OSCON 2026 Regular Session Presentation Abstracts
Find below the abstracts for OSCON 2026's Regular Session presentations.
March 23 Morning Session
Three tracks in the Amphitheater, Room 310, and Room 311.
We Make the Internet: An open source-powered journey to leave big tech behind and have fun on the computer
11:30am - Amphitheater
Presented by Max Turer
Lately, a lot of us have been feeling overwhelmed by what feels like an inescapable dependence on technology, and I recently made the decision many others have made to spend more time away from screens. However, as a programmer, digital artist, and creative coder, limiting my interaction with technology meant less time engaging with my hobbies, interests, and friends and community around the world. I tried reframing the problem: Is technology inherently stressing us all out, or could there be a better way to use it? Could it even be–radical idea incoming–fun to be online? In this talk, I’ll talk about my experience implementing my experiment: Replace as much proprietary software in my life as possible with open source options, and find a way to help those in my life with less or no programming experience do the same. I’ll discuss my various projects for all sorts of skill levels, from beginner linux distributions to building a fully open source portable music player, and hopefully encourage a few others to try some out, as well.
Life of an OSS Developer
11:30am - Room 310
Presented by Brandon Mitchell
Contributing to Open Source Software often involves developing a change for an upstream project to accept. But what does the ecosystem look like on the other side of that transaction? Why would some projects refuse a pull request? When does a contribution go from being helpful to harmful? And what's involved in becoming a maintainer of a well used project? This talk describes what it takes to be an OSS maintainer, to help you be a better contributor, and to consider if you want to be a maintainer yourself.
sdstudio: A Companion Package for Designing and Managing surveydown Surveys
11:30am - Room 311
Presented by Pingfan Hu
Survey research is fundamental to social sciences, market research, and data collection across numerous disciplines. Last year in GW OSCON, we introduced surveydown as an open-source, feature-packed package for programmable survey design & deployment. This year, we bring sdstudio, a GUI platform that dramatically increases the usability of surveydown for survey design, preview, and database management. Built using the Shiny framework, sdstudio features three main components: a visual survey builder with real-time code synchronization (the "Build" tab), an interactive preview system supporting both desktop and mobile views (the "Preview" tab), and a data management interface for response collection and analysis (the "Responses" tab).
Peerbots & Vizij: Embracing Modular Open Source Technologies in an Open Source System
11:50am - Amphitheater
Presented by Saad Elbeleidy
Come learn about the Peerbots platform and the modular approach we used to build it that allows us to incorporate open source modules such as Vizij.
The Peerbots platform includes (1) a free webapp that turns any screen into a controllable robot face, (2) a free controller to control expressions made by a social robot, (3) free authoring tools to define robot scripts for dialogue, (4) an open marketplace for authors to share their social robot scripts, and (5) an API for makers and builders to control the Peerbots Face.
Vizij is an open source ecosystem of software packages for designing, animating and deploying rendered robot faces. The Vizij ecosystem includes several tools that enable simple yet expressive and easily deployable robot faces.
In this talk, we'll discuss how Peerbots embraced and relies on Vizij as a key module for the Peerbots platform. We'll discuss how a modular approach allows us to grow the platform, contribute to other ecosystems, and create win-win opportunities. We'll go over examples of what's possible in social robotics on a budget for the maker, the researcher, and the curious.
Beyond Code Contributions: Network Position, Temporal Bursts, and Code Review Dynamics in Large-Scale Open Source Ecosystems
11:50am - Room 310
Presented by Sean P. Goggins
Open source software (OSS) projects rely on complex networks of contributors whose interactions drive innovation and sustainability. This talk presents a comprehensive analysis of OSS contributor networks using advanced graph neural networks and temporal network analysis on data spanning 25 years from the Cloud Native Computing Foundation ecosystem, encompassing sandbox, incubating, and graduated projects. Our analysis of thousands of contributors across hundreds of repositories reveals that OSS networks exhibit strong power-law distributions in influence, with the top 1% of contributors controlling a substantial portion of network influence. Statistical analysis reveals significant correlations between specific action types (commits, pull requests, issues) and contributor influence, with multiple regression models explaining substantial variance in influence metrics. Temporal analysis shows that network density, clustering coefficients, and modularity exhibit statistically significant temporal trends, with distinct regime changes coinciding with major project milestones. Structural integrity simulations show that Bridge contributors, despite representing a small fraction of the network, have a disproportionate impact on network cohesion when removed. Our findings provide empirical evidence for strategic contributor-retention policies and actionable insights into community health metrics.
