As educational technology continues to evolve, developers are promising revolutionary changes brought about by artificial intelligence (AI) in both learning and teaching methods. However, a recent report from Stanford University reveals that the evidence supporting these claims of effectiveness is not as robust as expected, raising important questions about genuine student learning.
The study highlights a notable tension: while AI can enhance student performance while it’s being used, those improvements often diminish once the technology is removed.
CONTEXT AND SCALE OF RESEARCH
Despite the rapid integration of AI in educational settings, comprehensive research on its impact remains scarce. Most studies explore whether students using AI tools achieve better performance, yet fail to determine if the technology itself is a contributing factor.
Although the report evaluated over 800 studies on AI in education, its conclusions rested on around 20 high-quality causal studies that are specifically designed to assess whether AI directly affects student outcomes.
“Causal research is crucial in determining how a tool, like an AI application, influences students and educators. This focus differs from merely analyzing survey data or descriptive statistics,” said Lily Fesler, a senior researcher at the AI Hub for Education at Stanford and co-author of the report. “This emphasis on causal studies is why we have framed our analysis this way.”
She pointed out that this distinction is vital for school leaders who are considering adopting AI. Without causal evidence, improvements may stem from factors like teacher experience, student motivation, or classroom dynamics rather than the tool itself.
However, among the numerous studies, few meet the rigorous criteria required for causal analysis. Researchers noted that this creates a volatile market with limited proof of the actual impact of AI in education.
“The broad conclusion is that it’s still premature to determine whether AI is speeding up education or fundamentally changing it. Current research indicates hints of both,” noted Chris Agnew, co-author and managing director of the AI Hub for Education. “The overall situation is mixed and necessitates further, more intentional research.”
Agnew also mentioned that much of the existing research pertains to currently available tools, which tend to be primarily one-to-one, chatbot-style products. Moreover, the data often fails to reflect how AI is actually applied within classrooms.
EVIDENT UPSIDES
Both Agnew and Fesler acknowledge that while evidence of AI’s benefits in education exists, it primarily highlights specific and limited advantages, especially when students receive step-by-step support and timely feedback.
Fesler indicated that studies reveal AI tools can enhance students’ performance in structured tasks, such as solving math problems and providing critiques on writing assignments.
The report also discovered that AI can significantly alter how teachers allocate their time. Tools can decrease the hours spent on tasks like grading, lesson planning, and providing feedback—sometimes by as much as 30 percent—without compromising lesson quality.
However, Fesler cautioned that time saved does not always equate to a reduction in teachers’ workloads.
“While we know that AI saves teachers time, it doesn’t necessarily mean they spend less time working at home,” Fesler explained. “Instead, they may reinvest that saved time into other tasks they believe will benefit student learning in the classroom.”
Agnew emphasized the broader potential for AI to enhance accessibility and empower students to take control of their own learning experiences.
“There is a genuine opportunity for students to develop agency,” he noted, suggesting that AI could be particularly beneficial for those with diverse learning needs.
PROMINENT DOWNSIDES
The report consistently expresses concern over the implications of removing AI tools.
Research shows that although students often perform better while using AI tools, they struggle to replicate these results independently, with many of the gains appearing to diminish or vanish once the technology is withdrawn.
“We have seen several instances where AI can boost student performance during its use, but it’s crucial to recognize that these improvements often do not carry over when the tools are not available,” Fesler remarked. “There’s growing evidence that indicates students may not perform well once these tools are removed.”
Fesler described this phenomenon as cognitive offloading, where students rely on AI instead of developing their own comprehension.
“If students predominantly depend on AI to complete assignments, they may engage less in the task and think less critically,” Fesler commented. “While performance on the task may improve, they are likely not internalizing and learning effectively from the experience.”
Agnew pointed out that these risks extend beyond educational settings.
“My concerns include the erosion of fundamental skill development, a decline in critical thinking abilities, and shifts in young people’s self-perception due to technology,” he stated.
Additionally, the design of the AI tools plays a critical role. Tools that facilitate guided reasoning tend to yield better educational outcomes than those that merely provide answers.
“It significantly hinges on the pedagogical design of the AI tool,” Fesler explained. “While there are certainly methods to engage students cognitively when using an AI tool, this is not an automatic outcome; such engagement must be intentionally designed.”
POLICY RELEVANCE
For district leaders, the findings present an essential consideration when evaluating AI tools: discerning when such technologies function as effective aids versus when they may serve as crutches in the learning process.
“Any product making significant claims about its outcomes should be scrutinized cautiously, as the research is still in its early stages,” Agnew urged, stressing the need for leaders to question vendors about whether tools truly enhance independent learning rather than just improving performance during usage.
Researchers also emphasized the importance of not replacing fundamental teaching elements with technology.
“The promise of AI as a beneficial tool does not imply that it is beneficial for children to spend more time solely in front of computers or their phones,” Fesler asserted, reiterating the centrality of relationships and in-person learning experiences.
