European factfulness AI

March 2025

Post-truth narratives increasingly dominate social and political debates, where opinion often outweighs evidence. Even critical thinkers struggle to distinguish fact from misinformation when venturing outside their areas of expertise.

The aim of misinformation is often not to make people believe, but to make sure that nobody believes anything. In the long run, this deprives people of the power of thought and judgement, as has been pointed out by Hannah Arendt, a political theorist and philosopher.

Was Hitler a communist? Who started the war in Ukraine in 2022? Are migrants more criminal than the domestic population? In which EU countries is the freedom of speech guaranteed?

Seemingly straightforward questions to which most of us would have an immediate answer. However, querying the internet, social media or various AI tools does not necessarily provide the facts needed to answer the questions. Blunt opinions, fake news and actual evidence often seem to be mixed in the answers. A mix that can even differ depending on who asks the question and how it is asked.

Following my earlier blog post on “Escape the Pied Pipers: Back to evidence (Öffnet in neuem Fenster)”, I had several discussions with friends and colleagues. All of them raised the issue of agency and the need for a reliable and trustworthy source to turn to for facts and fact-checking. There seems to be a need for something like a modern, AI-powered equivalent of the national encyclopaedias of the past.

The key question, of course, is who will ensure that such a tool is a trusted source? And could it not only act as a recognised reference point, but also proactively communicate facts and figures?

Imagine an AI tool you can trust

To a large extent, it can be argued that general AI tools such as DeepSeek, Gemini, Grok, ChatGPT, etc. already allow us to do the fact-checking we are looking for. But are we always sure that we can trust them?

To what extent are their results simply a reflection of what is available online, rather than the most accurate information? Is there not a risk that they reflect the most popular answers rather than the most correct ones? In the worst case, could some algorithms even be tuned to favour certain narratives?

A European factfulness AI would not indiscriminately scrape online content, but rely on accredited sources, including peer-reviewed scientific journals, independent research organisations and official public institutions. Think of it as a modern AI-powered encyclopaedia, where every answer is fact-checked and transparently sourced.

Instead of harvesting everything that is out there, this AI would work on the basis of validated input, ensuring fact-based, easy-to-understand answers that also include contextual information and clear references.

European network of independent research organisations

Of course, such an encyclopaedic AI tool could be tweaked and twisted in many directions. Its trustworthiness would depend on the organisations behind it. One non-negotiable key criterion should be independence from economic and political interests. It would probably need to be based on a network of institutions rather than just one.

Some ideas for possible institutions which could be considered for such a network mentioned in the discussions include the Leibniz Association, Max Planck Institutes or Helmholtz Association in Germany, Luxembourg Institute of Socio-Economic Research (LISER), Netherlands Organisation for Applied Scientific Research (TNO), French National Centre for Scientific Research (CNRS), VTT Technical Research Centre of Finland, Swedish Research Council (Vetenskapsrådet) or Stockholm Environment Institute (SEI). One might consider whether to involve leading foundations such as Nobel Foundation, Open Society Foundations or Gapminder Foundation. This list is certainly not exhaustive, as many other institutions could contribute. In addition, there are a number of private and non-profit organisations already active in the field that could be relevant partners in such an endeavour.

Ensuring neutrality and transparency will be crucial. A governance framework must define how sources are selected, how facts are verified and how bias is minimised. Public trust will depend on maintaining clear, open processes and a commitment to scientific integrity.

Add a communications arm

The role of such a network of trusted organisations could go beyond simply building and feeding a European AI on facts. A dedicated communication strategy could help to actively disseminate facts and figures to the general public. Combining existing efforts to counter post-factual narratives is just as important as providing a reliable, low-barrier reference point for fact-checking.

The communication would also need to address and clarifying the tensions between facts, opinions, personal perceptions and feelings.

Start with EU KI and innovation ambitions

Who’s supposed to pay for that, who has that much money? Certainly, all this will require resources - resources that are currently in short supply. But a European approach could also help to boost AI development within the EU and clearly differentiate it from market- or policy-driven AI developments elsewhere. Why not consider integrating this idea into the European InvestAI initiative or the upcoming Horizon Europe research and innovation funding under the EU's post-2027 multiannual financial framework?

In addition to EU funding, relevant stakeholders would need to come together to prepare for the development of a European factfulness AI. Trusted, politically and economically independent organisations should take the lead in this initiative.

To answer the question

This approach could be one step forward in ensuring that future debates are based on facts rather than opinions. By promoting a culture of factfulness, this AI could help restore trust in public debates and prevent facts from being drowned out by misinformation. Trustworthy information is the foundation of an informed democracy. It allows people to get reliable answers to the questions above, and many more and more complex questions.

Answers to the initial questions:

• No, Adolf Hitler was not a communist.

• The war in Ukraine in 2022 was started by Russia, which launched a full-scale invasion of Ukraine on 24 February 2022.

• No, migrants are not inherently more criminal than the domestic population. Multiple studies from various countries show that migration status itself is not a predictor of criminal behaviour.

• While legally guaranteed in all EU countries, the practical implementation of freedom of speech varies due to differences in legal traditions, political landscapes, and media environments. Some key distinctions include: Countries such as Sweden, Finland, Denmark, the Netherlands, and Germany consistently rank high in press freedom indexes, with strong legal protections against government interference. In some EU member states, concerns have been raised about political interference in media, legal threats against journalists, or the misuse of defamation laws to silence criticism.

by Kai Böhme and Wilhelm Geilen