> ## Content Index > Fetch the complete content index at: https://digitalteachernotes.com/llms.txt > Use this file to discover other available public pages before exploring further. # Can Technology Help Students Become Better Learners? What 33 Studies Found - URL: https://digitalteachernotes.com/can-technology-help-students-become-better-learners-what-33-studies-found/ - Published: 2026-10-17T04:21:25.000Z - Updated: 2026-10-17T04:21:25.000Z - Description: A new meta-analysis found that technology can support students’ ability to plan, monitor and reflect on learning, but most of the evidence still relies on what students say rather than what they demonstrably do. - Author: Digital Teacher - Tags: Teaching Ideas, Digital Tools A good learner does more than know the answer. They also know how to learn. They can ask: **Do I actually understand this?** **What strategy should I try?** **Where did I go wrong?** **Should I keep going or change my approach?** Psychologists call this **metacognition**. And a new 2026 meta-analysis asked whether technology can help students develop it. The researchers combined: - **33 studies** - **94 effect sizes** - traditional digital learning tools - adaptive AI systems - generative AI tools The overall result was encouraging. Students using technology-supported interventions showed higher metacognitive outcomes than comparison groups, with an average effect of: **g = 0.65** That is a moderate effect. But there is an important catch. Most of the evidence came from students **reporting that they had become better at planning, monitoring or reflecting**. Very few studies actually measured whether students demonstrated those skills independently. So the conclusion is not: **“Technology makes students better learners.”** It is closer to: **“Technology designed to support reflection and self-regulation appears promising, but we still need much better evidence that the improvement is real.”** [Read the full open-access meta-analysis](https://doi.org/10.1007/s10648-026-10227-3?ref=digitalteachernotes.com) ## The short version The meta-analysis found that: - **33 studies** contributed **94 effect sizes** - the overall effect was **g = 0.65** - effects varied substantially between studies - 20 studies used non-AI technology - 4 used conventional non-generative AI - 9 used generative AI - most interventions supported reflection, monitoring and strategic control - guided interventions showed larger effects than unguided ones - immediate feedback showed larger effects than delayed feedback - system-directed support showed larger effects than leaving learners fully in control - those moderator differences became less certain after correcting for multiple statistical tests - **31 of 33 studies relied on self-report measures** - no included study was judged to have a low overall risk of bias The most useful takeaway is: **Technology may help students think about their learning, but good design matters more than simply adding technology.** ## What is metacognition? Metacognition is sometimes described as: **thinking about thinking.** That sounds abstract. In practice, it is very concrete. Before a task, a student might ask: **What do I already know?** **What strategy should I use?** That is planning. During the task: **Do I understand this?** **Am I making progress?** That is monitoring. Afterwards: **What worked?** **What should I do differently next time?** That is evaluation and reflection. Students who can regulate these processes effectively are not simply completing tasks. They are learning how to manage their own learning. ## What can technology do here? A digital tool can support metacognition in several ways. For example, it might ask a student: > How confident are you in this answer? Then, after revealing the result: > Your confidence was high, but your answer was incorrect. What might you check next time? Or it might prompt: > Before you begin, choose a strategy. Then later: > Did that strategy work? Technology can also: - provide hints - show progress - prompt reflection - ask students to explain decisions - provide immediate feedback - suggest another strategy - help students compare their prediction with the actual result None of these requires AI. But AI can make the prompts and feedback more responsive. ## Reflection was everywhere The researchers mapped the interventions according to which metacognitive processes they supported. The most common was **evaluation and reflection**. It appeared in: **29 of the 33 studies** Monitoring appeared in: **26 studies** Strategic control also appeared in: **26 studies** Planning appeared less frequently: **19 studies** That is interesting. Educational technology seems particularly well suited to interrupting a learning activity and asking: **What just happened?** It is less commonly used to help students decide what to do before they begin. ## Guided technology looked stronger One of the exploratory findings was that students appeared to benefit more when the technology provided explicit guidance. Interventions with guidance produced an estimated effect of: **g = 0.76** Interventions without explicit guidance: **g = 0.48** That difference makes intuitive sense. Simply giving students a dashboard, chatbot or reflection tool does not guarantee that they know how to use it productively. Guidance might include: - explaining how to use a tool - prompting students at the right moment - giving examples of good reflection - suggesting strategies - structuring the learning process In other words: **Self-regulated learning does not necessarily mean completely unsupported learning.** ## Immediate feedback also looked stronger The same pattern appeared with feedback timing. Interventions providing immediate feedback showed: **g = 0.74** Delayed-feedback interventions showed: **g = 0.41** Again, this was an exploratory comparison rather than a definitive causal finding. But it suggests a useful principle. Metacognition is most valuable when students can still do something with the information. If feedback arrives immediately, they can ask: **What should I change now?** If it arrives days later, the opportunity to regulate the original learning process may already be gone. ## More learner control was not automatically better This may be the most counterintuitive finding. Interventions where the **system played a stronger role in directing the learning process** had an estimated effect of: **g = 0.78** More learner-directed interventions showed: **g = 0.50** We often assume that giving students more control is automatically beneficial. But students who are still developing self-regulation may not yet know: - when to stop - when to review - which strategy to choose - when they misunderstand something A well-designed system can provide structure until students become better at managing those decisions themselves. That resembles scaffolding in ordinary teaching. Support first. Independence later. ## But these comparisons need caution There is an important statistical detail. The researchers tested several possible moderators. Guidance, immediate feedback and system-directed control initially showed significant differences. But after adjusting for the fact that many comparisons were being tested simultaneously, those differences no longer met the study’s stricter significance threshold. So we should not turn them into rules such as: **“Immediate feedback definitely doubles the effect.”