Artificial intelligence can help a teacher analyze student data and work, connect their instruction to state and district goals, and tailor lesson plans.
When it comes to professional portfolios, is that a critical tool for improvement—or cheating?
The spread of AI in education has led to widespread policy discussions around its use for classroom instruction and students’ work. But those debates have almost entirely skipped over the technology’s potential use in teacher-evaluation systems, particularly the portfolios that make up a key element of certification—in some states for initial licensing exams, in others to renew or obtain permanent certificates, and even for advanced credentials like national-board certification.
Yet teachers building their portfolios now can face vague, conflicting, or nonexistant rules from states and districts, teachers’ colleges and professional organizations, and even testing organizations themselves on how to use the technology ethically and effectively.
Without clear guidance, teachers, like students, can be tempted to just “cut and paste” language provided by AI—and principals, like teachers, often aren’t able to identify inappropriate AI use, according to Leah Van Dassor, the president of the St. Paul, Minn., teachers’ union, which recently included language in its labor contract to shape administrators’ AI use in teacher evaluations and potential staff displacement. (The union proposed but ultimately did not include language around teacher use in professional evaluations.)
“People are using AI, and they don’t see it as a danger, but it is like a frog in hot water; it sneaks up, and all of a sudden, it’s like, ‘Oh yeah, I just crossed the line, but I didn’t realize it,’” said Van Dassor.
Confusion over state requirements can also limit innovation in meaningful approaches to using AI for teaching portfolios.
“We have to be very strategic with the way that we’re teaching our [student-teachers] to use AI to actually help them in the long term in their educational context,” said Amanda Hurlbut, an associate professor and the head of educational leadership at Texas Women’s University. “I have to say, ‘You can’t use AI for this or this because it’s not allowed on the test,’ and ultimately, the test determines whether or not you get certified. You can’t get a job if you don’t get certified.”
Teacher policy a lower priority in AI discussions
In theory, teaching portfolios provide a more nuanced picture of a teacher’s development and effectiveness than a certification test alone—which is why a majority of states allow their use in some capacity, and more than a dozen require portfolios as part of initial teacher licensure.
Typically, they allow teachers to present lesson plans and video recordings, data and examples of student work, and their own reflections on their instructional practice, creating “living demonstrations of teachers’ readiness for the classroom,” according to Cheryl Holcomb-McCoy, the president and CEO of the American Association of Colleges for Teacher Education.
States have prioritized guidance around the high-profile issues of student privacy, wellness, and academic integrity but have not done much to address the teacher-quality implications, in part because teacher professional development and assessments are generally considered a labor issue, according to Christian Pinedo, the vice president of the AI Education Project (aiEDU), a nonprofit focused on helping teachers understand and prepare for the technology in K-12 and a member of an AI working group at the California education department.
California officials have heard anecdotes of teacher-candidates using AI to develop instructional portfolios required for its state-required test, the California Teaching Performance Assessment. But the state has no official policy on how they do so and no screening for AI use.
A state working group is developing new guidance around AI use in education, but it has focused so far on student use and privacy and, to a lesser extent, how principals can use AI when evaluating teachers, rather than how teachers use AI themselves for professional development.
California’s approach mirrors that of the nation at large.
None of the dozen or so states that require teachers to submit portfolios for initial or recertification has specific guidance for how teachers can use AI in developing these professional presentations. Some, like Arkansas and South Carolina, have no statewide AI policy at all.
The most important thing [in teacher portfolios] is really talking about what's happening in their classrooms, and those conversations are not an AI job. That is the job of the teacher with a mentor.
Others, like Alabama and Tennessee, have left it to individual districts to set AI practices, while Maryland sees portfolios as the responsibility of teacher-preparation programs. Georgia and North Carolina each have released new teacher AI guidelines recently, but these focus on classroom and student use, not teachers’ own use for professional development. (North Carolina’s newly released guidance for principal portfolio assessments don’t address the use of AI, either.)
“There’s been very, very little guidance, and the teachers are looking for the guidance,” said Maria Elena Guzman, the director of pedagogy for AFT Local 2 in New York City and an instructor for the National Academy for AI Instruction, a five-year, $23 million partnership between the American Federation of Teachers and three of the largest AI developers—Anthropic, Microsoft, and OpenAI. She noted that the nation’s largest school district, New York City, has not formally addressed teacher professional use of the tools.
“It’s almost like no one wants to talk about this,” she said. “Everybody wants teachers to be using some aspects of it, but no one’s really talking about it.”
