STAAR Rescoring Requests Triple After More Texas Schools Question Computer Grading

TL;DR

For listenersOffer from Amazon

Turn the school run and nap time into listening time

• Thousands of audiobooks, podcasts and originals
• Listen on your phone, tablet or Echo — also offline
• Cancel anytime

Try Audible free →Free trial for new members

As an affiliate, we earn on qualifying purchases.Hide

Requests to rescore open-ended responses on Texas STAAR reading exams rose from 21,620 last year to 64,489 this spring, according to state data reported by The 74. Human reviews raised more than 13,300 scores, although the state said fewer than 0.5% of all open-ended reading responses changed.

Texas schools submitted 64,489 requests this spring to rescore open-ended responses on STAAR reading exams, nearly three times the 21,620 requests recorded last year, according to Texas Education Agency data reported by The 74. Human reviewers raised more than 13,300 reading scores, as districts questioned results from a system that uses automated scoring alongside human review.

The number of requests was 2,758 in spring 2024, rose to 21,620 the following year and reached 64,489 this spring. The state corrected roughly twice as many scores as in the previous year. Of the more than 13,300 reading scores that increased after review, 58% had initially been graded by a computer and 42% by a human. Humans conducted all rescoring reviews.

More than 250 school districts and public charter networks requested reviews, including Cypress-Fairbanks, Dallas, Katy and Northside. Nearly 90% of participating districts and networks had at least one submitted reading exam score increase, agency data cited by The 74 showed. The increase rate among applicants does not mean that most STAAR responses changed: the agency said fewer than 0.5% of all open-ended responses on reading exams received a score change.

More than 3.2 million students took the spring reading assessments, covering grades three through eight and high school English I and English II exams. The agency automatically rescored extended written responses that fell one point short of the next performance level at no cost. Some districts said that process let them use their own requests for students who were two points away.

Why Score Reviews Matter

STAAR results affect more than an individual test record. Students’ results can determine whether they meet a grade-level performance category, face remediation or meet requirements relevant to graduation. A small score adjustment near a cutoff can change a student’s reported performance level, though the source does not specify how many of the 13,300 increases moved students into a higher category.

The results also feed into school accountability ratings. Texas campuses and districts receive A–F ratings based in part on how many students reach specified STAAR performance levels. State law allows the education commissioner to intervene in campuses that underperform over multiple years, including appointing a conservator or replacing a locally elected school board. The rescoring figures do not establish that any particular rating or intervention changed, but they show why districts may scrutinize scores closely.

The growing number of requests also puts attention on whether educators trust the scoring process. University of Texas at Austin professor Rachel White told The 74 that the requests suggest teachers who know their students’ writing are increasingly disagreeing with assigned scores. That is her interpretation of the trend; the request count alone does not show why each district sought a review or whether the initial score was wrong.

How Texas Scores Written Responses

The Texas Education Agency began using an automated scoring engine in December 2023 for open-ended STAAR questions, including reading and English assessments. The change followed a redesign of the test that reduced multiple-choice questions and added more written responses. To develop the system, the agency gathered responses previously scored by people and programmed the engine to assign scores based on characteristics in those examples.

The agency describes the arrangement as a hybrid scoring model: human experts contribute to the process, and humans manually review a quarter of exam scores, according to the source report. The agency has stressed that its scoring engine is not generative AI such as ChatGPT or Google Gemini and does not learn from one response to the next.

The automated approach drew criticism after a rise in zero scores on writing prompts during the first exam cycle scored by computers. State officials pointed to other possible factors in the low scores, including students retaking the exam. The available reporting does not resolve how much any factor contributed to those results.

“That traction of superintendents talking to each other — (and) media continuing to talk about it — made more districts aware that not only can they do this process, but actually they should, and they need to.”

— Allison Matney, director of the Texas Center for School Accountability

Questions Behind the Requests

The published totals do not explain why each district requested a review, which responses were disputed or how many students moved to a higher performance category after a score change. The fact that 58% of the increased scores had originally been computer-graded does not, on its own, show that computer scoring caused more errors: the source does not give the share of all responses initially scored by each method in a way that would establish comparative error rates.

It is also unclear how much of the jump reflects more awareness of the process, a greater number of disputed scores, district capacity to submit requests, or other factors. The article reports Matney’s view that awareness spread among superintendents and White’s concern about trust; those are attributed explanations, not a determination by the state. The source material does not provide a full breakdown of request outcomes by district, test or scoring method.

Districts Await Score Outcomes

Schools and families can use the results of human reviews to understand whether particular scores changed and whether a student’s performance level was affected. The state’s reported totals cover the spring rescoring cycle; the source material does not give a date for another review window or say whether Texas plans to change its scoring system.

Further reporting or agency data would be needed to show how many students crossed performance cutoffs, how districts’ requests were distributed, and whether the next assessment cycle brings another rise in appeals. For now, the confirmed picture is a sharp increase in requests and more than 13,300 raised scores, alongside the agency’s statement that changes affected less than 0.5% of open-ended reading responses overall.

Key Questions

How many STAAR rescoring requests did Texas receive?

The state received 64,489 requests this spring to review open-ended responses on reading exams, up from 21,620 last year and 2,758 in spring 2024.

How many scores changed after review?

Human reviewers raised more than 13,300 reading scores. The Texas Education Agency said fewer than 0.5% of all open-ended responses on reading exams received changes.

Were the responses rescored by computers?

No. According to the report, humans conducted all rescoring reviews. The initial scoring process uses an automated engine for some open-ended responses alongside human review.

Why could a score change matter to a student?

A change may affect whether a student reaches one of the STAAR performance levels. Depending on the student and assessment, that level can relate to remediation or graduation requirements; the source does not say how many rescored students crossed a cutoff.

Does the increase prove that computer grading is inaccurate?

No. The increase in requests and corrected scores does not establish the error rate of automated scoring. The source does not provide enough comparative data to determine whether computer-graded responses were more likely than human-graded responses to change on review.

Source: rss