Lucid Ledger, One Year Later: Building an Open-Source Wage Protection Infrastructure
11:50am - Room 311
Presented by Emmanuel Teitelbaum
Last year at the GW Open Source Conference, we introduced Lucid Ledger as a proof-of-concept for using smart contracts and institutional design to reduce wage theft and labor exploitation. In this talk, Emmanuel Teitelbaum will present a technical and architectural update on the project’s development over the past year, including a redesigned user interface, a fully integrated backend and database, and progress on core smart contracts, escrow logic, authentication, and oracle integration.
The second part of the talk reflects on the practical challenges of building a complex, open-source system with limited funding and a small, student-led team. These challenges included breaking a large, interdependent system into tractable development tasks; matching tasks to contributors with differing technical backgrounds; coordinating UI/UX, backend, database, contract, and oracle development; and making feature decisions in the absence of a formal product design process or a clear MVP. The team also encountered a familiar catch-22: meaningful market feedback was hard to obtain without a working system, yet standard “lean” guidance discourages substantial development prior to user validation.
The final part of the talk looks forward. Recent advances in AI-assisted development have substantially lowered coordination and implementation costs, making it feasible to iterate faster with fewer resources. It concludes by outlining concrete opportunities for collaboration across development, testing, documentation, demos, and market research, and by inviting contributors interested in helping shape the next phase of Lucid Ledger as an open-source project.
Building Cities, Not Empires: The Business of Open Source
12:10pm - Amphitheater
Presented by Dan Funk
How do you build a sustainable business when your entire competitive advantage is giving everything away?
Early in his career, Dan wrote Visual Basic in a world where Microsoft controlled the documentation, never admitted bugs, and made fixing real problems impossible. When a friend introduced him to Linux (and proceeded to remotely fill his desktop with swimming Koi using xFish) everything changed. He discovered that open source wasn't just about free code; it was about freedom from artificial constraints.
Fast forward to today: Dan's team maintains SpiffWorkflow, a Python-based BPMN workflow engine with a 17-year open source lineage. They have learned that succeeding in open source requires rejecting traditional business wisdom about "moats" and embracing a different model entirely.
This talk explores the three pillars that make open source work:
- Open Standards: Why BPMN creates transferable skills and enables entire ecosystems (unlike walled-garden app stores)
- Open Ideas: The delicate balance between intellectual property and knowledge sharing that's driven innovation since the Enlightenment* * Community: GitHub stars aren't vanity metrics; they're proof your project actually exists
In open source, success isn't about building walls around your castle. It is about being the most valuable vendor in the thriving city you helped construct. Build cities, not empires.
Funding Open Source as a Public Good
12:10pm - Room 310
Presented by Jonathan Starr
Open source software is everywhere, from the servers running hospitals to the code powering scientific research. Yet institutions still struggle to recognize it as a public good. Open Source Endowment is a registered 501c3 that makes the case for open source as critical public infrastructure, akin to roads, libraries, and basic research. Our view of the open source community as an institution in its own right forces the question: what if the open source "institution" had its own sustainable source of funding?
OSE's endowment model opens new possibilities for how we fund the software the world depends on, with an approach that is permanent, community-owned, and independent.
This talk will walk through OSE's positions on open source as a public good and our vision for funding it moving forward. We'll explore how our community-driven, data-first endowment model, can provide permanent, stable support for critical projects and shift the conversation from short-term fixes to systemic, long-term solutions, defined and owned by the community itself.
