Texas introduces AI grading for standardized tests

Some educators have raised concerns about the accuracy of the new system.

Texas - USA

Texas students face a new era in standardised testing as the state rolls out an AI-powered scoring system to evaluate open-ended exam questions. The Texas Education Agency (TEA) is implementing an ‘automated scoring engine’ employing natural language processing technology akin to chatbots like OpenAI’s ChatGPT. With plans to replace a majority of human graders, TEA anticipates annual savings of $15–20 million, reducing the need for temporary scorers from 6,000 in 2023 to under 2,000 this year.

The State of Texas Assessments of Academic Readiness (STAAR) exams, revamped last year to include fewer multiple-choice questions, now feature up to seven times more open-ended inquiries. TEA’s director of student assessment, Jose Rios, cites the time-intensive nature of scoring these responses as a driving factor behind the shift. Despite initial training using 3,000 human-graded exam responses and implemented safety nets, including human rescoring for a quarter of computer-graded results and ambiguous AI-confounding answers, concerns linger among educators.

While TEA is optimistic about cost savings, some educators, like Lewisville Independent School District superintendent Lori Rapp, remain cautious. Rapp notes a ‘drastic increase’ in zero-scored constructed responses during the system’s limited trial, raising questions about test question integrity versus automated scoring accuracy. The move towards AI-driven grading aligns with a broader trend in education, with AI essay-scoring engines already in use across 21 states, albeit with mixed success. TEA emphasises distinctions between its ‘closed system’ scoring engine and broader AI, highlighting the importance of transparency and accountability in its implementation.

Why does it matter?

As Texas students navigate this new grading landscape, concerns about fairness and accountability emerge. With generative AI tools already raising issues of academic integrity and equity, questions arise about the consistency and impartiality of AI grading. As the rollout progresses, stakeholders will be watching closely to assess the impact of AI on standardised testing and its implications for education policy and practice.