The challenge
The client wanted to simulate how candidates of different proficiency levels would perform in a large-scale exam.
Prototyped AI-driven virtual students to simulate exam proficiency distributions for assessment research.
Ref: IMSFT113118
Video story · 1:01
The challenge
The client wanted to simulate how candidates of different proficiency levels would perform in a large-scale exam.
The approach
Kumulus integrated AI components into an application covering architecture, APIs, user experience and operational controls, using a virtual-student architecture parameterised to replicate proficiency distributions.
The outcome
Gave assessment researchers a way to calibrate questions before testing them with real candidates. Virtual student agents simulate exam performance across proficiency levels to estimate the difficulty and discrimination of ENEM items in a controlled digital environment.
Could this work for your organisation?
The AI Audit shows where data and AI will deliver measurable value for you first.