​
General concepts

​
Evaluator

A tool designed to measure the quality of materials generated by AI-powered edtech applications. They assess various aspects for pedagogical alignment and help identify areas for improvement.

​
Rubrics

A structured framework used to evaluate a concept based on learning science. It forms the foundation for a family of evaluators.

​
Dimension

A specific facet or attribute measured within a rubric. An individual evaluator is typically based on a single rubric dimension.
Rubrics And Evaluators Rubric Example Sv

​
Evaluator family

A collection of evaluators that score AI-generated content across multiple dimensions of one or more rubrics.

​
Accuracy

A measure of how closely the evaluator’s score aligns with curated (and often human-annotated) validation datasets. It is typically expressed as a percentage and indicates the evaluator’s reliability.

​
Early release

An evaluator that is released early in its development because we believe it provides useful capabilities for research and experimentation. While stable, it is small in scope and still under active development. It is intended for users to test and provide feedback to shape its iterative improvement.

​
Literacy evaluator family concepts

​
Quantitative text analysis

An objective measure of the difficulty of understanding a text passage. This analysis uses formulas to analyze features like word length, sentence length, and syllable count (e.g., Flesch-Kincaid Grade Level).

​
Qualitative text analysis

A deeper look at a text’s features, such as its structure, language, purpose, and the knowledge it demands from the reader.

​
Background knowledge

The prior knowledge a student is expected to have, which influences their ability to understand a text passage. This includes both curriculum-based knowledge and life experience.