Worth a Thousand Words: Recommendations for Authoring Image Descriptions

Isabela Presedo-Floyd, Tania Allard

Funding Area: Capacity building / Construcción de capacidad

Problem Statement: The Internet—from research tools to publishing platforms to documentation for critical digital infrastructure—is filled with images. Missing image descriptions (or alt text) remain the second most common accessibility failure on the Internet. The World Health Organization estimates at least 16% of the global population has a disability. Yet, accessibility expertise is rare in open source and tech spaces. With new content published every few seconds, undescribed or poorly described images further exclude people with disabilities. Consider a diagram, a flow chart, or a data visualization on the documentation of a digital infrastructure project. Readers are expected to gain information through visual intake of images; without alt text, that information is lost to many disabled people. Sharing accessibility expertise is required for open source and research ecosystems to deliver on their promises of inclusion, transparency, and access. In the last 2 years, the authors have improved image descriptions in open source projects while disseminating knowledge on writing alt text. While these efforts have been successful, they have been limited by timezone availability and lack of a central reference point. An effort to document, share, and publicize expertise is needed to address this pan-Internet problem. The outcomes of this proposal can positively impact broad open research, open scholarship, open source communities, and the millions of disabled people interacting with them.

Proposed Activities: Months 0-2 1.Revisit the draft of image description guidelines and other private sources to identify content gaps and needing updating or expanding. 2.Post the first draft of guidelines on GitHub and as a public webpage for consultation and feedback. The team will carry out several outreach activities, such as •Targeted outreach to existing collaborators (e.g., open source projects such as Jupyter, NumPy, SciPy, Matplotlib, and initiatives such as the Turing Way and Metadocencia). •Open community outreach through public channels 3.Logistic and operational work for paid structured interviews with people with disabilities and accessibility advocates and experts. Months 2-4 1.Respond to and triage requests and feedback gathered through GitHub. 2.Facilitate structured interviews. 3.Synthesise and publish results from structured interviews. 4.Close the open solicitation period and synthesize the feedback. Months 4-6 1.Incorporate feedback collected. 2.Publish the updated set of alt-text and image description guidelines. 3.Disseminate resources among communities and projects. 4.Hold two open-office sessions for broader community members (i.e., open source projects, researchers, educators, and content creators) interested in adopting the alt-text guidelines as part of their content creation and documentation practices. Stretch goal: reach out to open source and open research Spanish-speaking communities for future translation and localization of the image description authoring guidelines. Note that any translation and localization work falls out of scope due to time and resource constraints for this application. Thus we’d need to identify potential/future funding sources to make this work happen. Expertise and resources needed 1.User experience research: •Design and conduct structured interviews •Synthesise and publish research findings and feedback •Experience engaging with people with disabilities 2.Contributor experience and community stewardship: •Respond to and triage GitHub collected feedback. •Generating documentation sites through open source documentation engines (such as Sphinx or Docusaurus) and writing documentation for a diverse audience. •Organizing and facilitating accessible and inclusive community-focused events 3.Accessibility expertise: •Knowledge of WCAG 2.1 AA requirements for content authoring and production •Awareness of different types of disabilities and assistive technology. •Knowledge of digital accessibility and disability inclusion. •Experience creating accessible content. •Experience creating content that reflects diversity and inclusion. 4.General open source-related expertise: •Open source knowledge •Git/GitHub practical knowledge •Familiarity with open source and open research communities and open practices •Experience working with a diverse range of community members. Resources: Access to a computer and Internet, survey tools, virtual call platform, Git/GitHub, and real-time collaboration documents.

Openness: This work builds on 16 open alt-text workshops with open source scientific computing community members. Following open source principles, the resources generated for and during the events are public for anyone to use, reuse, or modify. These events were designed with a co-creation and co-participation model in mind, both for their development and how they are run to: 1. Raise awareness on the importance of alt-text for the accessibility of content published on the web. 2. Contribute in the form of alt-text descriptions to several open source projects in the scientific computing ecosystem. 3. Mentor the participants and share best practices for useful and contextualized image descriptions. Some of the outcomes and outputs of these events are: 1. The development and maintenance of open resources that could be reused by other community members willing to run similar events with their communities with little overhead. 2. These events enabled multiple first-time contributions through mentorship and support from maintainers or core devs of the participating projects. The proposed work will continue to center on community co-creation and participation by including an open feedback solicitation period and focused interviews. We will also hold open office hours to support and guide community members interested in better understanding and integrating the alt-text guidelines in their content creation processes. Generated resources will be released under open licenses (CC-BY 4.0).