CanisLupus2.0: Microbiome Analysis Dashboard
12:10pm - Room 311
Presented by Philip Yamoah Appiah
CanisLupus2.0 is an open-source, R Shiny–based web application that streamlines microbiome data exploration and visualization for researchers and students without requiring advanced programming skills. Built on the phyloseq ecosystem, the app allows users to upload ASV/OTU tables, taxonomic assignments, metadata, and phylogenetic trees, then interactively generate diversity analyses, ordinations, and community composition plots within a reproducible workflow. This talk will introduce the motivation and design of CanisLupus2.0, demonstrate key features using real microbiome datasets, and highlight how careful UI/UX choices and modular code support extensibility, teaching, and collaboration in open science. Attendees will learn how to deploy the app locally or on a server, adapt modules to their own pipelines, and contribute enhancements via GitHub to grow a community-driven tool for accessible microbiome analytics.
March 23 Afternoon Sessions
Three tracks in the amphitheater, room 310, and 311.
Open Practices, Open Places: A Librarian’s Guide to Open Source
2:45pm - Amphitheater
Presented by Rosemary Pauley
Libraries are more than bookshelves; they’re places where people learn, make, and share. This talk discusses how open source software and practices fit naturally into libraries. Rosemary will explain simple, practical ways libraries can use open source tools for cataloging, discovery, events, etc., and how librarians can teach basic programming skills to patrons. This session also covers open source in makerspaces, low-cost, community-focused labs where students and patrons learn skills in electronics, 3D printing, crafts, and creative problem solving using open designs and free tools. Through short examples from real libraries and organizations, attendees will discover how open source reduces costs, avoids vendor lock-in, and supports community needs. This session focuses on easy steps librarians can take to embrace open practices, such as choosing projects, running workshops, partnering with tech groups, and contributing back to projects in small, meaningful ways. Attendees will leave with a short list of starter tools, a better understanding of the ways in which libraries already have relationships with openness, and ideas for building inclusive, hands-on learning in their libraries.
Open Source in US Census Bureau Geographic Update Applications
2:45pm - Room 310
Presented by Maria Panaccione-Tweve
This talk will cover how the Census Bureau is using open-source GIS tools to create web-based GIS solutions that are innovative, adaptable, and versatile.
The Census Bureau is enhancing its Geographic Update Partnership Software (GUPS) to more fully utilize open- source and web-based technologies. Built on open- source software, GUPS enables partners to load, display, and edit geospatial data, as well as output standardized files that the Census Bureau then uses to update the Master Address File/Topologically Integrated Geographic Encoding and Referencing (MAF/TIGER) System.
Previous versions of GUPS were only available as a desktop application. Now, with GUPS Web, GUPS is also available as a web application built on QGIS, OpenLayers, and GeoServer. Continuing off the successful implementation of functionality such as real time sharing of data with partners and automated spatial analysis principles, GUPS Web is introducing non-face-based editing for a more streamlined user experience.
In this presentation, representatives from the Census Bureau will showcase the technology used in developing the GUPS Web application and explore how GUPS evolved into a cloud native and license-free web-based GIS application.
From Credibility Frameworks to Verifiable Research Pipelines for AI-driven Science
2:45pm - Room 311
Presented by Tina Morrison
EQTYLab's culture centers on supporting open science and Web3; we're developing a "trust infrastructure" to support research integrity when data and code are shared as part of publication, enabling replication with verifiable verifiable research pipelines.
Building on 17 years as a regulator, Dr. Morrison developed frameworks enabling regulators to assess methodology credibility within their context of use (COU). However, AI-driven research challenges reliance on social trust, and assessment complexity fosters skepticism rather than confidence.
Traditional assessment focuses on verification (correct implementation?) and validation (appropriate methods?). Dr. Morrison argues that verifiable pipelines—where the computer attests to software builds and compute actions—shift assessment from "did the code really run as stated?" to questions critical for research integrity, like, is the pipeline relevant to the stated COU?
High-stakes biomedical research should adopt verifiable pipelines that automatically generate compact, cryptographic "evidence bundles" at each step: software compilation, data curation, model training, inference, and report generation, all digitally signed by the computer. This enables confirmation of what built, what ran, where, with which inputs/outputs, and under which controls—without re-running work or accessing protected information.
