The state of Texas is implementing artificial intelligence (AI) to score the written portion of the Texas Assessment of Academic Readiness (STAAR) exam, aiming to streamline the grading process and potentially save the state millions of dollars each year. The revised exam, launching in 2023, emphasizes open-ended questions rather than multiple-choice questions, resulting in more structured response items.
The transition to AI grading is motivated by a cost-saving measure, with the Texas Education Agency (TEA) estimating potential savings of up to $20 million per year compared to hiring human graders from third-party contractors. While machines have not completely replaced human scorers, the number of temporary scorers required has been significantly reduced, from about 6,000 to less than 2,000.
The AI scoring engine uses natural language processing to evaluate responses and was trained on a dataset of 3,000 previously rated responses. However, some responses (about a quarter) will still be re-evaluated by human raters, particularly those deemed challenging for AI, such as non-English responses or responses containing slang.
Chris Rozunick, director of assessment development at TEA, emphasizing that AI systems are not autonomous and do not “learn” from every response. Instead, it relies on initial training to provide consistent ratings similar to humans. Despite the integration of artificial intelligence, strict quality control processes remain in place to ensure accuracy and fairness of scoring.
The shift to AI grading marks a significant change in the way assessment is conducted, emphasizing the benefits of technology in educational assessment while still recognizing the important role of human oversight and expertise in the process.
Submitted under . Read more about AI (artificial intelligence).