So, you’ve probably heard the argument comparing the adoption of AI to that of calculators in mathematics - both examples of technologies that support cognition in some way, or do they? The implied statement is “Calculators shortcut mathematical thinking, yet we’ve used them successfully for years. They even support higher-level thinking."  However, we should ask: What do we know about calculators’ impact on math learning? How are they used in practice?  And, what can we deduce from this case about AI adoption? Let's answer the calculator question first, and then move to GenAI:

An excellent review by McNeil et al.,(2025) focuses on the ability at the core of the calculator's question:  Arithmetic  Fluency -  mastery of basic operations within the 1–10 range. They describe the debate regarding the cognitive components underlying this ability: those who argue that reasoning and understanding number relationships  (let's call it 'Making meaning') are sufficient vs. those who argue that practice leading to automaticity (i.e. 'memorisation') is equally essential.

According to the "Meaning only" approach, using calculators contributes to learning because it frees up cognitive resources for more beneficial strategy use,  or offloading ( exporting part of the cognitive workload to a third party).

But it turns out that offloading has its price. Though short-term performance improves, evidence shows a decline in long-term retention.

The review then explores how both meaning-making and memorisation are interdependent and jointly necessary for arithmetic fluency. They provide a thorough and convincing research-based cognitive explanation. Then lay out a series of recommendations for implementation. These include early support, explicit instruction, effective practice (including time-limited practice), and time for strategy comparison, discussion and reflection. The recommendations do not include, you guessed it, using calculators.

Conclusion

Calculators may solve problems "on the spot," but they hinder the development of foundational skills essential for future success. What we don’t practice ourselves does not become a useful ability. Therefore, it is better to use calculators only after human ability has been established (after acquiring arithmetic fluency in elementary school).

And what does this mean for AI?

AI can shortcut far more complex tasks than basic calculation, like summarising texts and writing essays -  skills built over years, during high school and well into higher education. When we use AI to improve performance in a specific academic task, we may impair the development of writing skills. 

This does not mean avoiding artificial intelligence altogether, but if we want to use the calculators case adequately, we should conclude that encouraging students to use AI to reduce workload and enhance outcomes should come only after basic skills have been developed with focused instruction and practice.

We all experience and feel today that after acquiring the basic skills, GenAI opens up a whole world of possibilities for outcome enhancement and productivity enhancement. As educators, our responsibility is to ensure strong foundations and to guide thoughtful, evidence-informed use of AI, avoiding the avoidable traps. One of the most important distinctions when striving to adopt GenAI responsibly in education is to remember the qualitative difference between expert users and those who are still on the track to achieving expertise.

Reference

McNeil, N. M., Jordan, N. C., Viegut, A. A., & Ansari, D. (2025). What the science of learning teaches us about arithmetic fluency. Psychological Science in the Public Interest, 26(1), 10-57. https://journals.sagepub.com/doi/pdf/10.1177/15291006241287726