UX Research: Enhancing Graph Accessibility for Blind Users

Graphs are a common method of communicating information to audiences but are inaccessible to users who are blind. As a part of this research, we posed two questions:

  • What information do blind users desire to obtain and retain from a graph description?

  • How can designers improve written graph descriptions for users who are blind?

My Role

Initially, I worked on this project with 3 other classmates as a part of my Masters Program. After completing a pilot study, consisting of usability tests and semi-structured interviews with blind participants, two of my classmates and I decided to involve blind users in the co-construction of graph descriptions. We conducted participatory design sessions with blind users and later conducted an evaluative usability test on revised graph descriptions. My role including creating usability test and participatory design guides, moderating sessions, taking field notes, leading analysis sessions, and testing the digital prototype employed for our evaluative usability test.

Pilot Study

Semi-Structured Interviews & Usability Tests 

 

Simplest Graph Description used in Usability Tests

We conducted semi-structured interviews and usability tests with 4 blind participants. Based on a review of prior literature, we constructed three graph descriptions of varying complexity to employ during our usability tests.  
 
During each session, we presented participants with the graph descriptions and asked questions to assess their understanding of the content and their satisfaction with the description. Additionally, we asked questions about our participants’ knowledge and experience with graphs.

Analysis

I led a remote session with team members to analyze the data gathered from our semi-structured interviews and usability tests. Over the course of 2-hour Google Hangout, we used Stormboard and affinity diagramming techniques to derive themes from the data gathered.

Time lapse of our affinity diagram process using Stormboard, courtesy of team member Casey Hudetz

Findings

We found that blind users desire to understand the type of graph (i.e. bar or pie), the graph’s scope or range of values, the purpose or context for the graph, the categories, values and trends. In addition, we found that blind users desire graph descriptions that are broken down into sections so that they can return to specific sections if they wish to review particular details.

Outcomes

Based on our findings, we created three guidelines for designers to follow to enhance the accessibility of graphs for blind users along with a paper describing our process and findings.

State the Context

Designers must state the graph’s purpose and context. For instance, a graph may serve to support an argument or to display detailed findings.

Provide Access to Data on Demand

Designers must provide access to the graph’s data on demand. Designers must include a clear hierarchy of raw data in their descriptions that breaks down the graph into sections so that users can skip to different sections.

Minimize Cognitive Load

Designers must follow techniques to increase memory retention and reduce cognitive load. These techniques include chunking information together, using clear declarative writing, and removing repetitive descriptions.

Participatory Design 

Participatory Design Sessions

Revised Graph Description Format used for Participatory Design Sessions

Using the insights gathered from our pilot study, we made several revisions to the graph descriptions & our study:

  • To broaden our understanding of how to design descriptions for different types of graphs, we decided to use one bar graph and two line graphs
  • We divided the graph description into individual sections (Overview, Context, Trend, Values) based on preferences indicated by participants in the pilot
  • We reduced any redundant text and added a statement at the end of the description instructing participants to repeat a particular section if they preferred

We then conducted participatory design sessions with 3 blind users to elicit their feedback on the revised graph descriptions and questions that we had created.

Analysis

Initially, we inductively analyzed our Participatory Design transcripts using Dedoose. After one round of individual coding, I facilitated a collaborative analysis session with my teammates. Using a white board, post-it notes, and markers, we uncovered themes about our participants' prior experiences with graphs and their preferences.

Affinity Diagram of Data Gathered from Participatory Design Sessions

Findings

We uncovered various insights about participants' past experiences with graphs and their desires for a graph description. Some of the insights are shown below.

Evaluative Usability Tests

Usability Tests

Using the insights gathered from both the pilot study and participatory design sessions, we made several revisions to our graph descriptions & our study including:

Screenshot of Digital Prototype Employed for Usability Study

  • Added Visual Description of the Graph and Purpose sections to help users understand the graph's purpose and gain a spatial awareness of the content

  • Added headings for each graph description section

  • Included graphs with articles to simulate how people might use the graph descriptions in their daily lives

  • Created a website with the graphs, articles, and descriptions to assess the effectiveness of our descriptions when used with screen readers

  • Revised assessment questions used during the study

We conducted usability tests with 7 people who were blind. During each usability session, we instructed users to navigate the graph descriptions on the website on their own. Participants used a MacBook, an external keyboard, and VoiceOver to hear the contents of the website. 

Photo Taken from a Usability Session

Analysis

To assess the effectiveness of each graph description, we compared the answers to each assessment question. We also compared the clarity ratings across all participants and compared these for each graph. We each extracted quotes from the recorded interviews. Over a 2-hour Google Hangout session, we used Stormboard and affinity diagramming techniques to identify common and salient themes.

Findings

On average, participants reported moderate satisfaction with our graph descriptions.

Mean Satisfaction Ratings by Question & Graph

Participants' ability to answer graph content questions varied greatly. Overall, participants answered more questions related to graphs 1 and 2 (graph descriptions presented within an article) than graph 3 (standalone graph).

% of Participants that Answered Graph Content Questions Correctly

We uncovered various challenges faced by participants when navigating the graph descriptions including:

  • Difficulties using VoiceOver to navigate graph descriptions
  • Confusion distinguishing graph article and description
  • Participants' Mental model of the graph doesn't match the description
  • Difficulties drawing conclusions from the data table

Outcome

From the insights uncovered from our pilot study, participatory design sessions, and evaluative usability tests, we created guidelines for designers to follow in order to enhance the accessibility of graphs. Currently, we are working to submit a paper of our work to an HCI conference.