Digital Assessment Design - Principles and Directions

After all these years working on digital assessment, it is time to take stock of the fundamental principles that guide this approach. I propose to initiate an update of the design principles, particularly in the era of artificial intelligence. I call on your experience and contributions to enrich this approach and better define the scope of the field.

But before anything else, I would like to anchor this reflection in the project of a forthcoming versatile, free, and open-source platform, as we have envisioned for years. Let’s consider the immense revolution brought by AI, which now makes it possible to achieve projects that would have once required unrealistic development costs (cf. article 14/01/2025).

Here is a first draft of the fundamental principles guiding such a project. If certain elements of this list seem debatable, so is their order:

1. Versatility of Uses

Digital assessment is a vast field. The multiplication of small applications specialized in specific uses is a hindrance to its expansion. A modern platform must offer extensive possibilities: beyond a secure exam mode, it should also include a learning mode—focused on students and self-assessment, and a live mode—similar to classroom engagement tools like Kahoot. The platform must be capable of providing adaptive or conditional assessments, and even AI-driven assessment agents.

2. Scalability

The platform must integrate new functionalities through object interfaces, support themes (CSS-based appearances), and plugins (new interactions, additional features). Scalability is essential to ensure that the tool adapts to users, rather than forcing users to adapt to the tool.

3. Security

Security must cover tests, items, user data, and results. It is essential to establish a web exposure policy for high-stakes assessments and to offer local test deployments. Access to content—from creation to test-taking—must be protected. The platform must also ensure compliance with GDPR regulations.

4. Efficiency

Streamlined deployment using Docker-based solutions for easy installation and management. Smooth user interface for designing assessments. Optimized test execution speed, with minimal response times and reactive servers. Cloud-native infrastructure with auto-scaling capabilities (e.g., Kubernetes, serverless computing).

A Node.js-based backend to unify backend and frontend development using JavaScript.

5. Modularity

A modular design makes it easier to maintain the code and develop it as sub-projects. The chosen stack—Node + Express + MongoDB + Vue3 + Vite + Pinia—follows a highly efficient modular approach, which I have personally tested and validated.

6. Compatibility

Adopt JSON as the core compatibility format for legacy formats such as XML, QTI, PCI. AI can now easily generate conversion routines to ensure seamless integration.

7. Portability

A modern platform should be integrable into existing technical environments (e.g., LTI, SCORM). For LTI integration, there are already available Express.js modules: ltijs, JSON Web Tokens (JWT), OpenID Authentication

8. Community-Driven Development

The project must rely on user communities at multiple levels: Teachers and students to share best practices and improve the toolset. Developers to ensure widespread adoption of technologies, based on open-source standards and long-term sustainability.

9. Accessibility & Inclusion

Ensure compatibility with screen readers. Allow customization of display settings (contrast, font size, speech synthesis, etc.). Adapt to specific learning needs (extra time, alternative formats). Use accessibility evaluation tools to assess compliance with standards.

10. Integration of Artificial Intelligence

AI can automate grading and evaluation: Open-ended questions, oral responses, handwritten answers, and even graphical items (e.g., modified diagrams). Until now, manual grading platforms were required for evaluation (I personally developed one for the DEPP a few years ago).

AI can assist in test generation, analyze student performance, and generate reports, such as: Student Reports, Teacher Reports (for one or multiple classes), Psychometric Studies

11. Open Research & Open Data

Encourage scientific and collaborative approaches: Provide APIs for anonymous data export to improve assessment methodologies.

12. Comprehensive and Dynamic Documentation

With AI-assisted documentation generation, detailed technical explanations are no longer a barrier to adoption.

13. Sustainability

The project should be maintained within an institutional framework, ensuring community-driven decision-making through a democratic process.

14. Transparent Governance

A project of this scope should be built on trustworthy individuals and institutions that support free and open education. The licensing model should: Promote free and unrestricted use for public institutions. Require a fair contribution from private-sector partners and for-profit organizations.

Voilà!

This is just an initial draft, and I hope this approach sparks interest and further discussion on how to integrate modern technologies into the educational ecosystem effectively.