School IRL: combined technology-driven learning and student safety
Do you remember your first school trip and how exciting it felt gaining new knowledge outside of a conventional classroom setting? Students in disadvantaged positions may not ever experience that feeling considering field trips are the first element schools eliminate during budget cuts. To address this, we proposed School IRL, an app combining AR and AI technologies to enable real world experiences to students during classroom lessons, all with the goal to democratize experiential learning.
While developing the app concept, my contribution related to Evaluation and Governance stages, where I conducted a survey with 21 students to test whether our idea addressed students’ needs and identify improvement opportunities after users interacted with the prototype. Moreover, I assessed the results against governance and security issues and decided on responsible design choices.
Important decisions
An important decision that impacted the evaluation was the realization that building an actual AR tool and testing it in a real classroom setting was infeasible due to financial constraints and need for advanced technology. This forced us to constrain our evaluation to a prototype with fixed projections of predetermined lessons, simulating what a capture would look like. These design constraints made me question how we could draw accurate conclusions about user behaviour if they could not interact with the app as intended. I recognized that in this situation we could not analyse student performance based on conventional vs AR-assisted lesson comparisons. So, I decided to focus on perceived adoption intention instead, based on the technology adoption model dimensions, usefulness and ease of use (Santini, 2024). If the users showed positive engagement after interacting with the prototype, I could draw conclusion from their feedback and analyse if there was an adoption intent.
A second important decision I made was in the analysis phase, when deciding whether we should scale the app’s features based on user feedback or whether to constrain AI input to specific areas of the app to increase the safety of use. The results of the evaluation showed that users wanted several interactive features like voice narration, a chatbot and personalised explanations, however I recognized that these additions would make it harder to control student-facing AI output and interactions. If our team moved forward with user suggestions and embedded AI in student-facing interfaces there would be a Blackbox effect with limited oversight over AI input in lessons. Another aspect which I underestimated was the influence of legislation. Considering that the target audience of School IRL are minor students the app needed to comply with underage user requirements (General Data Protection Regulation, 2020). Consequently, we decided not to add a chatbot element and have minimal AI input in student-facing app components. The lesson I learned from this decision is that there needs to be a trade-off between user requests and safety and that ultimately the legal space sets the boundary to what can be achieved. Surprisingly while I expected all these safety regulations to constraint the concept for School IRL, they brought out new perspectives and forced our team to focus more on student wellbeing at the centre of the system. My mindset shifted from creating value for students through technology to placing more weight on the safety of the user experience.
More changes after feedback
The first changes to the app I proposed were based on AI Act and GDPR legislation and focused on protecting personal data and limiting AI to lesson development. However, as I was reading about concrete examples of students using their mobile devices during class and highlighted risks and benefits, I recognized that the key element for the success of such initiatives was teacher engagement (Tao et al., 2026). This insight made me realise that the implementation of the app depended on designing the system around teachers as the people in the loop. If teachers had control to override AI outputs in lesson design and supervise how students used the app during class, governance risks would decrease. I learned that value should not only be centred around the target user but also around other stakeholders’ who have a strong influence on the entire value delivery process. Consequently, the logic moving forward changed, focusing on designing the app in a way that would allow teachers to intervene in the student-app interactions.
Collaboration
Collaboration with my teammates taught me the importance of alignment and transparent communication when working on strategy development projects. I learned that each individual decision at any stage has a strong influence on the overall direction of the strategy and that it is important to collectively agree what areas of the app need to change and in what way. I also learned that clear communication is the base for alignment towards a shared vision, because each team member may have a different perspective of the app’s value proposition and on what levels technology is embedded in it. I initially viewed the AI interface as a layer throughout the whole system while other team members viewed it as only applicable to individual parts such as lesson generation or assessment. Therefore, it was important to discuss different expectations and decide on clear areas where AI would be present and how this would impact the value added to the users.
New perspective on AI and business strategy
In the beginning of the project, I was convinced that integrating sophisticated technologies like AI and AR into various dimensions of an app adds more depth to its services and subsequently increases the value for users. In my mind the purpose of the technologies was to improve user interaction at every stage, so it made sense finding multiple different use cases for them. This was also the logic I initially followed when analysing user feedback, thinking that adding different features would make the final product more appealing. In contrast, I learned that while users my request certain features their needs sometimes conflict with key security principles and legislation. It is better to focus on a few technology applications and be able to clearly track the process through which they add value to users, instead of having multiple AI features and losing track of interactions. The lesson I am taking with me from the course and AI Strategy Lab is that less is sometimes more. AI should not be implemented everywhere if oversight cannot be guaranteed and that there sometimes needs to be a compromise between what is desired and what is safe for the user.
References
General Data Protection Regulation. (2020, February 10). Conditions applicable to child’s consent in relation to information society services – General Data Protection Regulation — GDPR. https://guides.gdprinfo.eu/en-article-8
Santini, F., Sampaio, C. H., Rasul, T., Ladeira, W. J., Kar, A. K., Perin, M. G., & Azhar, M. (2024). Understanding students’ technology acceptance behaviour: A meta-analytic study. Technology in Society, 81, 102798. https://doi.org/10.1016/j.techsoc.2024.102798
Tao, S., Lan, M., Wang, M., & Li, H. (2026). Potential risks of generative artificial intelligence integration into K-12 education: A scoping review. Computers and Education Artificial Intelligence, 10, 100561. https://doi.org/10.1016/j.caeai.2026.100561