The talk showcases a viable prototype with: (1) verify-as-you-go certificates for builds packaged with data/code availability statements; (2) COU-and-claims mapping results to evidence bundles; and (3) verifiable manifests for offline third-party integrity confirmation, verifiable at any future time.
Using AI-model builds and real-world data examples, we'll demonstrate how verifiable infrastructure enhances trust and shifts focus to relevance of data/ methods—essential for open science maintaining rigorous integrity standards with opensource tools.
Fostering Clinical Research Transparency @ Mass General Brigham Through an Open Science License Library
3:05pm - Amphitheater
Presented by Marvin Barksdale
Despite representing critical approaches in Mass General Brigham’s “research to publication to commercialization” pipeline, open science distribution was largely the wild-wild-west before the MGB OSPO. MGB clinical researchers were operating in the fringe – grey areas of NIH mandated open science best practices and MGB IP and Data policy. Researchers lacked clarity on the enterprise’s position on open distributing system data and IP, and the system lacked clarity on the scale and impact of open science activity. As a Bus-Dev Director at Innovation (MGB’s Tech Transfer Office) Marvin Barksdale observed a growing number of invention disclosures from researchers exploring open science distribution options across a spectrum of activities including open access research publication, creative commons licensing, and traditional open source distribution, all with an eye toward commercialization. What started off as engaging with MGB's General Counsel’s Office to develop open source licensing recommendations quickly revealed gaps between MGB’s licensing provisions and the most popular OSI approved permissive open source Licenses. This talk is an overview of the MGB Open Science License Library which aims to balance clinical research risk mitigation with fostering the range of modern open science activities and outcomes.
It Is Time: Succession planning, Shepherding, and Sunsetting Open Source Projects
3:05pm - Room 310
Presented by Lucinda Wade
In the world of Open Source there are plenty of talks, resources, and guides about where to begin, and how to grow your project. But what happens at the end? How do you know when It Is Time? And how do you say goodbye with compassion and dignity?
In this talk, members of the Open Source Program Office at the Digital Service at CMS.gov will be sharing how they use repository metrics, maturity models, and archival checklists for succession planning, stewardship, and sunsetting of Open Source projects.
Translating Research to Practice Using Open Source
3:05pm - Room 311
Presented by Mike Sanders
There is a fundamental disconnect between public policy research and open source software development in the US. While our European counterparts embrace low cost, open solutions to fundamental policy problems, US governments continue to pay private companies for one off, bespoke solutions. Over the last two decades in public policy research Mike has come to learn that frequently the issue isn't that government want to get stuck in a loop of repeated development but that they lack the experience with open source solutions to build sufficient trust in the process to learn the skills necessary to implement open solutions.
With the rise of AI and the removal of the need to know how to code to create open source solutions, the cost of acquiring the skills to develop open solutions has never been lower. Why then have policy leaders continued to outsource even basic development tasks? In this session Mike will explore the barriers to adoption of open-source code at the state government level. Mike will provide a theory of change and initial evidence why this transition is slow to occur in the US and how top down action from the federal government might enable deeper penetration of civtech and other policy focused OS projects.
Scaling Open-Source in Public Health: Lessons from 20+ years Engineering and Implementing in Resource-Constrained Settings
3:25pm - Amphitheater
Presented by Jan Flowers
The University of Washington's Digital Initiatives Group (DIGI) (https://digi.uw.edu/) will present case studies and approaches from its 20+ years co-designing, engineering, and implementing open source digital health global public goods for and with resource-constrained countries. DIGI has architected and deployed these systems in 25+ countries, including the US, demonstrating that sustainable, scalable, high-quality and trusted health information systems are achievable without proprietary vendor lock-in. This talk will also include details on leading the Communities of Practice that sustain these digital health "Global Goods" and how the GW community can get involved.
