The Texas Education Agency will use an automated grading engine to grade copies of certain STAAR tests. The agency sees it as a way to save $15 to $20 million per year by reducing the use of contract workers.
To train it, the agency collected more than 3,000 answers to the exam, which were subject to “two rounds of human grading” The agency also ensures that when the computer is in doubt about the grade to be awarded, the copy will be re-examined by a human corrector. This will automatically be the case for types of responses that its programming does not recognize, such as those in a foreign language or using slang. In 2023 the agency had hired 6,000 proofreaders...It plans to recruit fewer than 2,000 this year. The scoring engine used is programmed to assign the same grades as a human. It will assess “reading, writing, science and social sciences” The Texas Tribune.