Should I tell people (students and parents) I use AI for grading and reviewing formative assessments?
TL;DR: Yes, but keep reading…
In a recent product interview, a teacher asked me if students or parents could tell if they were that they were using an AI to grade papers and review formative assessments. Later that day, a former colleague read my Facebook post in which where I explained why I chose to build Braide and raised concerns about potential AI bias. All of this comes on top of me recently reading a NY Times “The Ethicist” column that asked this very same question.
So, the appropriate use of AI grading software and teacher productivity tools and the disclosure of that use are clearly relevant and timely topics.
I think there are two questions to consider here:
Should teachers use AI to evaluate formative assessments and homework?
Should they disclose that use?
Should teachers use AI to evaluate formative assessments and homework?
I think there are two things to consider here: the impact on teachers and the impact on students. Today, teachers work an average of 15 unpaid hours per week. Much of that time is spent grading and reviewing formative assessments. A typical public school teacher teaches approximately 25 students per section and 5 sections at a time, meaning, they are grading or evaluating about 125 individuals for every formative assessment. In reality, the teacher is left with three options:
Grade/review fast and give less feedback each time
Grade/review infrequently and give feedback less often
Both A and B
The very human response to unpaid, after-hours work is C) both. This doesn’t serve students very well either. The research is clear: frequent, low-stakes formative assessments paired with timely and actionable feedback benefit students immensely. They understandably aren’t getting that when teachers are overworked and underresourced. What’s more, humans that are tired and overworked tend to inject error into the process fatigued, overworked teachers are more prone to mistakes. A recent study found that 60% of human grades are inaccurate when compared to testing assessments: 40% overestimate student proficiency and 20% underestimate student proficiency.
In all of our product discovery interviews with teachers, they volunteered all sorts of ways they inadvertently inject bias and error:
Inconsistency: One of our teachers described grading harder for the first few responses until he remembered the grade level he was teaching and readjusted his expectations for the rest of the grading.
Fatigue: By the end of reviewing 100+ student responses, teachers are trying to get through them as quickly as possible.
Priors/Expectations: After reviewing those 100+ assignments it is very tempting to give a quick glance and say to oneself, “This looks like a 3. Student X always gets a 3. 3 it is.”
Feelings: One teacher told us the very real feedback that it was easier to grade and give feedback to interesting (aka strong) responses, which may be the students that absorb the most feedback but not the ones that need the most feedback.
Lastly, the teachers we have spoken to acknowledge they provide less feedback and assign fewer formative assessments than students need due to lack of time and energy. It’s pretty amazing how candid people are when they don’t feel judged.
So, what does AI do better than humans?
AI is a long way from doing everything or even many things better than humans. BUT, there are some things that AI is good at:
Consistency: AI applies the same standards to every student, the same energy level, and the same focus.
Efficiency: AI grades assignments rapidly and provides a consistent level of feedback to every student and saves teachers time.
Scalability: AI can handle large volumes of grading across multiple classes, schools, or even regions. If you are a school or a district you can get consistent data on student proficiency without waiting for standardized tests.
Data-Driven Insights: AI can offer detailed performance analytics - identifying patterns, strengths, and areas for improvement at class and individual levels.
For students, the advantages are clear: more frequent and consistent feedback, and the student’s grade is no longer dictated by getting the teacher that grades “too hard” or “too easy.”
What are the downsides of using AI for grading and reviewing formative assessments?
AI is an imperfect tool and not a replacement for teachers. There are a few really clear issues to keep in mind when using AI. One is an issue with the tech and 2 are about how one USES the tech.
AI Error: AI models are still in their infancy, and they struggle as assignments and assessments get more complex. They make different mistakes than humans, but they still make mistakes.
Disconnection from Students: Completely turning over grading and assessment to an AI tool will not allow teachers to see where their students are excelling or struggling.
Trust Issues: Administrators, teachers, students, or parents who have not worked with AI may be skeptical of AI’s ability to assess work fairly and accurately.
AI has not reached a level where it is ready for hands-off, high-stakes tasks. It needs human oversight and should be viewed as a productivity tool NOT a replacement of a human expert. At Braide we want the teacher to review every Braide result and the feedback given to students.
Should one tell students and parents that AI is being used for grading and why or why not?
Early in my career, a mentor shared the “The Washington Post Test” when I posed a thorny issue I was wrestling with. In a nutshell, the “Washington Post Test” boils down to: “What would you do if you knew that your course of action would make the front page of the Washington Post?” With that as the test, it’s really a personal decision, but if you aren’t comfortable sharing it with your students and their parents it probably isn’t right for you.
What do teachers using Braide say about how AI impacts their work:
"I would have made far more mistakes than Braide. I am just blown away! It is fantastic!”
AND
"It gives me a ton of backing when a parent comes to me to talk about a grade.”
AND
“I had 2 parents complain about grades. Braide’s feedback gave me the chance to explain the grade to those parents with confidence. It was awesome!”
AND
"The number of hours you save me is incredible. But it’s not just the hours. It’s the fact that I gave valid feedback in a timely manner. I’ve never given this much feedback to my students.”
All of those quotes indicate teachers that are happy to use AI. More than that they are happy with the results their students receive.
