"Customization is Key": Reconfigurable Content Tokens for Accessible Data Visualizations

TL;DR

This work tackles the heterogeneity of BLV users by introducing a token-based customization model for screen-reader accessible visualizations, implemented as extensions to the Olli toolkit. It formalizes customization as sequences of content tokens with affordances, directions, and brevity, plus a persistent settings menu and an ephemeral command box to cover both long-term and task-specific needs. A 13-person study with BLV participants shows that customization can improve autonomy and efficiency in data exploration, though it introduces learning overhead and varies with user experience. The authors also outline future directions, including adding an interpreting affordance and extending the model to multisensory representations, highlighting the practical potential to tailor accessibility to diverse tasks and user profiles.

Abstract

Customization is crucial for making visualizations accessible to blind and low-vision (BLV) people with widely-varying needs. But what makes for usable or useful customization? We identify four design goals for how BLV people should be able to customize screen-reader-accessible visualizations: presence, or what content is included; verbosity, or how concisely content is presented; ordering, or how content is sequenced; and, duration, or how long customizations are active. To meet these goals, we model a customization as a sequence of content tokens, each with a set of adjustable properties. We instantiate our model by extending Olli, an open-source accessible visualization toolkit, with a settings menu and command box for persistent and ephemeral customization respectively. Through a study with 13 BLV participants, we find that customization increases the ease of identifying and remembering information. However, customization also introduces additional complexity, making it more helpful for users familiar with similar tools.

Figures (6)

• Figure 1: The relationship between the structure of a tree-shaped navigable hierarchy and the visualization's corresponding Olli hierarchy. The left figure is adapted from Zong, Lee, Lundgard et al. zong_rich_2022. It shows four levels of the tree and how users can use arrow keys to move between them. The right figure shows how the four levels are instantiated in Olli. Each level of the tree is collapsed by default (in this case the y-axis and legend) and expands as the user moves into the level. This allows users to "zoom in" on selections of the data broken out by any of the three fields (in this case flipper length, body mass, and species).
• Figure 2: Our design specification for customization. We define customizations that meet the four design goals we identify: they contain an ordered list of tokens, which addresses presence and ordering (DG1, DG3); each token has its own brevity, controlling verbosity (DG2), and each customization has a duration (DG4).
• Figure 3: Two different customizations of Olli hierarchies for a chart showing five technology companies' stock prices between 2000 and 2010. (a) the visualization; (b) a customization that includes more tokens, with longer brevity, more suitable for novice users who need additional assistance in forming the correct mental model of the graph; (c) a customization with fewer, brief tokens more suitable for an expert user who might have a well-formed mental model of the chart.
• Figure 4: The Olli settings menu. "Facet", "Axis", "Section", and "Datapoint" correspond to the four levels of the Olli hierarchy. The user can set a separate persistent customization for each level, with three default options of high, medium, and low, as well as the option to create new customizations.
• Figure 5: The user interface for creating a new customization for the settings menu. Each customization is specific to one hierarchy level. For each token that can be included in that hierarchy level, the user can choose whether to exclude it, include it with a short brevity, or with a long brevity. They can also choose to reorder tokens. This creates a complete customization that meets the presence, verbosity, and ordering goals (DG1, 2, 3).