This story first appeared at The 74, a nonprofit news site covering education.
This summer, everyone from homeschoolers to large urban districts like Los Angeles Unified is trying to process what artificial intelligence will mean for the coming school year. Educators find themselves at a crossroads — AI’s promise for revolutionizing education is tantalizing, yet fraught with challenges. Amid the excitement and the angst, and the desire to recover from COVID learning losses, a powerful but often overlooked tool for boosting student achievement lies hidden in plain sight: strategic testing.
By harnessing AI to create frequent, low-stakes assessments, teachers can unlock the scientifically proven benefits of the testing effect — a phenomenon in which students learn more by taking tests than from studying. This summer, it is worth challenging assumptions about testing and exploring how AI-powered strategic assessments can not only boost student learning, but save teachers valuable time and make their jobs easier.
Summer is the perfect time to consider the benefits of frequent, low-stakes strategic assessments powered by artificial intelligence.
Unlocking the promise of AI requires first understanding the testing effect: Students’ long-term retention of material improves dramatically — 50% better — through exam-taking than through sophisticated studying techniques like concept mapping. This effect isn’t limited to rote memorization; it enhances students’ inference-making and understanding of complex concepts. The advantages emerge for multiple disciplines (e.g., science and language learning) and across age groups, and even extend to learners with neurologically based memory impairment. Additionally, teaching students about the phenomenon of the testing effect can boost their confidence.
Unfortunately, in most classrooms, opportunities for students to practice retrieving, connecting and organizing knowledge through testing happen rarely — think occasional quizzes or infrequent unit tests. This isn’t surprising, given all the pressures teachers face. Developing and grading tests is time-intensive — not to mention thankless — work.
But AI tools like ChatGPT and Claude 3.5 Sonnet can change that. They can generate diverse, personalized assessments quickly, potentially helping teachers leverage the testing effect more effectively — converting a counterintuitive research finding into a classroom practice that could save time and help students learn more. With AI handling the creation and analysis of tests, educators can easily incorporate frequent, low-stakes assessments into their lesson plans.
How does it work?
To illustrate, we asked ChatGPT to create a 10-minute test on natural resources for sixth graders in Maryland. In less than 10 seconds, the tool provided options for multiple choice, true/false, short answer, matching and diagram interpretation questions. We even got a creative thinking essay prompt: “If you were a superhero tasked with protecting Earth’s natural resources, what would be your superpower and why?” By picking and choosing test items, running the prompt a second time and lightly editing a couple of questions, we had a compelling quiz, created in 10 minutes. The AI tool also provided comprehensive instructions and an answer key.
Teachers can tailor this process in dozens of ways. They can input key concepts and learning objectives to fit their curriculum needs. They can fine-tune test questions for relevance and difficulty. They can inform ChatGPT about the class’s interests to bolster student engagement.
What about grading? Not only can AI grade test papers and even essays, but it can guide students in assessing their own work and that of their classmates. For example, when students grade each other’s assignments, they can check their feedback against ChatGPT. Doing so provides another opportunity to practice recalling key material. Teachers’ evaluations and personalized feedback will remain critical, but these do not have to happen every time.
Take, for example, a language class with a ChatGPT-generated vocabulary test. For objective parts of exams, like multiple-choice questions, students might self-assess by using ChatGPT to grade these items quickly. For tasks like sentence construction, students might engage in peer assessment to gain new insights from classmates on word choices and sentence structure. Teachers can step in for more complex tasks such as creative writing. Rotating among AI, peers and teachers lightens the grading load significantly while ensuring diverse, rich feedback.
Embracing AI-assisted strategic testing could create a more effective and fulfilling educational experience for students and teachers alike. As educators navigate the evolving landscape of AI in education, strategic testing offers a balanced approach. It leverages AI’s capabilities to enhance teaching and learning while preserving the crucial role of human teachers in the classroom. This summer, as educators reflect and plan for the future, they should reconsider testing not as a mere assessment tool, but as a powerful catalyst for learning.
***
Xue Wang is a fourth-year Ph.D. candidate at Johns Hopkins University’s School of Education, focusing on learner autonomy, meta-analysis and AI tools for education research.
Hunter Gehlbach is a professor and director of the Ph.D. program at the Johns Hopkins School of Education. He has worked for Panorama Education and is an occasional consultant.
The 74 is a nonprofit news organization covering America’s education system from early childhood through college and career. The 74’s journalists aim to challenge the status quo, expose corruption and inequality, spotlight solutions, confront the impact of systemic racism, and champion the heroes bringing positive change to our schools. Sign up for free newsletters from The 74.
