When an AP Calculus Reader Puts AI-assisted Grading to the Test

Whitney High School is small, but its reputation is anything but. The Cerritos campus offers 16 Advanced Placement (AP) courses and ranks first in California, with a name for academic excellence that reaches well beyond its borders.

Cindy Carlson has spent 26 years teaching math in ABC Unified School District, nine of them at Whitney High School. She teaches AP Precalculus, AP Calculus and Math I. But her expertise doesn't stop at the classroom door: she's also an AP Calculus Reader, scoring the College Board's exams. A tech-savvy teacher, Cindy is often the first in her department to try a new tool, and the person willing to train her colleagues when they're ready to give it a shot themselves. So when AI-assisted grading started making promises, she went looking.

She wasn't after a machine that would grade for her. She wanted an assistant that could grade like her.

Could an AI tool make it easy to get students' work onto a platform, take a first pass at a stack of advanced math papers, and meet the standards of an AP Reader? Could it eliminate the friction of the paper-based workflow while fitting seamlessly into the way she already taught and graded?

In other words, could AI become an extension of her craft, rather than another tool she had to work around?

And just like that, the pilot was on.

The first hurdle was everything surrounding grading: the paperwork, the data entry, the small administrative tasks that notoriously burden a teacher’s routine. Before Cindy could get started, we needed to make sure the setup wouldn't get in her way, so she received training and got familiar with the platform.

Then came the grading itself.

Cindy built her own grading scale and ran warm-up exercises through the platform to see what came back. Our challenge was to make the grading notes as flexible as she needed them to be and, paper by paper, bring the AI-assisted first pass closer to her professional judgment.

In advanced math, students can take different paths to a solution. They may restart their work, offer multiple answers, or show their reasoning in ways that call for partial credit. Cindy needed Ed.ai to account for those variations while maintaining the level of rigor expected of an AP Calculus Reader.

There was a learning curve, too. As Cindy got to know Ed, she became better at explaining exactly what she expected it to look for.

After 10 days of using Ed.ai, Cindy has graded 400 papers across her AP Calculus AB and AP Precalculus classes. Ed.ai has changed how she handles the volume of quizzes, tests, and warm-ups. Her students can now get feedback they can review on their own.

Of course, the pilot hasn't been without its challenges. With up to 35 students per class and multiple classes to manage, getting papers into the system remains time-consuming. Taking each picture by hand quickly becomes overwhelming, and a batch scanner would help, although Cindy doesn't yet have access to one at school.

Turnaround time is another concern. Long assignments take time to grade, and Cindy would welcome the option to release feedback automatically once grading is complete, for some assignments.

Still, she has already seen what the tool can make possible. When Cindy was out for a week, her students still did their warm-ups. She uploaded their papers remotely, and they could access Ed’s feedback and correct their work without ever getting their papers back.

Her students have noticed the difference, too. Her AP Precalculus classes have been using Snorkl for warm-ups, but told her they prefer the feedback they received from Ed.

Students upload their work through a link or QR code that Cindy posts in Google Classroom. Next up: connecting Google Classroom to Ed.ai, so her students can open the assignment and submit their work directly from there.

Cindy hopes that, eventually, Ed.ai could become something her school or district provides for teachers.