Technology Alone Doesn’t Cause Learning: Learning Activities Matter

A device is not an instructional method

Educational technology is often evaluated through nouns: laptop, tablet, chatbot, simulation, virtual reality. Learning happens through verbs: retrieve, compare, explain, construct, practise, receive feedback, revise and collaborate.

That difference is more than wordplay. A device can deliver a lecture, host a discussion, simulate a system or distract from all three. Two classrooms can use the same platform while learners do entirely different cognitive work. If an evaluation asks only whether “technology” was present, the category is too broad to explain the outcome.

A more useful chain begins upstream with an affordance, moves through the activity the learner actually performs, examines the support around that activity, and ends with an independently measured capability. Technology may enable the chain. It does not cause learning by its presence.

Look beyond the substitution

A 2024 systematic review examined 28 meta-analyses of technology-enhanced learning in higher education. Those syntheses represented 1,286 primary-study effects; 45 eligible effects entered a second-order meta-analysis organised by learning activity and cognitive support. When technology merely substituted for nontechnology instruction, there was no substantial change in cognitive outcomes on average. Better outcomes appeared when technology supported specific learning activities or enabled more advanced forms of activity. ¹

This finding is a useful antidote to novelty claims, but it has boundaries. The review focused on higher education. It combined meta-analyses of varied technologies and designs, and second-order synthesis introduces overlap and abstraction. “No substantial substitution effect” is not “technology never matters.” It means that changing the delivery medium without changing the learning process is a weak theory of improvement.

An older debate reached a related conclusion from another direction. Richard Clark argued that media deliver instruction but do not themselves influence achievement any more than a delivery vehicle causes nutrition. ³ Critics and later researchers have rightly examined cases where a medium’s affordances change what can be represented or done. The most productive reading today is not a ban on media effects. It is a demand to identify the mechanism.

The technology-to-learning chain

Four questions make that mechanism visible.

1. Affordance: What can this technology make possible, such as rapid feedback, dynamic representation, communication across distance, safe simulation or adaptation?
2. Activity: What does the learner actually do and think: watch, select, retrieve, explain, build or discuss?
3. Support: How are attention, sequence, feedback and strategy guided?
4. Outcome: What can the learner later remember, explain, use or transfer?

Each link can fail. A simulation may make an invisible process manipulable, yet leave learners clicking without a question to investigate. An AI system may generate feedback instantly, yet respond to the wrong misconception. A collaborative document may permit interaction, while the group divides the work so no one integrates the whole idea.

This chain also clarifies what counts as evidence. Login frequency verifies access. Completion verifies an event. Satisfaction captures experience. None independently establishes that the target capability changed.

Activity is a stronger unit of design

The ICAP framework distinguishes passive, active, constructive and interactive modes of cognitive engagement. Receiving information is passive; physically manipulating it without adding new ideas is active; generating an inference or explanation is constructive; building on another person’s reasoning can be interactive. The framework predicts deeper learning as engagement moves toward constructive and genuinely interactive activity, under appropriate conditions. ²

The labels are not a tournament in which every discussion defeats every explanation. A confused group can interact unproductively, while a carefully designed explanation can be exactly what a novice needs. ICAP is valuable because it asks what the learner produces and how the activity engages knowledge.

A 2014 meta-analysis of 225 studies found that active learning improved performance in science, engineering and mathematics courses relative to traditional lecturing on average. ¹² Technology can support those activities through polling, shared models, immediate practice or simulation, but the device is not the active ingredient by itself.

Designers can therefore start with the verb. If the goal is causal reasoning, ask learners to predict, manipulate and explain a system. If the goal is discrimination between similar concepts, ask them to compare cases and justify a choice. If the goal is a procedure, combine modelling with progressively independent practice. Only then choose the medium that makes the activity more feasible, accurate, accessible or scalable.

Technology can add real value

Rejecting technological determinism does not require technological pessimism. Meta-analyses across decades often find positive average effects for classroom technology, online and blended learning, and mobile learning. ^{4 5 6 7} The variation inside those averages is the point: outcomes depend on instructional function, comparison quality, time, subject, learner and implementation.

