A STAAR Rescore Is Free If
the Score Changes. If Not, $50.

In TEA’s December 2023 STAAR document, the rescore fee is waived if the score changes. If not, the Texas Tribune reported October 7, 2026, it is $50 per test.

By William Douglas Baker

Founding Editor

One Texas Education Agency (TEA) document from December 2023 describes how a written answer on the English-language State of Texas Assessments of Academic Readiness (STAAR) is scored, and when a rescore fee is waived. Step 3 of the document, “Scoring Process for STAAR Constructed Responses,” says of TEA’s automated scoring engine (ASE): “All student responses are scored for the first time by the ASE.” Step 5 says district testing personnel can request a rescore “for a fee,” and that “If the score on the constructed-response question changes, the fee is waived.” This page reads the version TEA links on its STAAR page on October 23, 2026.

In TEA’s December 2023 document, Step 3’s engine-first scoring is for the English-language version of STAAR; for Spanish-language assessments, the document says, student responses to constructed-response questions are “100 percent human scored and do not get scored by the ASE.” Step 3 also names condition codes, which the ASE assigns, and gives one example: responses that “use words not seen in the responses used to train the ASE.”

The rescore, and its fee

In Step 5 of TEA’s December 2023 document, “district testing personnel can request that the student’s response be rescored for a fee. Rescore requests are sent for human scoring.” In the Texas Tribune’s words, “Humans conducted all re-scores.” As of October 23, 2026, TEA’s December 2023 document states no fee amount. The Tribune reported on October 7, 2026: “School districts must pay $50 per test if they request rescoring and the student’s score does not change. The rescore is free to districts if a score changes.” As of October 23, 2026, no source this edition read reconciles TEA’s “the student’s response” and “the score on the constructed-response question” with “$50 per test.” Education Service Center Region 13’s April 30, 2024 explainer, which is not TEA’s, says: “If the score improves, there is no fee. If there is no change in the score, the school is charged $50.” The service center writes “improves”; TEA and the Tribune write “changes.” This page keeps the words apart.

On what a rescore could mean, Allison Matney, director of the Texas Center for School Accountability, said, the Tribune reports: “For an individual student, that could mean anything from now being able to graduate to now (they) don’t have to do remediation hours in the next grade level.” Dallas ISD officials said, in a statement the Tribune reports, that the district submitted a rescoring request because they felt students benefit when human scorers review their answers. The Tribune also reports that the district submitted more than 5,100 reading tests for rescoring and that 34% returned with a higher score.

How an answer reaches a person

Step 3 describes a sample: “At least 25 percent of student responses for each grade/subject and course are then routed to trained human scorers,” and TEA’s December 2023 document says this set of second scores is used “to monitor the ASE scoring.” Step 2 of the document says the ASE “calculates confidence values that indicate the degree to which the ASE is confident the score it has assigned matches the score a human would assign,” and Step 3 says the “low confidence responses are often those responses that are on the border between two score points.” Step 3 adds that responses routed for human scoring “maintain the score assigned by the human scorer as the score of record.” On our reading, TEA’s December 2023 document names three routes by which an English-language STAAR written answer reaches a human scorer.

The first is a sample of at least 25 percent. The second is responses to which the engine assigns condition codes or that it identifies as “low confidence.” The third is a rescore that district testing personnel request for a fee, in Step 5.

The Tribune reports another rescore: TEA “automatically rescored extended constructed student responses at no cost if the exams were one point from qualifying the next performance level.” As of October 23, 2026, TEA’s December 2023 document describes no automatic rescore, and the Tribune’s article gives no administration, TEA document or count for it.

The slide that labels a share

In TEA’s December 2023 document, the path where the engine assigns the score is a box on the page-4 figure: “Responses assigned score by ASE–this becomes the score of record.” On slide 26 of “Hybrid Scoring Key Questions,” TEA’s March 2024 slides print “Approx. 75% responses” under the box “Scenario 1: Auto scoring engine assigns score of record.” One bracket, on slides 26 and 27, under the two human-scoring scenarios reads “At least 25% responses,” and slide 21 says: “All scores go through auto scoring engine; 25% are double human scored (one score of record, one for auditing).”

As of October 23, 2026, no source this edition read reports a measured count or share of STAAR scores of record by who set them, for any named administration. TEA’s December 2023 document prints a floor, “at least 25 percent,” for each grade, subject and course, in Step 3. The slides name no administration, count or denominator, do not mention Spanish-language responses, and do not say whether their figures are planned or measured.

The counts, and one split

The Texas Tribune reports that TEA began using an automated scoring engine in December 2023, and that “In spring 2024, the state had 2,758 requests to regrade open-ended response scores on reading exams. Then, 21,620 requests last year. This spring, that jumped to 64,489.” The article is dated October 7, 2026, so “last year” is 2025 and “this spring” is spring 2026. The Tribune reports that “The state improved more than 13,300 reading scores after the reviews.”

Of the more than 13,300 improved scores, the Tribune reports, “humans initially assigned grades to 42% while computers graded 58%.” The Tribune also reports that “less than .5% of total open-ended responses on reading exams saw changes.” As of October 23, 2026, that is the only split by scorer in the sources this edition read. None of them splits the 64,489 requests by scorer, says whether “initially assigned” means the score of record before the rescore, or gives the number of responses behind “less than .5%.” None reconciles the 42/58 with TEA’s December 2023 Step 3 sentence, “All student responses are scored for the first time by the ASE.” This page draws no conflict between them and does not reconcile them.

What TEA calls the engine

TEA’s December 2023 document says the ASE “is trained to emulate how humans would score student responses for each constructed-response question.” In a footnote on slide 28, TEA’s March 2024 slides say: “This kind of technology is different from AI in that AI is a computer using progressive learning algorithms to adapt, allowing the data to do the programming and essentially teaching itself.” State officials have stressed, the Tribune reports, that automated computer scoring is different from generative artificial intelligence, as the program does not “learn” from one response to the next.

The Tribune writes: “From the sample, the computer learned the characteristics of responses, and it was programmed to assign the same scores as a human.” Those are four descriptions. This page adds or chooses none.

Where the argument stops

As of October 23, 2026, TEA’s December 2023 document is the only version this edition located; nothing inspected dates a revised version or a 2026 process, so this page does not say how any 2026 administration was scored. TEA’s STAAR page links a file byte-identical to a September 1, 2025 archived capture. As of October 23, 2026, no source this edition read says what the fee pays for or who receives it. None says whether a rescore of a human-scored response goes to another scorer, or whether Spanish-language exams are inside the 64,489 or 13,300. None states a finding that any student’s original score was incorrect; the Tribune reports scores “improved” and, in its own voice, uses “corrections.”

The Agentics read, ours and not a finding: in the Inference Era, a system reads a professional’s work before any person does. Authority is not claimed. It is encoded. TEA’s December 2023 document lists ways a written answer reaches a person, including a rescore a district requests for a fee, waived if the score changes. Does anything in writing say as much about your work?