Demand for rescoring of students’ STAAR tests in Texas has surged dramatically, nearly tripling from last year. This significant increase reflects a growing trend among schools seeking review of written responses on reading exams, many of which were initially evaluated by computers.
The Texas Education Agency (TEA) implemented an automated scoring system in December 2023 to handle open-ended questions on the State of Texas Assessments of Academic Readiness (STAAR), which includes reading and English tests. While humans reviewed about 25% of the exam scores, the number of requests for rescoring open-ended responses on reading tests soared from 21,620 last year to 64,489 this spring, following a total of 2,758 requests in spring 2024.
Additionally, the number of score corrections doubled compared to the previous year, with TEA adjusting over 13,300 reading scores following reviews. Of these revised scores, 42% were initially graded by humans, while 58% were scored by computers, with all rescoring performed by human reviewers.
Allison Matney, director of the Texas Center for School Accountability, noted that it has traditionally been uncommon for district administrators to seek STAAR rescoring. She attributed the recent uptick in requests to increased awareness about the automated scoring and review processes among school officials.
“The conversations among superintendents and ongoing media coverage have made more districts realize that they can and should engage in this process,” Matney explained.
Despite initial criticism of the automated scoring system—especially after a notable rise in students receiving zeros on writing prompts in the first exams graded by computers—TEA officials attributed the low scores to various factors, including many students retaking the exam.
TEA has emphasized that its automated scoring system is distinct from generative artificial intelligence models like ChatGPT or Google’s Gemini, as it does not adapt or learn from individual responses.
Agency spokesperson Jake Kobersky highlighted the financial implications of hand-scoring essays, which would require four to five times the number of human scorers and incur an additional annual cost of $15 million to $20 million, leading to delayed results for districts. “TEA employs a hybrid model that prioritizes human judgment in scoring,” Kobersky stated, adding that this approach ensures quicker results for families and districts.
This spring, over 3.2 million Texas students, including those from third to eighth grades and high school students taking English I and II assessments, participated in the reading STAAR. Remarkably, less than 0.5% of total open-ended responses saw score changes.
TEA began using automated scoring after redesigning the STAAR to feature fewer multiple-choice questions and more open-ended questions, or constructed responses. The agency developed the scoring system by analyzing thousands of previously human-scored responses, enabling the program to replicate human grading standards.
This year, over 250 public charter networks and school districts, including Cypress-Fairbanks, Dallas, Katy, and Northside, submitted rescoring requests. Nearly 90% of these requests resulted in at least one reading exam score increase, according to agency data.
Rachel White, an associate professor focused on educational leadership and policy at the University of Texas at Austin, expressed concern over the necessity for school administrators to invest time and resources in scrutinizing test results and filing thousands of rescoring requests. “The tripling of requests indicates that educators who review their students’ essays are increasingly questioning the accuracy of the assigned scores,” White remarked, suggesting a broader issue of trust in the scoring process.
Public campuses and districts receive A-F accountability ratings based on student performance levels on the STAAR. Underperforming campuses that fail to meet expectations over multiple years may face actions from Texas education commissioner Mike Morath, including the appointment of a conservator or replacement of the local school board.
“If an instructional leader believes a test was scored inaccurately, even a couple of points can mean the difference between failing and passing,” Matney noted, highlighting the potential impact on students’ educational paths.
TEA provides automatic rescoring for extended constructed student responses at no cost if the exam scores are within one point of qualifying for the next performance level—ranging from did not meet to mastered. Some districts reported that this automated rescoring process allowed them to allocate funds more effectively and submit additional students for rescoring.
Officials from Dallas Independent School District (DISD) indicated that they submitted rescoring requests because they believe students benefit from human evaluation of their answers. The district submitted over 5,100 reading tests for rescoring, with 34% yielding higher scores.
DISD officials estimated that the combination of automatic rescoring and district-requested rescoring helped five campuses improve their overall ratings before public disclosure. Notably, one campus elevated its rating from a D to a C, while four others advanced from C to B.
School districts incur a fee of $50 per test for rescoring requests that do not result in score changes, although rescoring is free if a score adjustment occurs. North East ISD spokesperson Aubrey Chancellor mentioned that the district opted not to submit requests after reviewing multiple students’ scores, concluding that it would not benefit overall scores or campus ratings.
The TEA’s rescoring process was completed prior to the release of A-F accountability ratings in August.
