TLDR: ONION is a five-stage framework (Observe, Nurture, Integrate, Optimize, Normalize) for creating Entity-Relationship (ER) data models collaboratively. It aims to reduce bias and increase inclusivity by involving diverse stakeholders from the start, transforming unstructured input into structured models through progressive abstraction. Evaluated through real-world workshops, ONION proved effective in fostering mutual understanding and capturing varied perspectives for richer data models.
In today’s data-driven world, the way we organize and understand information is crucial. From simple spreadsheets to complex artificial intelligence systems, everything relies on a foundational element: the data model. These models define how data is structured, what pieces of information are related, and how they interact. Traditionally, designing these data models has often been a task for technical experts, which can sometimes lead to models that don’t fully capture the diverse needs and experiences of the people they affect.
A new framework called ONION, developed by Viktoriia Makovska, George Fletcher, and Julia Stoyanovich, aims to change this. ONION is a multi-layered approach to designing Entity-Relationship (ER) models – a common way to visualize data structures – that brings together insights from design justice, participatory AI, and conceptual modeling. The core idea is to make the data modeling process more inclusive, transparent, and less prone to designer bias by actively involving all stakeholders from the very beginning.
The ONION Framework: A Five-Stage Journey
The name ONION is an acronym for its five distinct stages: Observe, Nurture, Integrate, Optimize, and Normalize. This methodology guides participants through a progressive abstraction process, moving from broad, unstructured ideas to a precise, structured ER diagram.
The journey begins with the Observe stage, where participants are encouraged to share their initial needs and objectives in an open, free-form manner. This is a “zero-assumption” phase, designed to gather a wide range of real-world understanding without any predefined structure.
Next is the Nurture stage, where collaborative modeling truly takes shape. Participants, guided by facilitators, use shared visualization tools and activities like storytelling and sketching to co-create narratives that reflect system needs. This stage bridges everyday language with conceptual ideas, allowing for free expression and discussion.
The Integrate stage is where technical experts step in. They synthesize the diverse input from participants, identifying key entities, properties, and relationships to construct an initial ER model. This marks the midpoint, formalizing the informal insights gathered earlier.
In the Optimize stage, the draft model is brought back to the participants for validation. Through feedback and internal review, adjustments are made collaboratively to ensure that the technical interpretation still aligns with the participants’ original intentions and ideas.
Finally, the Normalize stage translates the collaboratively built model into a standard ER model, ready for technical implementation. An external technical expert reviews the model to ensure its validity, completeness, and suitability for development.
Roles and Real-World Application
The ONION framework defines three key roles: Participants (stakeholders providing input), Experts (synthesizing models and conducting technical reviews), and Facilitators (guiding the process and ensuring contextual relevance). The framework was evaluated through real-world workshops focused on sociotechnical systems in Ukraine, including a blood donation support system (Donor UA) and a volunteer coordination system (Volunteers Lab).
These workshops demonstrated ONION’s effectiveness in capturing diverse perspectives and transforming them into structured data representations. Participants, even those without prior data modeling experience, found the process intuitive and felt empowered to express their ideas. The use of visualizations and story-based activities was particularly helpful in bridging conceptual gaps and ensuring that contributions were integrated into the final models.
Also Read:
- Advanced AI Framework Accelerates Rare Disease Diagnosis
- AI Agents Reshaping Conceptual Engineering Design with Structured Language Models
Insights and Future Directions
The practical investigations revealed that ONION successfully supports structured abstraction while maintaining stakeholder inclusion. Participants reported gaining new insights and empathy for other roles, and felt that the diverse input led to fairer and more robust models. The framework helped uncover hidden complexities and biases that might otherwise be overlooked in traditional design processes.
While highly promising, the ONION framework also faces practical challenges, such as managing large groups in online settings, handling overly complex conceptual models, and the need for highly skilled facilitators. Future work will focus on scaling the framework, improving tooling, enhancing facilitator training, and adapting the methodology to other modeling languages.
The ONION framework represents a significant step towards more inclusive and collaborative data modeling, ensuring that the systems we build truly reflect the needs and values of all those they impact. For more detailed information, you can read the full research paper: ONION: A Multi-Layered Framework for Participatory ER Design.
