Sovereignty is more than just digital sovereignty. When we talk about the challenges of choosing European alternatives to American tech, we overlook factors that can have long-term implications for human sovereignty, such as perseverance, self-confidence, memory, independent thinking, and critical thinking.
The important discussions about digital sovereignty are spreading across the EU, businesses, and among experts in technology, data ethics, and cybersecurity. In this context, European alternatives to American language models are naturally highlighted as a critical choice. But the conversation often ends with a complaint about these very European alternatives.
“But they’re just not as good as the American ones,” is a common refrain from both tech optimists and employees who have grown accustomed to language models that solve tasks quickly.
“They make things too difficult and don’t just solve the task,” or “I have to maintain the overview or engage too much with the content myself,” are other frequent objections.
These are valid points, but perhaps the criticism itself is a sign that we are evaluating artificial intelligence using the same criteria we’ve applied to technology for decades. In doing so, we may not only surrender our digital sovereignty but also our human sovereignty. While we debate data, infrastructure, and technological independence, we should also question how much of our human capacity and coherence we are willing to outsource to (American) language models.
The interesting question may not be whether Mistral, Lumo, Ecosia AI, or other European language models are as good as ChatGPT, Gemini, or Claude. The interesting aspect is that we once again find ourselves in a situation where we measure technology by efficiency, convenience, and speed, while research is beginning to ask what these gains mean for humanity in the long run.
Here are five insights worth considering, which extend beyond the frictionless methods of the attention economy.
Persistence is Challenged
A new study from the US and UK showed that people who spent just about ten minutes using an AI chatbot for reading and math tasks subsequently performed worse when solving similar tasks alone. At the same time, there was a noticeable change in the participants afterward: They gave up faster than expected. Their persistence had, in other words, been altered. The researchers concluded that AI can not only affect our performance but also our willingness to stick to a task when it becomes difficult.
One criticism of European alternatives is that users must drive the answer and the task forward to a greater extent than with American models. That requires persistence.
Eroding Belief in our Own Abilities
The Canadian-American psychologist Albert Bandura described in the 1970s how people’s belief in their own abilities develops through experiences of mastering challenges. It is not success itself that strengthens our agency, but the experience of struggling with something difficult and gradually succeeding.
What happens to our sense of mastery if we increasingly receive the solution before we’ve even had the chance to struggle toward it or are even met with suggestions for improvement?
A side effect of American language models is that they often suggest improvements and optimizations themselves: “Should I give you a suggestion on how to become even sharper?” “Would you like me to optimize your message?”
This challenges our belief in our own abilities when we are not only prevented from struggling with what initially seems difficult but are also constantly met with suggestions for improvement.
Memory and Learning Must be Trained
A recent study from MIT showed that people who wrote essays with the help of ChatGPT had more difficulty remembering their own work afterward. They also performed worse when later writing without assistance. The study, led by researcher Nataliya Kosmyna from MIT Media Lab, examined the cognitive consequences of using AI in the writing process.
Like other studies, it suggests that when we write, argue, structure, and rewrite, we simultaneously train our ability to think. If we skip these steps because technology can deliver the result faster, we may risk losing more than time. It may take longer to use a European language model, but in doing so, we also train our ability to think.
Cognitive Outsourcing
We already use many digital apps to make daily life easier: calendars, notes, GPS, calculators. But several researchers point out that generative AI differs from previous technologies by not just supporting thinking but increasingly taking over parts of it.
When we use European language models, we are, to a greater extent, challenged to think for ourselves.
From Experience to Realization
The concept of Bildung (character formation) is often based on using children’s and young people’s experiences to create insights and shared understanding. Adults grew up with this, but children are particularly vulnerable to potentially missing out on this process. With American language models being smoother and more user-friendly, there is a risk that they will be chosen over European ones if children are not helped to choose the European alternatives.
Australian researcher Kristy Armitage points out that the dilemmas of using generative AI in relation to children and young people are still underexplored. Unlike adults, who have developed critical thinking, children still need to learn important skills. But how do you develop critical thinking, problem-solving, argumentation, and reflection if increasingly large parts of the process are handed over to a system that solves everything for you?
Research does not Point in one Direction
The history of technology is not just a story of loss. It is also a story of new opportunities.
When calculators became widespread, many feared that children would lose their mathematical skills. However, experience showed that technology could strengthen learning if introduced at the right time. First, you learn to calculate. Then, you use the calculator.
Sam Gilbert from University College London warns against concluding that AI necessarily weakens our thinking. The challenge is not whether we should use artificial intelligence or not. The challenge is rather how we use it – whether we use it as a replacement for our own thinking or as a tool to enhance it.
Perhaps this is why European language models deserve more attention than they often receive. Because they point to a possibility rarely discussed the preservation of human sovereignty.
Sources
Photo: Unsplash.com