Technology is especially promising when it changes a meaningful constraint.

Representation. Dynamic diagrams and simulations can expose processes that are too fast, slow, small, dangerous or abstract to observe directly.

Practice. A system can provide many varied problems, record attempts and make low-stakes retrieval easier to schedule.

Feedback. Well-grounded automation can shorten the delay between action and useful information, while escalating uncertainty to a human.

Construction. Learners can create models, code, explanations, media or arguments that would otherwise require inaccessible tools.

Interaction. People can compare reasoning across distance, preserve a discussion and jointly revise an artefact.

Access. Captions, adjustable text, alternative formats, translation and asynchronous participation can remove barriers, provided the implementation is accurate and users retain control.

Personalisation belongs on this list only when its mechanism is explicit. A 2024 meta-analysis of 47 higher-education interventions reported medium average effects for personalised technology-enhanced learning and used association-rule mining to explore combinations of learner modelling and adaptive support. ⁸ The evidence is promising, but it does not imply that any recommendation engine is pedagogically personal. A useful adaptation changes a task or support in response to valid evidence; a cosmetic adaptation merely changes presentation.

Support determines whether an affordance becomes learning

Access to a powerful tool does not teach learners how to use it strategically. Technology-supported self-regulation illustrates the gap. A 2024 umbrella review found recurring challenges in helping learners plan, monitor and reflect in digital environments, with scaffolding and instructional design central across the underlying reviews. ¹⁰

Collaborative inquiry shows the same pattern. A 2024 K–12 systematic review examined technology-enhanced inquiry and emphasised the role of teacher guidance, task structure, interaction and scaffolds. ¹¹ Putting students in a shared online space does not ensure that they ask productive questions, evaluate evidence or integrate perspectives.

Support can take many forms: a worked example before an open simulation; prompts that fade; feedback linked to a strategy; roles that distribute participation; a reflection after a choice; a teacher dashboard that surfaces reasoning rather than mere activity. The support should make the target thinking more likely, then recede when learners can carry it.

Compare against a real alternative

Claims about educational technology are only as meaningful as their comparison. “Students improved after using the app” cannot separate the app from practice, additional time, teacher attention, novelty or normal development.

A serious comparison asks:

• Did both groups spend similar time on the target activity?
• Did the technology group receive more feedback or smaller class ratios?
• Was the comparison ordinary practice, no instruction or an equally strong nontechnical method?
• Was the outcome aligned only with the software’s exact tasks?
• Did performance persist after a delay and transfer to a new context?
• Were learners who lacked devices, connectivity or prior digital experience represented?

A 2024 systematic review of validity criteria in technology-enhanced-learning research found fragmented use of validation frameworks. Criteria were rarely applied consistently, and conclusions varied with method. ⁹ That makes transparent outcome definitions, preregistration, attrition reporting and reproducible analysis especially important.

A five-question product review

Before adopting or building a feature, ask:

Goal: Which durable capability should change?

Activity: What will the learner do and think, rather than merely click?

Support: Which feedback, representation or scaffold makes that activity stronger?

Comparison: What happens without the technology under equally credible conditions?

Transfer: Does the gain survive a delay, a new task or removal of the tool?

This review keeps engineering metrics and educational outcomes in their proper places. Reliability, latency and adoption matter because a broken or unused tool cannot help. They are necessary operational evidence, not substitutes for learning evidence.

Pilot the mechanism, not just the software

Procurement and product pilots often ask whether users like a tool and whether it can be deployed. Those are necessary questions. A stronger pilot also writes down the proposed learning mechanism before launch.

For example: “The simulation will let learners vary one parameter, predict the result and receive feedback on the causal explanation.” That statement identifies an activity and a measurable outcome. It can be tested against an equally credible alternative. By contrast, “The simulation will increase engagement and transform learning” is too broad to falsify.

During the pilot, collect evidence along the chain. Check whether the affordance worked reliably, whether learners performed the intended activity, whether scaffolds were used as expected and whether the target capability changed. Interview learners about barriers that telemetry cannot interpret. Record implementation cost and educator workload, because a positive effect that depends on unsustainable support will not travel cleanly.

