AI Detection Software Accuracy Questioned as Educators Adopt Diverse Approaches to Generative AI

The proliferation of generative artificial intelligence (AI) in education has led to concerns regarding the accuracy and reliability of AI detection software, prompting varied responses from school districts, educators, and students. While some institutions and teachers are investing in AI detection tools, others advise against their use or are adapting teaching methods to address AI's presence. This highlights a broad spectrum of views on how to integrate or manage AI in academic settings.

Scrutiny Over AI Detection Software

Instances of students being accused of using AI in assignments, often based on results from AI detection software, have brought the tools' accuracy into question.

Ailsa Ostovitz, a junior at Eleanor Roosevelt High School in Maryland, was accused of using AI on three assignments, with one receiving a 30.76% probability score from an AI detection program. Ostovitz denied the accusations, and her grade was subsequently lowered.

Following a meeting between Ostovitz's mother, Stephanie Rizk, and the teacher, the teacher reportedly no longer believed AI was used. Ostovitz now pre-screens her assignments with multiple AI detection tools, adding approximately 30 minutes to her work.

Prince George's County Public Schools (PGCPS), Ostovitz's district, stated that the teacher used the AI detection tool independently. PGCPS does not fund such software and advises educators against relying on it due to documented inaccuracies and inconsistencies.

A nationally representative poll by the Center for Democracy and Technology indicated that over 40% of 6th- to 12th-grade teachers utilized AI detection tools during the previous school year.

However, research by Mike Perkins, an academic integrity and AI researcher at British University Vietnam, identified issues with popular AI detectors like Turnitin, GPTZero, and Copyleaks, finding they could misclassify both human-written and AI-generated text. Accuracy reportedly decreased when AI-generated text was modified.

Companies providing these tools also acknowledge limitations. Turnitin, which introduced an AI detection feature in 2023 and holds a three-year contract worth over $550,000 with Broward County Public Schools in Florida, advises that scores below 20% are less reliable.

Turnitin recommends its AI writing detection not be used as the sole basis for adverse actions against students due to potential inaccuracies.

Edward Tian, co-founder and CEO of GPTZero, describes his tool as a "tool in the toolkit" for educators, not a definitive judgment instrument, suggesting further examination for scores above 50%.

Some districts, like Broward County Public Schools, utilize Turnitin for facilitating conversation, feedback, and assisting with authentication requirements for programs such as International Baccalaureate (IB), though IB does not mandate such software.

Shaker Heights City School District in Ohio spends approximately $5,600 annually for GPTZero licenses for 27 teachers. Teacher John Grady uses GPTZero as an initial indicator, investigating further if an assignment receives an AI likelihood score exceeding 50%. He reported that approximately 75% of students admit to AI use when confronted under these circumstances.

Concerns Regarding Bias and Effectiveness

Concerns have been raised about potential biases in AI detection software and its overall effectiveness.

Zi Shi, a non-native English speaker and student at Shaker Heights, reported that his writing style, which includes word repetition, has been flagged by GPTZero. His teacher suggested that using grammar-checking software like Grammarly might have contributed to the flagging. Research on AI detection tools' bias against non-native English speakers is limited and has yielded contradictory results.

Carrie Cofer, an English teacher in the Cleveland Metropolitan School District, reported that a chapter from her Ph.D. dissertation was flagged as 89-91% AI-written by GPTZero during an experiment. Her district does not currently purchase AI detection software.

Educators like Cofer advocate against AI detection software, citing inefficiency and the likelihood of students finding workarounds, such as using "AI humanizer" programs or pre-screening their work with detectors.

Diverse Educational Approaches to AI Integration

Beyond detection, educators and institutions are adopting various strategies to address the presence of generative AI in learning.

Chanea Bond, an English teacher at Southwest High School in Fort Worth, Texas, has implemented an almost entirely analog approach in her classes, requiring handwritten assignments, daily journaling, and process-based grading. She aims to ensure students develop critical thinking and writing skills independently, noting that students initially using AI for thesis statements did not engage deeply with texts.

Dan Cryer, an English professor at Johnson County Community College, holds that using generative AI for college essays hinders students' skill development. He suggests educators minimize AI tool integration, emphasizing the educational process over the final product to foster critical thinking and research skills.

Conversely, some educators advocate for modeling responsible AI use. Brett Vogelsinger, an English teacher in Pennsylvania, allows AI on some assignments, requiring transparency from students.

Leslie Clement, an English, Spanish, and African studies professor at Johnson C. Smith University, views generative AI as a collaborative tool. She encourages students to use AI responsibly for outlining, obtaining feedback, and comparing information. She co-created a course, "African Diaspora and AI," to examine AI's global impact.

Hannah Elder, a pre-law junior at the University of North Carolina, uses generative AI for proofreading and checking work against rubrics. She advocates for integrating AI instruction into curricula to teach students beneficial and harmful uses of the technology.

A July survey by Inside Higher Ed and the Generation Lab indicated that approximately 85% of undergraduates use AI for coursework, including brainstorming, outlining, and studying, while about 19% use it to write full essays. Over half expressed mixed feelings about its utility.

Anjali Tatini, a pre-med sophomore at Duke University, uses AI tools like Google's Gemini to understand complex subjects, generate practice problems, and brainstorm ideas, considering it an accessible, on-demand tutor. She avoids using AI for writing entire assignments.

The adoption of AI in education varies at institutional and governmental levels. Miami-Dade County Public Schools provides students access to Google's Gemini chatbot. New Jersey has allocated grants to advance classroom AI use, and the U.S. Department of Education supports the "responsible adoption of AI." The Trump administration previously issued an executive order to expand AI education in K-12 schools.