Leaving teacher AI decisions up to labor negotiations can have its own drawbacks, though. The teacher contract in St. Paul, Minn., restricts administrators from using AI alone to make decisions on teacher evaluations, but it doesn’t guarantees any AI training for teachers or specify how teachers should use AI in their own professional learning. The state requires Minnesota requires teacher-candidates to follow the state code of ethics and get their professional portfolios notarized as authentic professional work, but does not detail AI use.
As at the state level, the district has focused mainly on how teachers should handle student AI use.
“It’s pretty cliche now, but it’s like [students] having somebody else write their papers—that was the biggest concern,” Van Dassor said.
In general, she said, districts and unions are unprepared to have deep practical conversations about AI: “It was unbelievably difficult to get [AI contract language] passed.”
Can technology discern when AI is used to cheat on portfolios?
The concern about the quality of teaching portfolios is not a new one. Critics have long suggested portfolios can be more vulnerable to cheating, with companies offering to review and edit teacher lesson plans and videos to ensure they pass muster with reviewers. AI makes this kind of review much cheaper and more widely available.
For example, teacher-candidates submitting portfolios for EdTPA, a national performance assessment system used in a majority of states for initial licensure or evaluation, are encouraged to use the technology for planning learning items and understanding parallels among the three portfolio tasks, according to a 2023 statement by Pearson Evaluation Systems, which administers edTPA, and the Stanford University Center for Assessment, Learning, and Equity, which developed the assessment. A 2023 FAQ warns that portfolio reviewers will use originality-detection software to catch improper submissions.
It’s unclear how much this could help, as AI detection software still has significant error rates for student work and virtually no evidence on identification in teacher professional work.
Pinedo of aiEDU argued that state and national portfolio systems need to develop assessment systems capable of evaluating the process of teacher work with AI, not just the products.
The National Board for Professional Teaching Standards has taken a hard line against most AI use for its prestigious national-board certification, considering it a “disqualifying breach” to use it when drafting or writing any of the written commentary or forms or requesting feedback on materials or scores, the organization said in an FAQ released in March.
While teachers can use AI to brainstorm teaching strategies or generate materials for their students, they must disclose any such use in their application and they are “encouraged to seek human feedback from a colleague” instead.
Guzman, the AFT member leading that union’s work, agrees thatteachers should use AI primarily for presentation, not substantive reflection. Her national AI academy has not included instruction on using AI in teachers’ professional evaluation work, though Guzman said some mentor-teachers do address the topic when working with mentees.
“The most important thing [in teacher portfolios] is really talking about what’s happening in their classrooms, and those conversations are not an AI job. That is the job of the teacher with a mentor,” Guzman said. “It’s so, so important that these tools only expand and refine what teachers already do well.”
Evolving thought partners
Some teacher education groups also think the evolution of AI means it’s time to rethink whether AI use in development and feedback should be allowed more broadly.
“It seemed like when AI came up, the first wave of concern was, ‘Well, the [teacher-candidates] are going to be dishonest about their scholarly work; how do we stop that? And then it was, ‘Are they going to lose their ability to really think critically?’” Holcomb-McCoy said. “We’ve been slower to get to the discussion of where the technology could be used meaningfully.”
Doing that, teacher education and AI experts say, means teaching educators to think of the technology less as a tool for generating products and more as a “thought partner” to help them explore their professional development.
What might that look like? aiEDU thinks it has one model. AI chatbots like Claude or ChatGPT create conversation logs, Pinedo said. These can be used to evaluate not only teachers’ professional reflections but also to document their discussions and work with AI programs, which can show teachers’ reasoning and whether they used AI to reflect or offload thought.
“When a teacher is going through a task with an AI chatbot, it gives us an ability to actually see, is this person using discernment?” Pinedo said. “Are they actually asking critical questions? Are they just trusting whatever the chatbot says, even if it’s wrong? Or are they challenging it and working with this tool in order to reach the goal in a constructive way?”
And while Holcomb-McCoy, Guzman, and other teacher educators say that teachers should prioritize help and feedback from human coaches and colleagues, they acknowledged that is a luxury not all teachers receive.
“I don’t think we can afford at this point to say, no, we’re not going to use AI, we should stop [teachers] from using AI in the portfolio process,” Holcomb-McCoy said. “If we say no, [teacher-candidates] are probably going to use it anyway. And then we’re setting them up for possibly getting into trouble or not being able to become teachers.”
