The enterprise-grade AI scoring engine

If you’re an existing customer, then your Customer Success Manager can bring you through the available pricing options. If you’re not yet a customer, then get in touch with us to learn more about licensing Learnosity.

In our testing and validation, we’ve found that our engine typically scores essays the same as, or within a few points of, teacher-scored essays in our data sets.

To conduct this evaluation, we sourced publicly available datasets of essays with professional scoring, as well as data from clients and industry partners. We then compared professional human scores to AI-generated scores.

We continue to evaluate the performance of our engine against multiple datasets.

We’re building on top of large language models (LLMs), which allows us to optimize for grading use cases while still benefiting from the rapid advancements in generative AI.

We use different LLMs for different use cases, tailored to meet the specific needs of grading tasks.

No, your data will not be used to train black box models.

For our release, we’ll be using LLMs through key enterprise providers with whom we have full contractual relationships and data processing agreements in place. This is the same approach we use for Author Aide, with contractual commitments that ensure nothing provided via the LLM APIs is used for training purposes.

We’ve validated the system with both general rubrics (e.g. for analytic or expository essays) and more specific rubrics tailored to individual assignments. More detailed and specific rubrics will generally yield better results.

Feedback Aide can also handle short response questions (up to 2,000 characters), taking a short ‘stimulus’, a ‘rubric’ of correct answers and a user ‘response’. It then evaluates the ‘response’ against the ‘rubric’, scoring the student and providing them constructive feedback based on which parts of the correct answer they achieved.