Hi John,

There’s a rapidly growing need for transparency, consistency, and rigor in AI tools for classrooms. Today, AI models are rarely evaluated on meaningful educational tasks—and we have an opportunity to change that. That’s why we're building the systems to ensure AI delivers on its promise. Recently, Jessamy Almquist on our team and Rose Wang at OpenAI co-authored an op-ed in The 74 explaining why rigorous evaluation must become essential infrastructure for AI in education. The piece highlights how our Evaluators tool helps assess whether AI-generated content meets pedagogical standards, using text complexity as a concrete example.

"As AI technologies become more widely adopted in schools, evaluation must move from an afterthought to essential infrastructure. Just as medicine relies on clinical trials and safety protocols, edtech needs open, transparent and credible ways to assess quality."

Our first Evaluators are focused on literacy for students in 3rd and 4th grades. They ensure AI-generated reading passages meet the depth and rigor students need to further their learning. By using Literacy Evaluators, you’re not just improving text quality — you’re helping ensure all students have access to materials at their level.

Literacy Evaluators assess qualitative dimensions — such as structure, vocabulary, and knowledge demands — that extend far beyond traditional readability scores. Built on the SCASS rubric from Student Achievement Partners and expert-annotated datasets, these evaluators provide transparent, research-based insights that help improve text quality and instructional value. 

• Sentence Structure

Analyzes a text’s syntax to assign a complexity rating — from slightly to exceedingly complex — to ensure AI-generated content is appropriately challenging for each grade level. By examining features like clause patterns and sentence variety, it offers a deeper view of text rigor and supports literacy growth.

• Vocabulary

Evaluates how challenging a text’s vocabulary will be for students by estimating background knowledge and analyzing word familiarity, specificity, and academic rigor. It helps educators and developers ensure AI-generated texts align with grade-level expectations, build knowledge, and strengthen reading comprehension.

• Grade Level Appropriateness

Determines whether a text is suitable for independent reading at a given grade level and identifies when additional scaffolding may be needed. By combining quantitative and qualitative analyses, users can ensure that their content is educationally accurate, standards-aligned, and adaptable across diverse classrooms.

Sincerely,

Sandra Liu Huang

Head of Education