Navigation auf uzh.ch

Suche

Department of Informatics Interactive Visual Data Analysis Group

IVDA Research Streams

Research at IVDA includes various methodologies and branches, all related to the meaningful combination of both strengths of humans and machines in the data analysis, knowledge generation, and decision-making process.

One of IVDA's assets is its support for data-centric, model-centric, and human-centric approaches, as well as their combinations, i.e., analysis processes that enable all three entities to benefit from (gray directions in the diagram).

In addition, our EEE Framework describes Human knowledge and preference Externalization, Model and output Explanation, and Data Exploration, as well as their combinations (orange, blue, and green directions in the diagram). This perspective on data analysis is about how entities can contribute to analytics processes, enabling modern forms of human-model-data interaction and collaboration.

IVDA research particularly studies data analytics challenges with complexities requiring many of these directions in unified interactive visual data analysis approaches and systems. Most sophisticated IVDA systems support all six directions between Humans, Models, and Data.

Links to Data-centric research: IVDA for Event Sequences | IVDA for Time Series
Links to Model-centric research: Interactive Machine Learning at IVDA
Links to Human-centric research: Data Humanism at IVDA

General research streams are as follows.

Exploratory Data Analysis

Exploratory Data Analysis (EDA) in Visual Analytics enables the intuitive discovery of new knowledge through interactive engagement with high-dimensional data. By leveraging dynamic visualizations, EDA facilitates pattern recognition, hypothesis generation, and insight extraction. Interactive tools empower users to explore complex datasets, guiding analysis through human intuition, machine learning, and visual feedback loops.

Links to application domains: Digital Libraries

Explainable Artificial Intelligence

Explainable Artificial Intelligence (XAI) in Visual Analytics integrates interpretable and trustworthy AI by combining human-centric visual exploration with algorithmic transparency. It enhances user understanding, enabling interaction with models via interpretable representations. Trustworthy AI fosters reliability, fairness, and accountability, ensuring that decision-making aligns with human intuition, fostering trust and informed insights.
 

Knowledge and Preference Externalization

Knowledge and Preference Externalization refers to the process of transforming tacit knowledge into explicit forms through interactive visual interfaces. Rooted in social sciences and pioneered in visualization research, it enables users to express preferences via annotations, scores, and rankings. This facilitates human-system interaction, linking feedback to ranking algorithms for decision support.
 

Links: Interactive Data Labeling at IVDA  | Interactive Item Ranking at IVDA

Interactive Data Labeling

Labeling data instances is a fundamental prerequisite for supervised machine learning. Our research studies the user-in-the-loop data labeling process from the machine learning perspective (leveraging active learning) and the interactive and visual perspective (enabling humans to select instances by themselves). We base our work on our “visual interactive labeling” (VIAL) methodology, which unifies both perspectives.

Links: Interactive Data Labeling at IVDA

Interactive Item Ranking

The human-centered ranking of multidimensional items is a non-trivial task. Comparing complex items to each other to create a ranking can be time-consuming and, especially if data sets are large, there is no guarantee that the result is satisfying. We characterize, design, and evaluate interactive solutions for the creation of item rankings and design and develop ranking explainers, to enable users to gain more trust into given ranking models.

Links: Interactive Item Ranking at IVDA

Interactive Relation Discovery

Interactive Relation Discovery enables the identification of dependencies, relations, and causalities between data attributes through interactive exploration. It supports research and practice by uncovering hidden patterns, enhancing data-driven decision-making. By fostering awareness of data interconnections, it improves model interpretability, supports hypothesis generation, and strengthens analytical reasoning across domains.

Links to application domains: Digital Libraries

Human-Model Collaboration

Human-Model Collaboration emphasizes understanding humans, models, and data as a foundation for designing effective communication and interaction. It fosters data-driven decision-making through intuitive dialogues between users and AI systems, leveraging the strengths of both. By aligning human insight with model capabilities, it enables more robust, transparent, and collaborative analytical processes.

Ethical Approaches to AI

Ethical Approaches to AI integrate ethical and humanistic principles into data-driven decision-making and machine learning applications. They aim to reflect human values, promote fairness, and reduce bias by ensuring transparency, accountability, and inclusivity. These approaches guide responsible AI development, aligning technological progress with societal well-being and individual rights.

Publications of the IVDA Group

ZORA Publication List

Download Options

Download as CSV Download as EndNote Download as BibTeX Download as XML

Publications

Pagination

Bereichs-Navigation

Unterseiten von Research