Challenges: Based on prior experience, the most significant challenges will likely be achieving high participation rates of individuals with disabilities in the open feedback solicitation period and identifying and recruiting people with disabilities within our community for the structured interviews. Many open source communities have disproportionately low participation and representation of disabled people due to systemic, attitudinal, and technological barriers. While some work is and has been done to raise disability awareness and improve the accessibility of a handful of tools of our ecosystem (even if marginally in some cases), most of our spaces, tools, and digital infrastructure remain remarkably inaccessible to many people with disabilities. We plan to mitigate some risks associated with such challenges by: 1. Conducting proactive and targeted recruitment of disabled people for the structured interviews. We will also rely on our network of collaborators across the open source and open research communities for this. 2. In the spirit of recognizing and not relying on free labor from an already marginalized group in our community, we have requested a sum in our budget to account for honoraria for each participant.

Neglectedness: Accessibility-focused design and development, including remediation work, has historically been one of the most neglected areas within tech. There is a notable lack of accessibility experts in the industry, making it significantly challenging to include accessibility as a standard practice across all development teams. This expertise is even more limited within open source projects, especially those heavily relying on free labor and where disciplines such as design have little representation (e.g., scientific computing). Also, minimal funding sources exist for maintenance and capacity-building research software. As such, the open source scientific community continues to rely on volunteers for most work unrelated to innovation but often fundamental for a project's sustainability. Accessibility, security, documentation, and user experience are massively underfunded, underresourced, and often neglected critical areas. To the authors' knowledge, there are no dedicated funding sources for accessibility-related work within open source. The team previously secured financing for accessibility remediation work within JupyterLab under the CZI EOSS grants program. However, such a project centered on design and development improvements to specific tools and user-testing rather than writing accessibility-focused documentation or guidelines. This is the first time the team has applied for funding to develop accessibility-focused documentation.

Success: This project proposal centers on multiple verticals such as community stewardship and contribution experience, disability inclusion and accessibility, and open source practices. Therefore we have identified success criteria aligned with and reflecting the effort needed on each of those verticals: 1. The image description documentation includes a framework for general image description guidance and specific recommendations for standard image types encountered in research, educational, and scientific computing content. This documentation has at least three examples of image description recommendations in use and contextualized for the relevant audiences. 2. The image description documentation is hosted somewhere public and accessible by the broader community under a permissive open license. Open and transparent processes exist to flag issues and provide feedback (such as a GitHub issue tracker). 3. There are clear contribution guidelines aimed at a diverse range of contributions and technical expertise. 4. The results of the public solicitation period and the structured interviews are open and published under an open and permissive license. All data should be aggregated and anonymized to protect the interviewees' personal information. 5. Feedback should include people representing at least three organizations, projects, or affiliations. 6. Two open office hours across different time zones are organized and conducted.

Total Budget: $11,060.00

Affiliations: Quansight LLC

LMIE Carveout: Our project does not fit within this category.

Team Skills: This project's scope requires rare skills in open source spaces. We've identified the following as critical: 1. Accessibility expertise. This team has worked on multiple accessibility projects centering Web Content Accessibility Guidelines conformance. We have shared accessibility knowledge in workshops and conference talks. 2. Image description expertise. Cumulatively, team members have written hundreds of image descriptions. We have experience with information-dense images related to code infrastructure and scientific research. Additionally, we have experience handling the differences in documentation infrastructure necessary to support image descriptions. 3. Experience authoring recommendations for cross-project use. We regularly document processes for a range of projects. Quansight Labs has a history of collaborating on open standards like the Consortium for Python Data APIs, and enhancement proposals across open source projects such as Core Python, NumPy, SciPy, Jupyter, conda-forge, conda, and SymPy. 4. Meeting the community where they are. We know how to find project forums, mailing lists, community calendars, and more while respecting local expectations. Our team members have a history of contribution as participants, organizers, and within leadership and decision-making positions. 5. Synchronous and asynchronous facilitation of community feedback. Team members have years of experience soliciting feedback in communities like Project Jupyter, napari, and Spyder.

How Did You Hear About This Call: Word of mouth (e.g. conversations and emails from IOI staff, friends, colleagues, etc.) / Boca a boca (por ejemplo, conversaciones y correos electrónicos del personal del IOI, amigos, colegas, etc.)

Submission Number: 127