STAAR rescoring requests raise questions about test accuracy and accountability

Texas school districts are increasingly asking the state to take a second look at students’ STAAR reading scores, raising questions about how written responses are evaluated, how much confidence schools and families can place in the results, and the fairness of those scores’ impact on school and district A-F accountability ratings.

The Texas Education Agency (TEA) received 64,489 requests to rescore open-ended responses on reading assessments this spring, nearly three times the 21,620 requests submitted the previous year. The state corrected more than 13,300 scores following those reviews, with human scorers handling every rescore. The figures point to two conclusions: One, districts need to closely examine results that they believe may not accurately reflect student performance and challenge those results if necessary; and two, the state should reflect on the use of AI to score these responses and evaluate if the model is accurate and successful.

Texas began using automated scoring technology for open-ended STAAR responses in December 2023. The system was introduced as part of the state’s redesigned assessments, which include more written responses and fewer multiple-choice questions. Under the current process, humans manually review a portion of scores initially assigned by the automated system. 

The growing number of rescoring requests suggest that districts are paying closer attention to how the technology evaluates student work. More than 250 school districts and charter networks requested reviews this year, and nearly 90% had at least one reading exam receive a higher score following review.

The corrections can have consequences beyond an individual student’s test results. STAAR performance is the primary input into the state’s A-F accountability ratings, which determine whether districts face state intervention. For students, a corrected score could also affect whether they meet a required performance level or must complete additional instructional support and tutoring required for students who are unsuccessful on the test.

Although TEA automatically reviews certain extended written responses when a student is one point away from the next performance level, districts may request additional reviews. Those requests come with a financial cost: Districts must pay $50 per test if a requested rescore does not change the score. That creates another challenge for school systems already working within limited budgets. District leaders must weigh which scores warrant further review against the cost of requesting it.

TEA has defended its hybrid scoring process, emphasizing that automated scoring is different from generative artificial intelligence and that human eyes and grading remain part of the system. The agency has also cited the expense and delays associated with having people score every written response, and Commissioner Morath has defended the model as a fiscally responsible use of taxpayer dollars.

While fiscal responsibility is important, it should not come at the expense of confidence in student assessments, particularly when the stakes of those assessments are so high. When districts identify potential scoring errors, they need a reliable and accessible way to have those concerns addressed without diverting limited resources from other educational priorities. 

The number of corrected scores also underscores why Texas’ accountability system should be evaluated by not only how quickly it produces results, but by how consistently and accurately the tests measure what students know and can do. Even when the overall percentage of responses requiring correction is small, the impact on an individual student or campus can be significant.

STAAR results carry considerable weight in Texas’ public education system. They help shape public perceptions of schools and contribute to accountability decisions that can lead to campus closures or state takeovers of districts. They also determine whether a student can be promoted or graduate. Educators and school administrators also bring valuable knowledge of student performance to the process. Their concerns about a score should be taken seriously, particularly when a score appears inconsistent with a student’s demonstrated abilities or written work. Accuracy and transparency are essential with stakes as high as these.

Technology may help the state process millions of assessments more efficiently, but test results must be a dependable measure of student learning, not simply a product of a faster scoring system. Texas should ensure that the tools used to evaluate students, as well as the accountability decisions based on those results, are accurate and dependable as well.