Interactive Visualization for Fostering Trust in ML
2023, Chau, Polo, Endert, Alex, Keim, Daniel A., Oelke, Daniela
The use of artificial intelligence continues to impact a broad variety of domains, application areas, and people. However, interpretability, understandability, responsibility, accountability, and fairness of the algorithms' results - all crucial for increasing humans' trust into the systems - are still largely missing. The purpose of this seminar is to understand how these components factor into the holistic view of trust. Further, this seminar seeks to identify design guidelines and best practices for how to build interactive visualization systems to calibrate trust.
Towards visual debugging for multi-target time series classification
2020, Schlegel, Udo, Cakmak, Eren, Arnout, Hiba, El-Assady, Mennatallah, Oelke, Daniela, Keim, Daniel A.
Multi-target classification of multivariate time series data poses a challenge in many real-world applications (e.g., predictive maintenance). Machine learning methods, such as random forests and neural networks, support training these classifiers. However, the debugging and analysis of possible misclassifications remain challenging due to the often complex relations between targets, classes, and the multivariate time series data. We propose a model-agnostic visual debugging workflow for multi-target time series classification that enables the examination of relations between targets, partially correct predictions, potential confusions, and the classified time series data. The workflow, as well as the prototype, aims to foster an in-depth analysis of multi-target classification results to identify potential causes of mispredictions visually. We demonstrate the usefulness of the workflow in the field of predictive maintenance in a usage scenario to show how users can iteratively explore and identify critical classes, as well as, relationships between targets.
Guidelines for Effective Usage of Text Highlighting Techniques
2016, Strobelt, Hendrik, Oelke, Daniela, Kwon, Bum Chul, Schreck, Tobias, Pfister, Hanspeter
Semi-automatic text analysis involves manual inspection of text. Often, different text annotations (like part-of-speech or named entities) are indicated by using distinctive text highlighting techniques. In typesetting there exist well-known formatting conventions, such as bold typeface, italics, or background coloring, that are useful for highlighting certain parts of a given text. Also, many advanced techniques for visualization and highlighting of text exist; yet, standard typesetting is common, and the effects of standard typesetting on the perception of text are not fully understood. As such, we surveyed and tested the effectiveness of common text highlighting techniques, both individually and in combination, to discover how to maximize pop-out effects while minimizing visual interference between techniques. To validate our findings, we conducted a series of crowdsourced experiments to determine: i) a ranking of nine commonly-used text highlighting techniques; ii) the degree of visual interference between pairs of text highlighting techniques; iii) the effectiveness of techniques for visual conjunctive search. Our results show that increasing font size works best as a single highlighting technique, and that there are significant visual interferences between some pairs of highlighting techniques. We discuss the pros and cons of different combinations as a design guideline to choose text highlighting techniques for text viewers.
Visual readability analysis : how to make your writings easier to read
2012-05, Oelke, Daniela, Spretke, David, Stoffel, Andreas, Keim, Daniel A.
We present a tool that is specifically designed to support a writer in revising a draft version of a document. In addition to showing which paragraphs and sentences are difficult to read and understand, we assist the reader in understanding why this is the case. This requires features that are expressive predictors of readability, and are also semantically understandable. In the first part of the paper, we, therefore, discuss a semiautomatic feature selection approach that is used to choose appropriate measures from a collection of 141 candidate readability features. In the second part, we present the visual analysis tool VisRA, which allows the user to analyze the feature values across the text and within single sentences. Users can choose between different visual representations accounting for differences in the size of the documents and the availability of information about the physical and logical layout of the documents. We put special emphasis on providing as much transparency as possible to ensure that the user can purposefully improve the readability of a sentence. Several case studies are presented that show the wide range of applicability of our tool. Furthermore, an in-depth evaluation assesses the quality of the measure and investigates how well users do in revising a text with the help of the tool.
The Role of Interactive Visualization in Fostering Trust in AI
2021, Beauxis-Aussalet, Emma, Behrisch, Michael, Borgo, Rita, Chau, Duen Horng, Collins, Christopher, El-Assady, Mennatallah, Keim, Daniel A., Oelke, Daniela, Schreck, Tobias, Strobelt, Hendrik
The increasing use of artificial intelligence (AI) technologies across application domains has prompted our society to pay closer attention to AI's trustworthiness, fairness, interpretability, and accountability. In order to foster trust in AI, it is important to consider the potential of interactive visualization, and how such visualizations help build trust in AI systems. This manifesto discusses the relevance of interactive visualizations and makes the following four claims: i) trust is not a technical problem, ii) trust is dynamic, iii) visualization cannot address all aspects of trust, and iv) visualization is crucial for human agency in AI.
Towards A Rigorous Evaluation Of XAI Methods On Time Series
2019-10, Schlegel, Udo, Arnout, Hiba, El-Assady, Mennatallah, Oelke, Daniela, Keim, Daniel A.
