Providing detailed (and consistent) feedback takes time and training. In a large chemistry course, TAs may only have time to leave a few words on each response, giving students limited insight into where they went right or wrong. At Stemble, we’ve always worked toward clear, consistent feedback through well-defined rubrics. Our design team has been exploring another way to put that guidance into practice: a hybrid grading approach that combines AI-generated feedback with manually assigned grades.

In this model, AI provides feedback on each individual question, guided by the assessment rubric. TAs can use that feedback as a second source of input as they review the student’s work, formulate their own comments, and assign marks. In this hybrid approach, the source of the final grade is the TA, not the AI.

The exciting part? Students can receive richer explanations of their work, and TAs have a more consistent starting point to support their review. The feedback can help students understand what they did well, where their reasoning needs attention, and how to improve.

A Small Example

Imagine a lab report asking students to explain how an experimental error affected their results. One student writes: “We spilled some product during transfer, which explains our low percent yield and why our product was less pure.”

With limited marking time, a TA might leave a brief comment such as “Yield and purity are different.” AI-generated feedback can help unpack that distinction:

“You’ve correctly identified that spilling product during transfer could reduce your percent yield because less product was recovered. However, losing some product does not, by itself, make the remaining material less pure. Purity describes the composition of your sample. What evidence from your experiment suggests that impurities were present?”

The TA reviews this feedback alongside the student’s report and the rubric. They can use the AI feedback to help formulate their own comments and assign the marks.

Another Option for Different Teaching Contexts

Comfort with AI varies across instructors, TAs, and students. For some courses, uncertainty about its role in grading can become a roadblock.

This hybrid approach gives those teams another way to use AI feedback within a workflow they’re comfortable with. Students know that a person determines their grade, while still benefiting from detailed feedback on individual responses. TAs retain responsibility for grading, with an additional source of guidance to support their decisions. The right balance will depend on the course, the teaching team, and their goals.

Hybrid grading adds another option alongside Stemble’s existing approaches, giving instructors flexibility in how they bring AI into their assessments. For Stemble, that flexibility serves the same ultimate goal: a better student experience, with clear, consistent feedback that helps students understand their mistakes, build confidence, and improve their learning outcomes.