Each case study will cover the open source software used, its impact,sustainability and governance models, and ethical AI opportunities, and perspectives on the future of software democratization. Case studies include: 1) Leading OpenMRS electronic medical records (https://openmrs.org/), used in over 80 countries supported by a community of practice with 5000+member 2) Stewarding the widely adopted OpenELIS Global laboratory information system scaled nationally across numerous countries, and leading OpenELIS U.S. serving U.S. State Public Health Laboratories. 3) Washington Department of Health projects including the Clinical Opioid Summary with Rx Integration app (https://project.cosri.app/) that helps clinicians determine opioid drug-seeking behavior by connecting medical records with Washington's State Prescription Drug Monitoring database, and MyHealth Summary where patients create customized health summaries from multiple data sources.
This session will demonstrate how open-source principles enable equitable, interoperable healthcare solutions across diverse contexts and offer practical strategies for collaborative digital health innovation.
Building for one, building for all
3:25pm - Room 310
Presented by Ray Bell
Effective IT governance requires balancing procurement with innovation. This presentation explores how the Department of IT’s Software Review Board collaborates with the AI Enablement Team to analyze software requests for 'build vs. buy' opportunities. We highlight a recent success story, 'youtube-to-docs,' where a request for a costly third-party tool was transformed into a secure, custom-built asset. Attendees will learn how internal development can reduce third-party risk, save taxpayer dollars, and rapidly meet business needs
Pixi: Fast, reproducible environments for Python, R and more without the headaches
3:25pm - Room 311
Presented by Daniel Kerchner
If you love the drama, frustration, and waiting time involved in fixing your Python/R/etc. environment so that you can run your analysis, then this talk is not for you. In this talk and live demo, I’ll introduce pixi.sh, a refreshingly simple way to build fast, reproducible environments without the usual dependency mess. We’ll see how pixi makes it easy to recreate results, share projects with collaborators, and move between different environments with confidence. If you care about reproducible research and prefer tools that stay out of your way, this quick tour of pixi might be where you meet your new favorite setup step.
March 24 Sessions
Three tracks in the amphitheater, room 310 and room 311.
Open-Source Auditing for Generative AI: Building Human-Centered Evaluation Infrastructure
1:00pm - Amphitheater
Presented by Monica Kodwani
Monica Kodwani will present an open-source audit tool designed to support human-centered evaluation of generative AI systems. As large language models become embedded in everyday workflows, there is growing need for transparent, reproducible, and privacy-aware methods to evaluate their behavior beyond benchmark accuracy.
This talk introduces an open-source auditing platform that enables researchers and practitioners to design structured audit workflows, interact with AI systems through configurable prompts, and systematically document model behaviors such as bias, factual errors, refusals, and subjective responses. The tool is modular by design, supporting multiple AI backends and allowing audit protocols to be shared, reused, and extended as open research artifacts.
The talk will cover the motivation behind the project, core architectural decisions, and lessons learned from deploying the tool in academic research studies. Monica will also discuss how open-source infrastructure can lower barriers to AI auditing, support open science practices, and empower students and researchers to critically examine AI systems without requiring access to proprietary evaluation pipelines.
Attendees will leave with a concrete understanding of how open-source tools can operationalize responsible AI principles and how they can contribute to or build upon this work in their own research and projects.
Commit to Community: Open Source as Social Infrastructure in Volunteer Civic Tech
1:00pm - Room 310
Presented by Kristijan Armeni
In volunteer-driven projects, human bandwidth is the primary constraint. When commitment is fleeting, traditional open source engineering practices come with dual benefit: they help scaling up communities, not just code. In this talk, we share lessons learned in technical leadership from the first year of the CIB Mango Tree project, a fully volunteer civic tech open source project for detecting inauthentic behavior in social media. By looking at practices like regular release cycles, continuous integration, and dependency selection, we'll unpack how technical decisions drive momentum across diverse team members — from new volunteers to project managers and outreach leads. A consistent pattern emerges: A regular release cycle streamlines code distribution, but also boosts the visibility of ongoing volunteers who see their contributions ship even when they can't commit long-term. Similarly, choosing right-sized code dependencies is about balancing technical complexity and performance, but equally about facilitating delivery such that outreach teams and community managers can work with tangible progress. Given the importance of contributor participation in volunteer projects, this perspective centers on individual and team throughput but sidesteps other aspects, like maintainer bandwidth. This is an unresolved trade-off that will only become sharper with the increased adoption of AI-assisted programming. We close with practical advice to aspiring open source contributors and maintainers that manage bandwidth-limited volunteer communities.
