Texas administration implements automatic correction for mandatory school assessments

At the end of the year, Texas schools will use computers to correct certain papers for students aged 8 to 14. The strategy would save money. But how will this work?

These computers will correct the STAAR test, the American initials of what is called the Assessment of Academic Readiness in the State of Texas. It is a standardized exam that all students take each year from the equivalent of CE2 to 4th grade. It makes it possible to gauge, as its name indicates, the overall level of a student in reading, writing, sciences and humanities.

Up to 4,000 proofreader positions saved

Until last year, this test consisted mainly of multiple choice questionnaires. But the new version involves more open responses with constructed text, which is more difficult to correct automatically. Now the technology that powers the digital proofreader works a bit like ChatGPT. The artificial intelligence was trained with 3,000 responses rated by humans. She analyzed these answers and the notes that accompanied them, to learn how to note like a human. According to the Texas administration, humans will check about a quarter of the answers scored by the computer. Knowing that if the computer is not sure of its mark or does not understand the answer – if the student has written in slang for example, a human will examine the copy in turn.

With this system, the Texas administration hopes to save 15 to 20 million dollars per year. Until now, to ensure the correction of these STAAR tests, Texas hired approximately 6,000 markers. Artificial intelligence should therefore greatly reduce human needs this year. Texas plans to hire more than 2,000 people.

An educational community taken by surprise by the initiative

THE Texas Tribune, the first media to report on this affair, reports that the announcement surprised many members of the teaching staff and parents of students. There is also skepticism about the ability of artificial intelligence to do this job properly and note a student’s creativity or thinking. And what happens to these students if the grades do not correspond to reality? Apparently a trial version tested last winter gave more bad grades than a human grader.

The results of these tests are also used by the administration to rate each school district. However, these academies are run in the United States by advisors elected by citizens, and the administration can possibly decide to change these advisors if the level of the academy’s students is judged unsatisfactory over time. Hence the importance of these tests, their grading and the grading system, which has consequences beyond just the students.