** The safer interpretation is: **These patterns are promising and worth investigating, but the evidence is not yet strong enough to establish them as universal design principles.** ## What about generative AI? This is probably the question many people will ask first. Nine of the 33 studies involved generative AI. Descriptively, the estimated effects were: - generative AI: **g = 0.83** - non-AI technology: **g = 0.64** - conventional non-generative AI: **g = 0.35** That makes generative AI look impressive. But the researchers found **no statistically significant difference between technology types**. And the number of studies was small: - 20 non-AI studies - 4 conventional-AI studies - 9 generative-AI studies So this is not evidence that ChatGPT-style systems are better for metacognition. The authors explicitly warn against that interpretation. The more interesting point may be that generative AI can support metacognition in ways older technology struggled to do. It can ask open-ended questions. Challenge reasoning. Offer alternative strategies. Ask students to justify an answer. But it can also do the opposite. It can remove the need to think. ## AI can support reflection or replace it Imagine a student writing an essay. One AI interaction might be: > Rewrite this paragraph for me. The student receives a better paragraph. But little metacognitive work was required. Another interaction might be: > Ask me three questions that will help me decide whether my argument is convincing. Do not rewrite it for me. Now the AI is supporting evaluation. The technology is identical. The cognitive role is completely different. That distinction may become increasingly important. AI can either: **do the thinking** or **support the student in thinking about their thinking.** The educational value depends partly on which role we give it. ## Here comes the biggest limitation The overall effect of **g = 0.65** sounds impressive. But we need to look at how metacognition was measured. Of the 33 studies: **31 used self-report measures.** Those studies produced: **90 of the 94 effect sizes.** Only two studies used performance, judgment or calibration measures. Only one included trace or process data. In other words, most researchers asked students questions similar to: > I monitor my understanding while learning. > I change strategies when something is not working. > I reflect on my performance. Students receiving a metacognitive intervention may become more aware of these concepts. That can make them rate themselves more highly. But it does not necessarily prove that they actually became better at regulating learning. ## Feeling more strategic is not the same as being more strategic This is a surprisingly important distinction. Imagine a student says: **“I am much better at noticing when I don’t understand something.”** That is useful. But ideally we would also test: Can they correctly identify when their answer is wrong? Can they predict their performance accurately? Do they switch strategies at the right moment? Do they behave differently when studying? Those are harder things to measure. And the current research base contains very little of that evidence. So the meta-analysis itself is unusually cautious. The average effect is positive. But what exactly improved remains less certain. ## The studies also had quality limitations Another important caveat: **none of the 33 included studies was judged to have a low overall risk of bias.** Nine randomized studies had some concerns. Most quasi-experimental studies were judged to have moderate risk. Two were rated as serious risk. That does not make the research useless. But it means the headline effect should not be treated as a precise estimate of what would happen in every classroom. The authors describe the evidence as promising, not definitive. ## Most of the evidence is from older students Only: **9 of the 33 studies** involved K–12 learners. The remaining 24 involved post-secondary students. So the evidence applies more strongly to university and college learning than to younger school students. That is another reason not to turn the findings into a universal classroom rule. We need much more research involving children and adolescents. ## What could teachers take from this now? Even with the limitations, the study points toward a useful design principle. Technology becomes more interesting when it asks students to do things like: **Plan** What is your strategy? **Monitor** Do you understand what you are doing? **Evaluate** Was your answer actually good? **Adjust** What should you change? That can happen through: - quizzes - digital journals - dashboards - games - learning platforms - AI tutors - chatbots The specific technology may matter less than the regulatory activity it supports. ## A better question for classroom AI Instead of asking: **“Can AI help students complete this task?”** we might ask: **“Can AI help students become better at deciding how to complete tasks themselves?”** That is a much higher bar. A tool that always provides the next answer may create dependency. A tool that helps students notice: **I don’t understand this yet** may build independence. That is the promise behind metacognitive technology. ## The biggest takeaway Across 33 studies, technology-supported interventions were associated with moderately better metacognitive outcomes. That is encouraging. But the evidence is also a warning against easy conclusions. Most studies measured what students **said** about their learning rather than what they demonstrably did. Effects varied widely. The research base is still small. And we cannot conclude that AI is better than other technologies. So perhaps the most useful conclusion is not: **Technology improves metacognition.** It is: **Technology can be designed to make students stop, monitor, reflect and adjust instead of simply moving to the next answer.** And that may be one of the most valuable things educational technology can do. ### Source **Yuan, Y. & Du, Y. (2026).** *Technology-Supported Interventions and Learners’ Metacognition: A Three-Level Meta-Analysis.* Educational Psychology Review, 38, 130. [Read the full open-access study](https://doi.org/10.1007/s10648-026-10227-3?ref=digitalteachernotes.com)