Finding legislative data
1:00pm - Room 311
Presented by Daniel Schuman
Open-source developers, journalists, and civic technologists increasingly rely on legislative data to build tools that improve public understanding of Congress—but finding that data remains far harder than it should be. In this talk, Daniel Schuman will present a practical, field-tested guide to locating, evaluating, and using data about the U.S. legislative branch for open-source projects.
In partnership with the House Digital Service, Schuman has compiled a comprehensive map of more than 130 official and non-official sources of congressional data, ranging from bills, votes, and committee records to staffing, budgets, oversight materials, and historical archives. These sources are organized both by subject matter and by the mechanisms required to access them, including APIs, bulk downloads, web scraping, records requests, and direct institutional channels.
The session will walk attendees through the most important and high-value data sources, explain how they fit together, and provide a practical framework for discovering additional datasets as new project needs arise. Participants will leave with a clear roadmap for sourcing reliable, authoritative legislative data for their own open-source work.
Open-Source GenAI for Government Forms, Referrals, and Caseworker Workflows
1:20pm - Amphitheater
Presented by Cory Trimm
Social services and government benefit programs often involve navigating complex forms and referral processes. Accessing these resources can require significant time, labor, and effort from staff and clients. This talk describes how Nava Public Benefit Corporation’s philanthropically-funded Nava Labs team prototyped and piloted two open-source, GenAI-powered tools — the Form-Filling Assistant and the Referral Generator — to address these problems.
The Form-Filling Assistant is an agentic tool that autonomously navigates multiple databases and benefit portals and then fills out application forms; caseworkers can then review the filled applications, address any missing or incorrect information, and submit. The Referral Generator optimizes case-management workflows by recommending suitable resources and step-by-step action plans to help address clients’ health and social needs. Guided by principles of human-centered design, both tools are built to enhance caseworker efficiency, reduce administrative burden, and improve client outcomes. The tools are being piloted with caseworkers in California, Texas, and Pennsylvania.
The talk will feature live demos and summarize the technical architecture behind these systems, including retrieval-augmented generation, MCP tool calling, and browser streaming. It will also share preliminary results on caseworker impact and lessons learned from implementing these tools in real public-service contexts with strict requirements around privacy and security.
Attendees will take away key lessons for building and deploying these and other open-source GenAI solutions.
Community engagement strategies for Open Source Communities
1:20pm - Room 310
Presented by Abigail Mesrenyame Dogbe
Strong open source communities don’t happen by accident, they are built through intentional, inclusive, and sustainable engagement strategies. This talk explores practical approaches to community engagement in open source, drawing from real-world experiences across global Python and open source communities. It will cover how communities onboard new contributors, support different contribution pathways beyond code, encourage belonging, and sustain participation over time. Attendees will learn what has worked, what hasn’t, and how factors like governance, communication, events, and recognition shape contributor experiences. Overall, this session offers actionable insights for building healthier, more resilient open source ecosystems.
Civic technology for the Web. Use cases of DC311, WMATA trains and Congress API
1:20pm - Room 311
Presented by Pedro Vicente
Use cases of DC311, WMATA trains annd Congress API
1) This is a C++ application that fetches congressional data from the official Congress API using HTTPS requests. It adds a C++ client to the API: https://github.com/LibraryOfCongress/api.congress.gov
The application retrieves information about members of Congress, including their biographical details, party affiliation, and terms of service, then saves the results to JSON files. It also includes a web interface built with the Wt framework that allows users to view the retrieved congressional data in a browser.
2) A web-based mapping application built with the Wt C++ framework that visualizes Washington DC's 311 non-emergency service requests. The application loads DC311 service request data from a SQLite database and renders them as interactive points on a map, allowing users to filter and explore city service patterns across Washington DC's wards.
