New AI Model Can Predict Student Performance Using Educational Games

An artificial intelligence (AI) model can now better determine the degree of students’ learning using educational games. Researchers from North Carolina State University created a model that can project student achievement on a test based on students’ previous experience with learning tools like educational games. The researchers tested the effectiveness of their model by analyzing student behavior while student participants played an educational game called Crystal Island. In the game, participants act as medical agents engaging in different tasks to investigate a disease outbreak on an island. Before and after playing the game, the participants took a survey and a 17-item test. According to one of the researchers, the team wanted the AI model to be capable of predicting whether or not students would answer each item on the test correctly just by analyzing how they played Crystal Island. Using data from 181 students, the AI identified similar game-playing behaviors among students who answered each test item correctly and among those who answered erroneously. The model then used this data to predict how a new student would answer the same test item. According to the researchers, this AI model was about 10% more accurate in predicting student performance than other AI models. The researchers believe their model can be used to improve teaching and learning results. For instance, the model can alert teachers if students need supplementary lessons. Another researcher said the AI model can open the door to more sophisticated educational programs that can adjust to students’ needs.