Explainable Artificial Intelligence (XAI) methods are typically deployed to explain and debug black-box machine learning models. However, most proposed XAI methods are black-boxes themselves and designed for images. Thus, they rely on visual interpretability to evaluate and prove explanations. In this work, we apply XAI methods previously used in the image and text-domain on time series. We present a methodology to test and evaluate various XAI methods on time series by introducing new verification techniques to incorporate the temporal dimension. We further conduct preliminary experiments to assess the quality of selected XAI method explanations with various verification methods on a range of datasets and inspecting quality metrics on it. We demonstrate that in our initial experiments, SHAP works robust for all models, but others like DeepLIFT, LRP, and Saliency Maps work better with specific architectures.
Feature-Based Visual Exploration of Text Classification
2015, Stoffel, Florian, Flekova, Lucie, Oelke, Daniela, Gurevych, Iryna, Keim, Daniel A.
There are many applications of text classification such as gender attribution in market research or the identification of forged product reviews on e-commerce sites. Although several automatic methods provide satisfying performance in most application cases, we see a gap in supporting the analyst to understand the results and derive knowledge for future application scenarios. In this paper, we present a visualization driven application that allows analysts to gain insight in text classification tasks such as sentiment detection or authorship attribution on feature level, built with a practitioner’s way of reasoning in mind, the Text Classification Analysis Process.
The Impact of Immersion on Cluster Identification Tasks
2020-01, Kraus, Matthias, Weiler, Niklas, Oelke, Daniela, Kehrer, Johannes, Keim, Daniel A., Fuchs, Johannes
Recent developments in technology encourage the use of head-mounted displays (HMDs) as a medium to explore visualizations in virtual realities (VRs). VR environments (VREs) enable new, more immersive visualization design spaces compared to traditional computer screens. Previous studies in different domains, such as medicine, psychology, and geology, report a positive effect of immersion, e.g., on learning performance or phobia treatment effectiveness. Our work presented in this paper assesses the applicability of those findings to a common task from the information visualization (InfoVis) domain. We conducted a quantitative user study to investigate the impact of immersion on cluster identification tasks in scatterplot visualizations. The main experiment was carried out with 18 participants in a within-subjects setting using four different visualizations, (1) a 2D scatterplot matrix on a screen, (2) a 3D scatterplot on a screen, (3) a 3D scatterplot miniature in a VRE and (4) a fully immersive 3D scatterplot in a VRE. The four visualization design spaces vary in their level of immersion, as shown in a supplementary study. The results of our main study indicate that task performance differs between the investigated visualization design spaces in terms of accuracy, efficiency, memorability, sense of orientation, and user preference. In particular, the 2D visualization on the screen performed worse compared to the 3D visualizations with regard to the measured variables. The study shows that an increased level of immersion can be a substantial benefit in the context of 3D data and cluster detection.
Lessons on Combining Topology and Geography : Visual Analytics for Electrical Outage Management
2016, Jäger, Alexander, Mittelstädt, Sebastian, Oelke, Daniela, Sander, Sonja, Platz, Axel, Bouwman, Gies, Keim, Daniel A.
Outage management in electrical networks is a complex task for operators and requires comprehensive overviews of the topology. At the same time valuable information for detecting the root cause may have geographical context such as digging activities or falling trees. Consequently, vendors of state-of-the-art SCADA systems started to integrate this valuable information source as well. However, in todays systems both views are separated, requiring operators to mentally connect the geographical and topological information. The wish of operators is to provide a comprehensive combination of both spaces in a single view. However, how to project geographical elements into the topology to support the workflow of real operators is yet unclear. In this paper, we present a design study for an interactive visualization system that provides a comprehensive overview for power grid operators. It provides full coverage of both spaces in order to measure how real operators make use of the geographical information. It bypasses the projection problem by interactive brushing-and-linking to support associative analysis. We extracted the mental-model of domain experts in real use cases and found a general bias source in sequential analysis of two spaces. We contribute our problem and task abstraction, lessons learned, and implications for future research.
Comparative Exploration of Document Collections : a Visual Analytics Approach
2014, Oelke, Daniela, Strobelt, Hendrik, Rohrdantz, Christian, Gurevych, Iryna, Deussen, Oliver
We present an analysis and visualization method for computing what distinguishes a given document collection from others. We determine topics that discriminate a subset of collections from the remaining ones by applying probabilistic topic modeling and subsequently approximating the two relevant criteria distinctiveness and characteristicness algorithmically through a set of heuristics. Furthermore, we suggest a novel visualization method called DiTop-View, in which topics are represented by glyphs (topic coins) that are arranged on a 2D plane. Topic coins are designed to encode all information necessary for performing comparative analyses such as the class membership of a topic, its most probable terms and the discriminative relations. We evaluate our topic analysis using statistical measures and a small user experiment and present an expert case study with researchers from political sciences analyzing two real-world datasets.