What Does Research Say About the Effectiveness of AI in Teaching and Learning?
bixstock/Shutterstock
Professors, departments, and colleges drafting AI policies should take into account what emerging research reveals about AI in 四虎影视 and learning. While it may be tempting to create course-specific chatbots, ask students to use AI for brainstorming, or encourage AI-assisted writing, faculty and staff should ask not only how AI can augment learning but also what aspects of learning and growth might be lost when tasks are offloaded onto AI.
This article considers some of the most important findings in recent research on AI in education.
Concerns Around Cognitive Offloading
One of the most significant themes in the emerging literature is cognitive offloading, the process by which individuals delegate mental tasks to external tools. This could be a simple task, like asking a navigation app to show how to get from point A to point B, or it could be a complex task, like asking ChatGPT to write an analysis comparing numerous documents from different sources.
A 2025 published in Societies examines 666 participants and finds a significant negative relationship between frequent AI use and critical-thinking ability, with cognitive offloading serving as a key mediating factor. The study concludes that users who routinely delegate cognitive tasks to AI are less likely to engage in independent reasoning and evaluation. Gerlich acknowledges the power and potential of AI tools, but argues that the use of these tools should be balanced with activities that require active engagement and learning apart from AI: "Educational interventions should therefore integrate critical thinking exercises and foster environments where students are encouraged to engage deeply with content rather than passively relying on AI tools."
Additional evidence comes from a randomized controlled trial published in Social Sciences & Humanities Open. Researchers assigned undergraduate students either to use ChatGPT while studying or to use traditional study methods. When tested unexpectedly 45 days later, students who had relied on ChatGPT demonstrated significantly lower knowledge retention than students who studied without AI assistance. The authors describe ChatGPT as a potential "cognitive crutch" that may improve short-term efficiency while reducing long-term retention.
At the same time, some researchers have found that cognitive offloading does not inevitably produce negative outcomes for students' critical thinking. A 2025 in the Forum for Linguistic Studies "[compared] traditional writing instruction with an AI-augmented pedagogy that delegated lower-order writing tasks to generative AI tools, allowing students to focus on analysis, evaluation, and reflection." The study found that the AI-enhanced group of students both produced stronger, more coherent essays and improved in their critical thinking (as measured by "standardized critical thinking assessments").
These early studies suggest that AI's educational value depends heavily on instructional design. AI is likely most beneficial when it supports or supplements, rather than replaces, the cognitive and critical work at the heart of learning.
Chatbots and Personalized Learning
Perhaps the most visible use of AI in higher education is the deployment of chatbots and conversational agents. A recent EDUCASE found that 37% of colleges have institutional licenses for AI chatbots, and 14% of colleges have developed their own bots. These tools can answer questions, provide feedback, explain concepts, and offer tutoring support at any time of day.
Research increasingly suggests that AI tutoring systems can be quite effective in delivering content. A 2026 Stanford Law School found that the answers provided by AI chatbots were deemed superior to answers provided by law faculty 75 percent of the time (as judged by other law faculty).
While chatbots are promising, they do have drawbacks. Researchers conducted a systematic review of 58 studies examining AI in computer programming education in the volume of the International Journal of Learning, Teaching and Educational Research. The review found evidence that chatbots and intelligent tutoring systems often improved learning outcomes through personalized feedback and on-demand assistance. However, the researchers also found that "overreliance" on AI can lead to "superficial learning," and they pointed out some concerns with errors in AI responses and challenges with academic integrity.
While chatbots can give students 24/7 access to course information and provide immediate answers, educators must consider whether instant assistance reduces productive struggle -- the process through which many important learning gains occur.
Mentorship and Human Relationships
As administrators and faculty create chatbots that provide a wide array of assistance to students with everything from college policies to course material, it is even more important to prioritize and make space for human relationships, mentorship, and personal interaction.
A 2026 of 26 studies in Educational Psychology Review found that there is a moderately positive correlation between strong faculty-student relationships and the likelihood of students to continue in their degree path.
AI tools and chatbots should not come at the expense of building meaningful and lasting relationships between students, faculty, and staff. Faculty can encourage questions from students and engage with students actively in the classroom, as well as make it clear to students that their office hours are there for a variety of purposes, from talking to students about professionalization and academic paths, to helping students understand course expectations. Staff and administrators can structure strong faculty-student mentorships and provide meetings and forums that allow for healthy communities within and between departments.
Structured Prompting and Critical AI Skills
One way to encourage students' critical thinking is to teach them how to evaluate AI outputs, engineer stronger prompts, and engage with all of the various dimensions of AI using ethical and critical reasoning.
When done correctly, using, crafting, and revising complex, structured prompts can encourage critical thinking. A 2025 in Data found that "structured prompting significantly reduces offloading and enhances both critical reasoning and reflective engagement."
Similarly, a 2025 published in Education Sciences found that students who used AI in an actively questioning way -- through a "collaborative dialogue" -- demonstrated moderate levels of critical thinking. Students who engaged with AI more passively, by simply accepting its responses, demonstrated low or no critical thinking.
Conclusion
AI, when used judiciously and selectively, promises to enhance learning and help center critical thinking. In this exciting but challenging educational ecosystem, faculty and administrators should consider how to protect and encourage students' independent thinking processes, foster human relationships, and help students engage cautiously and critically with AI. This might mean carefully carving out spaces for human struggle, human cognition, and interpersonal connections in courses and across campus.
