Abstract
One of the main reasons MCQ assessments are popular is that tests using text understanding is difficult and often erroneous. The emergence of large language models such as ChatGPT is not mature enough to help grading engineering assignments yet. Since MCQ tests are not well suited for summative assessment, scaling up eLearning for large number of students is difficult using non-MCQ tests. In this paper, we introduce a new eAssessment tool for database design courses that uses graphical conversations to understand learner’s mental model of cognitive state. We show that the model is capable of substituting NLP for authentic assessment for eLearning.