Northstar

Abstract

In order to democratize data science, we need to fundamentally rethink the current analytics stack, from the user interface to the "guts." Most importantly, enabling a broader range of users to unfold the potential of (their) data requires a change in the interface and the "protection" we offer them. On the one hand, visual interfaces for data science have to be intuitive, easy, and interactive to reach users without a strong background in computer science or statistics. On the other hand, we need to protect users from making false discoveries. Furthermore, it requires that technically involved (and often boring) tasks have to be automatically done by the system so that the user can focus on contributing their domain expertise to the problem. In this paper, we present Northstar, the Interactive Data Science System, which we have developed over the last 4 years to explore designs that make advanced analytics and model building more accessible.

MetaExplorer : Facilitating Reasoning with Epistemic Uncertainty in Meta-analysis

“…In part because of this lack of integration of model checking with VA tools, analysts may not always scrutinize patterns discovered using such tools to ask how exactly they might arise. Exceptions include early research systems developed by and for statisticians [4,56], NorthStar created for predictive modeling [33], and recent research systems developed to study novel interfaces for eliciting analysts' expectations via natural language [9] and sketching [32]. In contrast to these efforts, we developed EVM to study how model check visualizations might benefit the broader populations of analysts that tools like Tableau target, assuming neither that our users would be statisticians nor that realizing model checking necessarily requires a new elicitation medium.…”

Q4EDA: A Novel Strategy for Textual Information Retrieval Based on User Interactions with Visual Representations of Time Series

“…In order to support users in visual data analysis, some tools and techniques allow for a non-obstructive exploratory approach through visual interactivity [14,21], among the most relevant is NorthStar [14], which goes in-depth into the difficulties of providing an exploratory system that follows responsive and real-time guidelines for intuitive and engaging user exploratory analysis while at the same time utilizing automatic problem detectors throughout the entire workflow to reduce the amount of potential bias or incorrect insights generated through the exploration. However, although a certain level of "query conversion" is performed through these tools or techniques, they still expect what is essentially a text-based query to be constructed manually through their visual interface.…”

“…Instead, queries are again constructed manually using the Cadence system through a drag-and-drop interface to perform the search. This procedure is arguably most similar to the interface of NorthStar [14] as opposed to Q4EDA.…”

“…However, when considering structured datasets, up to our knowledge, there is no mechanism to use pieces of visual data, such as user selections in a line chart, to search for related information using general-purpose SEs. Instead of using visual selections, there are solutions that either use visualizations as outputs [11][12][13] or as a visual interface to manually construct what is essentially a text-based query [14][15][16][17]. Even though such visualizations are shown to aid the information gathering processes, in these works, the use of visual selections as visual search queries to fetch relevant information from generalpurpose SEs is an unexplored concept.…”

Amplifying Domain Expertise in Clinical Data Pipelines (Preprint)

“…There is active research on interactive and human-in-the-loop systems in many computer science disciplines. The database and visualization communities have produced numerous tools [3][4][5][6][7][8] to aid data scientists with data wrangling and analysis. At the decision-making stage, the machine learning community has looked at making black box models explainable [2,[9][10][11][12], while the human-computer interaction (HCI) community has been studying how differences in explainability affect decision making [13,14].…”