Search Results

Now showing 1 - 4 of 4
  • Item
    Map Augmentation and Sketching for Cycling Experience Elicitation
    (The Eurographics Association, 2024) Reljan-Delaney, Mirela; Wood, Jo D.; Taylor, Alex S.; Hunter, David; Slingsby, Aidan
    This work examines the use of maps for knowledge elicitation in the sphere of urban cycling. The study involved running 14 distinct workshops, each serving as a unique data collection session for a particular individual. In each workshop, the participant was provided with 12 different renditions of the geographical areas as well as drawing materials. The geographical area renditions contained regions specified by the participant as cycling locations during the preparatory correspondence. The outputs were analysed for patterns in map augmentations and thematic content in the sketches. We have found that participants engaged deeply with the map augmentation process expressing their preferences and giving new insights. Themes such as connectivity, scenic beauty, and temporality emerged prominently from the analysed data, shedding light on the subjective experiences and preferences of urban cyclists.
  • Item
    The Misclassification Likelihood Matrix: Some Classes Are More Likely To Be Misclassified Than Others
    (The Eurographics Association, 2024) Sikar, Daniel; Garcez, Artur d'Avila; Bloomfield, Robin; Weyde, Tillman; Peeroo, Kaleem; Singh, Naman; Hutchinson, Maeve; Laksono, Dany; Reljan-Delaney, Mirela; Hunter, David; Slingsby, Aidan
    This study introduces the Misclassification Likelihood Matrix (MLM) as a novel tool for quantifying the reliability of neural network predictions under distribution shifts. The MLM is obtained by leveraging softmax outputs and clustering techniques to measure the distances between the predictions of a trained neural network and class centroids. By analyzing these distances, the MLM provides a comprehensive view of the model's misclassification tendencies, enabling decision-makers to identify the most common and critical sources of errors. The MLM allows for the prioritization of model improvements and the establishment of decision thresholds based on acceptable risk levels. The approach is evaluated on the MNIST dataset using a Convolutional Neural Network (CNN) and a perturbed version of the dataset to simulate distribution shifts. The results demonstrate the effectiveness of the MLM in assessing the reliability of predictions and highlight its potential in enhancing the interpretability and risk mitigation capabilities of neural networks. The implications of this work extend beyond image classification, with ongoing applications in autonomous systems, such as self-driving cars, to improve the safety and reliability of decision-making in complex, real-world environments.
  • Item
    Use of Notebooks and Role of Map features in Mapping Minority Women Bicycle Riding
    (The Eurographics Association, 2024) Reljan-Delaney, Mirela; Wood, Jo D.; Taylor, Alex S.; Hunter, David; Slingsby, Aidan
    Visualization has greatly enhanced our understanding of cycling trends [fL18], enabled the depiction and analysis of largescale cycling data [BWB14, RMGZALD18], and facilitated the tracking and interpretation of personal behaviour through the dashboards of personal tracking devices [NKKW20]. Data and visualization can be either vast and generalized or intimate and personal. There are significant challenges associated with big data as certain subgroups are underrepresented in data collection, making their presence difficult to detect and more targeted and smaller data collection can complement and expose facets of the population that are not visible in big data. Ethnic minority women cyclists are one such group. Research into their attitudes and cycling habits is often outdated [Lim10] or originates from contexts where their ethnicity is the majority [GOF∗22]. This study aims to shed light on the experiences ofMuslim and BAME women cyclists, uncovering hidden realities and challenging dominant narratives. A small group of ethnic minority women participated in the research, keeping diaries of their cycling experiences and using GPS trackers. The collected data was presented back to them in the form of individual data notebooks, combining technology, visualization, and ultimately qualitative analysis. This empirical work provides a fresh perspective on how female cyclists interact with their environment and offers valuable understanding of the preferences and challenges faced by this growing and vibrant group. This paper builds upon the previously published work [RDWT23], shifting the focus away from the methodological execution of the study and instead emphasizing the participants' interactions with the maps and the unique insights gained.
  • Item
    Exploring Data Analysts' Uncertainty Reasoning Strategies for Effective Uncertainty Visualization Design
    (The Eurographics Association, 2024) Kinkeldey, Christoph; Reljan-Delaney, Mirela; Panagiotidou, Georgia; Dykes, Jason; Hunter, David; Slingsby, Aidan
    Despite its proven positive effects, visual data analysis rarely includes information about data uncertainty. Building on past research, we explore the hypothesis that effective uncertainty visualizations must support reasoning strategies that enable data analysts to utilize uncertainty information ('uncertainty reasoning strategies', UnReSt). Through this work, we seek to gain insights into the reasoning strategies employed by domain experts for incorporating uncertainty into their visual analysis. Additionally, we aim to explore effective ways of designing uncertainty visualizations that support these strategies. For this purpose, we developed a methodology involving online meetings that included think-aloud protocols and interviews. We applied the methodology in a user study with five domain experts from the field of epidemiology. Our findings identify, describe, and discuss the UnReSt employed by our participants, allowing for initial recommendations as a foundation for future design guidelines: uncertainty visualization should (i) visually support data analysts in adapting or developing UnReSt, (ii) not facilitate ignoring the uncertainty, (iii) aid in the definition of acceptable levels of uncertainty, and (iv) not hide uncertain parts of the data by default. We reflect on the methodology we developed and applied in our study, addressing challenges related to the recruiting process, the examination of an existing tool along with familiar tasks and data, the design of bespoke prototypes in collaboration with visualization experts, and the timing of the meetings. We encourage visualization researchers to adapt this methodology to gain deeper insights into the UnReSt of data analysts and how uncertainty visualization can effectively support them. The supplemental materials can be found at https://osf.io/s2nwf/.