Providence Public Schools says its Amira reading tutor does not store student voice recordings, but recordings from Amira Assess, the benchmark and dyslexia screening test, are kept for 48 hours and then deleted. The district's statement on Amira, updated Oct. 29, describes the two products as having different audio-retention settings.
The statement follows an Oct. 22 letter to families that said the district had chosen a setting under which new student voice recordings are not retained. The letter did not distinguish between the two products. The updated statement says families "have raised additional questions" about recordings, assessment options and how the platform responds to speech differences.
Two products, two settings
Amira Tutor is the practice tool; it listens as students read aloud and gives feedback on accuracy and fluency. The statement says recordings are not stored in it. The Oct. 22 letter said students in grades K-5 use Amira.
Amira Assess is given three times a year. The statement says recordings are held for 48 hours for quality assurance and human review if a scoring concern arises. It says Amira does not use Providence student recordings or other student data to train or improve its AI models.
Families may instead choose a nonverbal, audio-off assessment that students complete on a computer without speaking. The statement calls it an approved dyslexia screening tool but says it does not give the same information about a student's oral reading. It encourages families with concerns to talk with their child's teacher.
Teachers cannot replay Tutor sessions
The statement says the district opted in August not to record students and that under that setting no new recordings are retained, but it also says Amira Assess recordings are held for 48 hours. It does not explain how the two fit together, and the district was not asked to reconcile them for this article. The statement also says the district can and has requested deletion of all recordings, and that recordings are kept only for the period the district or a state agency requires; it does not say how long earlier recordings were kept. It also says that because recordings are not retained in Amira Tutor, teachers cannot play back a session to check for speech-recognition errors, including for students with speech differences such as a lisp. The statement says Amira told the district that retained recordings let teachers listen to a session, review possible recognition errors and correct scoring.
The district reports that its internal monitoring shows students using Amira about three times a week for 11 minutes a session. It lists the recommended use as 30 to 60 minutes a week, with 30 minutes as the target the district calls research-proven. It cites a 2022-23 independent evaluation in Utah that found higher reading scores for first and second graders who used the program for 30 minutes a week over 30 weeks, with effect sizes of 0.41 and 0.45, and notes that results varied by grade and level of use.
The Oct. 22 letter said 110 minutes of platform use is about 7 percent of the 1,650 weekly instructional minutes used in the family guide's calculation. It said the district uses Amira as its dyslexia screening tool, which it said the Rhode Island Department of Education requires. It also described data-sharing agreements with vendors and tools to review and block applications that do not meet district privacy requirements. Families are directed to their child's teacher or principal.
