Inspiration

Essay assessment is an essential part of education, but evaluating written answers consistently can be challenging. Different teachers may interpret assessment criteria differently, leading to disagreements in scores and feedback.

We were inspired to build a platform that makes essay assessment more collaborative, transparent, and accountable without taking decision-making authority away from teachers.

What it does

ESAINA (Explainable Essay Assessment Platform) is a web-based platform designed to support collaborative essay assessment through AI-assisted evaluation.

The platform enables teachers to:

Create assessments, questions, and scoring rubrics.
Evaluate student essays collaboratively with multiple raters.
Identify scoring disagreements between evaluators.
Get AI-generated second opinions with supporting evidence and feedback.
Review AI recommendations and make the final assessment decision.
Track assessment decisions through an assessment history.

Our core principle is simple: AI assists teachers, but teachers remain in control.

How we built it

We designed ESAINA as a web application using Next.js and TypeScript, with Tailwind CSS and shadcn/ui for the user interface.

The backend is built around Next.js API routes, Prisma ORM, and PostgreSQL through Supabase.

For AI-assisted assessment, the platform integrates a Large Language Model (LLM) API to analyze student answers based on the provided questions, reference answers, and scoring rubrics. The AI generates structured assessment recommendations that teachers can review before making their final decisions.

Challenges we faced

One of our main challenges was designing an assessment workflow that balances AI assistance with human judgment.

We needed to ensure that AI recommendations were not treated as unquestionable results. Instead, the system was designed to help identify differences in evaluation, provide additional insights, and support teachers in making more consistent decisions.

Another challenge was integrating AI-generated responses into a structured assessment workflow while maintaining data consistency and usability.

What we learned

Through this project, we learned that building an educational AI solution is not just about integrating a language model. It also requires understanding real user workflows, designing meaningful human-AI interactions, and ensuring that technology supports rather than replaces human expertise.

What's next

We envision ESAINA becoming a more comprehensive assessment platform with improved evaluation analytics, teacher calibration, and more transparent feedback mechanisms to support better learning outcomes.