralized matching is a ubiquitous resource allocation problem. In a centralized matching problem, each agent has a preference list ranking the other agents and a central planner is responsible for matching the agents manually or with an algorithm. While algorithms can find a matching which optimizes some performance metrics, they are used as a black box and preclude the central planner from applying his domain knowledge to find a matching which aligns better with the user tasks. Furthermore, the existing matching visualization techniques (i.e. bipartite graph and adjacency matrix) fail in helping the central planner understand the differences between matchings. In this paper, we present VisMatchmaker, a visualization system which allows the central planner to explore alternatives to an algorithm-generated matching. We identified three common tasks in the process of matching adjustment: problem detection, matching recommendation and matching evaluation. We classified matching comparison into three levels and designed visualization techniques for them, including the number line view and the stacked graph view. Two types of algorithmic support, namely direct assignment and range search, and their interactive operations are also provided to enable the user to apply his domain knowledge in matching adjustment.