The Texas Education Agency will replace thousands of human scorers with new AI for the STAAR exams this week.
According to a report by The Texas Tribune, the TEA expects the new AI to save about $15-20 million a year by reducing the number of temporary graders it hires. In 2023, the TEA hired 6,000 test scorers, while this year it will only need less than 2,000 scorers.
The TEA described its model as an “automated scoring engine” that would use natural language processing, the same technology used by chatbots like OpenAI’s ChatGPT or Google Gemini, though the agency says its model will not act as a system capable of learning.
The system has been trained on 2,000 exam answers already graded by a human, and is designed to mimic a human grader. The model will focus on grading open-ended responses.
As a safeguard, a quarter of all computer-graded scores will be rescored by a human. The TEA said some responses will confuse the system, such as responses with non-English answers or excessive use of slang.
Some teachers have expressed concern about using AI to grade exams. When testing the system, Lewisville Independent School District Superintendent Lori Rapp said her district saw a “drastic increase” in answers receiving a zero from the automated system.
Kevin Brow, the executive director for the Texas Association of School Administrators, said that human graders give students credit for originality in their writing, and that the system could force students to abandon their originality.