Source code available at: https://github.com/pedro-vicente/wt.extra
The service is available online at: https://pedro-vicente.net:9444/
3) A real-time web application for tracking Washington DC Metro Red Line trains with interpolated positioning between stations. It uses a C++ client for the Washington Metropolitan Area Transit Authority API. https://developer.wmata.com/apis
Source code available at: https://github.com/pedro-vicente/wt.wmata
The service is available online at: https://pedro-vicente.net:9446/
The State of Open Source AI: Exploring Trends in AI Policy and Model Releases
1:40pm - Amphitheater
Presented by Gabriel Toscano
The October 2024 release of the Open Source AI Definition (OSAID 1.0) by the Open Source Initiative (OSI) sought to anchor the term “Open Source AI” in clear, unambiguous standards. Yet openness in AI remains nascent and inconsistently applied. Open source software has long relied on licenses to communicate permissions and restrictions, but AI systems differ because source code represents only one component. The OSAID identifies additional essential elements, including access to model weights and either training data or meaningful documentation of its provenance. At present, developers release these components inconsistently, and no common mechanism exists for disclosing weights or data. Meanwhile, many organizations market systems as “open” while withholding components or imposing significant use limitations through restrictive software licenses.
This talk summarizes findings from a research project examining how the OSAID 1.0 operates in practice amid evolving technical and policy environments. Through quantitative and qualitative analysis, this project delineates key trends in Open Source AI, industry norms surrounding AI model releases, and emerging regulatory efforts.
Clear definitions are essential for legal and policy decision-making concerning complex technologies. By grounding Open Source AI in transparent, verifiable, standardized criteria, this work supports more consistent interpretation across legal, policy, and technical domains. The project provides evidence-based recommendations for engagement with policymakers and contributes to ongoing efforts to refine the OSAID beyond its initial release.
Barriers to Open Source Leadership
1:40pm - Room 310
Presented by Shauna Gordon-McKeon
Many hands make light work, but open source projects seldom have the hands they need. The pathway from newcomer to maintainer is full of barriers that lead to drop out—and burn out for the overworked people that remain.
While newcomer experience, the first step on the pathway from newcomer to maintainer, has received some attention, there is very little guidance on how to help someone fully integrate into the community and take on leadership roles. This talk provides a brief survey of barriers "further down the pathway", as well as insights on how to address those barriers.
Barriers will be grouped into four domains: technical, cultural, financial, and governance.
Technical barriers include building competencies in underlying languages, frameworks and subject domains, as well as learning the design decisions behind the specific projects. Potential solutions include targeted recruitment, group study, and 'context-building tasks.
Cultural barriers include difficulty in navigating conflicting goals and needs (for example, a contributor who needs mentorship and a maintainer who needs a break); potential solutions include adopting needs-focused language and roadmaps.
Financial barriers such as low/no pay for contributors, and governance barriers such as a lack of clear decision-making, often overlap. Many projects lack the internal structure that would motivate organizations with a financial stake in the project to invest. Open source does not often select for, nor does it teach, governance or business skills. But by working intentionally, we can bring these skills into our projects and communities.
Making Data Ethics Actionable
1:40pm - Room 311
Presented by Jay Qi
Anyone doing data science, machine learning, or artificial intelligence work should be intentional about ethics—especially in open source software and open data, where formal accountability mechanisms may be limited. While important work is happening to develop governance structures and policies, translating ethical principles into everyday development practice can be challenging for individuals and teams.
In this talk, we’ll focus on how to make data ethics practical and actionable for developers. We’ll introduce a checklist-based approach that helps teams systematically reflect on ethical risks and tradeoffs as part of their normal workflows.
We’ll demonstrate how to integrate an ethics checklist into existing data science and software development processes using deon (deon.drivendata.org), a lightweight, open-source command-line tool. Through real-world case studies, you’ll see how simple, structured prompts can surface ethical concerns early—before they become costly or harmful.
Attendees will leave with concrete techniques for incorporating ethical intentionality into open source data projects, along with a practical tool they can immediately adopt or adapt in their own work.