Recent Artificial Intelligence (AI) advances have sparked heated discussions across various industries.

Media representations have ranged from scenarios of super-intelligent computers capable of unthinkable conquests to apocalyptic scenarios of possible extinction.

This hype comes from the prospect of using AI to solve any problem without having specific skills. Extreme and opposite reactions towards AI are therefore emerging, some potentially harmful. On the one hand, excessive trust in artificial intelligence can lead to the use of incorrect suggestions in delicate contexts and even cause atrophy of our skills over time; on the other hand, fear and negative attitudes towards AI could compromise the vast social benefits that the use of these technologies can generate.

The hype and lack of information about the current limitations of AI, such as its “hallucinations,” distract from other important issues, such as the impact on inequality, copyright, privacy, and the long-term effects of such pervasive systems, connected both to social media and to mobile devices and the web, on our personal choices and on the community in which we live.

Projects such as Courage (http://upf.edu/web/courage), funded by the Volkswagen Foundation, are therefore crucial, which aim to understand the complex effects of AI, propose targeted approaches, promote the common good, and, above all, develop educational interventions for conscious and safe use.

Traditionally, the task of AI was optimisation. Examples in education could range from pathfinding to error checking or from learner modelling to content adaptation. However, AI is now entering the realm of creation, and is no longer constrained to optimisation. With this comes complexity that also carries doubt with it. The notions of right and wrong no longer have clearly defined boundaries, AI has entered the world of opinions and value judgements.

And of course, from the perspective of Technology-Enhanced Learning, this is welcome. In our society, learning to work with opinions and values - learning to think critically - has long become an educational objective even from early school years. However, effective critical thinking is not purely about facts. To be convincing it needs to include a reflection on epistemology - what is one's perception of knowledge and how it emerges. Applications of AI are no longer restricted to cases when we have thousands of training examples, deterministic oracles, or even the luxury of absolute truth.

As a consequence, what we witness now is that even established research methods fall short when it comes to making important advancements in human knowledge (although marginal performance improvements are daily news). This might indicate the need to go back to fundamental questions about the scientific method - to review our hidden assumptions and make them explicit, to make an effort to engage in transparent and reproducible science.

One place to look for potential answers is interdisciplinary research. Beyond the well-established combination of educational research, AI and gamification, I am part of a research effort that has joined forces with professional psychologists, scholars from the humanities, and museum practitioners to try to redefine how we can benefit from AI (and technology in general) to make our society more resilient through education and awareness.