Decide in advance what would count as success, no effect or harm. A transparent stopping rule makes it easier to retire an attractive feature when its mechanism fails. It also makes it easier to improve a modest feature when the mechanism is sound but delivery is weak.

What the evidence does not show

The evidence does not show that technology is neutral in every sense. Tools shape attention, access, privacy, participation and whose language or knowledge is represented. Nor does it show that all media are interchangeable: some representations and interactions are genuinely difficult without particular technologies.

It does not show that a positive average effect will survive local implementation. Training, curriculum fit, infrastructure, support and opportunity cost can change the result. Meta-analyses combine studies with different technologies, publication periods and comparison conditions. Second-order meta-analysis adds another layer of distance from the original intervention. ^{1 4}

The evidence also does not support judging learning by engagement alone. A compelling system can increase time on platform while encouraging shallow activity. Conversely, a brief tool may enable a productive explanation and then get out of the way.

Finally, saying that learning activities matter is still incomplete. Prior knowledge, relationships, language, culture, health and material conditions shape whether an activity is possible and what it means. The activity is a better explanatory unit than the device, not a complete theory of education.

Start with the verb

The serious case for educational technology is not that screens modernise learning. It is that carefully chosen tools can enable worthwhile activity, provide timely support and widen access to experiences that were previously scarce.

That case becomes stronger when claims become narrower. Name the capability. Describe the activity. Explain the support. Compare it fairly. Test whether the learner can still perform later and elsewhere.

Technology enables. Learners learn. The work of design is to connect the two without confusing one for the other.

References

1. Sailer, M., Maier, R., Berger, S., Kastorff, T., & Stegmann, K. “Learning activities in technology-enhanced learning: A systematic review of meta-analyses and second-order meta-analysis in higher education.” Learning and Individual Differences, 112, 102446. Source.
2. Chi, M. T. H., & Wylie, R. “The ICAP Framework: Linking Cognitive Engagement to Active Learning Outcomes.” Educational Psychologist, 49(4), 219–243. Source.
3. Clark, R. E. “Media will never influence learning.” Educational Technology Research and Development, 42, 21–29. Source.
4. Tamim, R. M., et al. “What Forty Years of Research Says About the Impact of Technology on Learning: A Second-Order Meta-Analysis and Validation Study.” Review of Educational Research, 81(1), 4–28. Source.
5. Schmid, R. F., et al. “The effects of technology use in postsecondary education: A meta-analysis of classroom applications.” Computers & Education, 72, 271–291. Source.
6. Means, B., Toyama, Y., Murphy, R., & Baki, M. “The Effectiveness of Online and Blended Learning: A Meta-Analysis of the Empirical Literature.” Teachers College Record, 115(3), 1–47. Source.
7. Sung, Y.-T., Chang, K.-E., & Liu, T.-C. “The effects of integrating mobile devices with teaching and learning on students’ learning performance: A meta-analysis and research synthesis.” Computers & Education, 94, 252–275. Source.
8. Hooshyar, D., et al. “The effectiveness of personalized technology-enhanced learning in higher education: A meta-analysis with association rule mining.” Computers & Education, 223, 105169. Source.
9. van Haastrecht, M., Haas, M., Brinkhuis, M., & Spruit, M. “Understanding validity criteria in technology-enhanced learning: A systematic literature review.” Computers & Education, 220, 105128. Source.
10. Edisherashvili, N., et al. “Challenges in Promoting Self-Regulated Learning in Technology Supported Learning Environments: An Umbrella Review of Systematic Reviews and Meta-Analyses.” Technology, Knowledge and Learning, 29, 1809–1830. Source.
11. Amarasinghe, I., et al. “Technology-Enhanced Collaborative Inquiry in K–12 Classrooms: A Systematic Review of Empirical Studies.” Science & Education. Source.
12. Freeman, S., et al. “Active learning increases student performance in science, engineering, and mathematics.” Proceedings of the National Academy of Sciences, 111(23), 8410–8415. Source.