Arlington ISD trustees were told Aug. 21 that a vendor error prevented the state from processing the district's June rescore submissions for AI-scored open-response (ECR) items, and that rescored human reviews, once completed, changed outcomes at dozens of campuses.

The rescore review was prompted by the district's decision to submit selected English open-response and STAR reading assessments for human rescoring. Doctor Natalie Lopez, who led the district update to the board, said Arlington submitted 4,370 assessments in June for human rescoring and was later informed by testing vendor Cambium that "our rescore results were not — didn't make it all the way through." She said TEA later expedited an investigation and rescoring but "these results are not going to make it in time for the August 15 rating released to the public."

The rescored files, Doctor Lopez said, produced measurable score increases: districtwide, 21% of submitted tests showed a raw- or scale-score increase and 14% resulted in a performance-level change. At the school level she reported that six campuses' letter grades would have risen on the Aug. 15 release (Owsley Junior High, Duff Elementary, Hill Elementary, Peach Elementary, Berry Elementary and, after further analysis, Arlington High). Districtwide she said 30% of campuses showed overall score growth in at least one accountability domain after the rescoring.

Why this matters: Texas school accountability and local perceptions of school quality rely heavily on STAR and EOC outcomes. Trustee Richardson warned the board that "when we read the news headlines ... that is what our community reads is Arlington is rated this many schools as an F," and said that incorrect public reporting harms families and property decisions. Trustee Fowler said she was "angry" that the AI grading error and the state's process could require districts to pay to correct the mistake: "They're charging us to go back and change what they messed up on."

What happened and next steps: Doctor Lopez described the district's selection process for rescoring (prioritizing papers closest to a higher performance level) and said the district's submissions included STAR EOC English I and II (1,131 tests) and STAR reading grades 3 (3,239 tests). Cambium initially showed the files as processed in its upload history but later told the district the files had not reached the appeals/rescore queue. Doctor Lopez said TEA's Kevin Mallendrokala, director of assessment, "was very responsive" and told the district, "This is really a very bizarre and unfortunate thing that has happened to Arlington ISD, to the campuses, and to the students." TEA expedited rescoring, but the agency told the district those corrected scores would not appear in the public August 15 accountability release; instead, Doctor Lopez said, the district will pursue appeals to update the published ratings.

District direction: Superintendent Smith told trustees the district is assembling detailed documentation for appeals. He said the district will file appeals for the campuses Doctor Lopez identified and set a target appeal submission date aligned to state guidance; Doctor Smith told the board the appeal submission deadline the district is preparing for is Sept. 12. "Right now, Natalie and her team are working day in and day out and making sure we have all the appropriate data," he said.

Board reaction and context: Trustees expressed frustration at the role of AI in scoring. Trustee McMorrow asked whether the miscoring affected graduation; Doctor Lopez said no — the district received corrected results before summer graduations and the main concern is accurate accountability reporting. Multiple trustees said the situation underlined broader concerns with the state accountability system and the practical hurdles districts face when attempting to request rescoring: short turnaround windows, the cost of mass human rescoring and limited staff capacity. Trustee Mike observed that only a fraction of the district's total assessments were submitted and asked rhetorically what would happen if the district could afford to resubmit more tests.

What the rescoring changed (district summaries provided to the board): Doctor Lopez gave the district totals and examples: districtwide 21% of the 4,370 tests submitted improved; 14% led to a performance-level increase. She said Owsley Junior High would have moved from a B to an A; Duff Elementary from a B to an A; Hill Elementary from a C to a B; Peach Elementary from a D to a C; Berry Elementary from a forced F to a D. She also said further analysis during the week indicated Arlington High School would have had a letter-grade increase as well.

What trustees asked staff to do: Trustees asked staff to proceed with appeals and to continue documenting the errors and their effects on campus ratings for legislators and the public. Doctor Smith said he had spoken with the TEA commissioner and that TEA did respond to expedite rescoring once the problem was identified, but trustees urged legislative and systemic fixes to how scoring and appeals are handled.

The board took no formal vote on rescoring at the Aug. 21 meeting; staff were directed to compile and submit appeals and supporting documentation to TEA.