Articles in Library Universe.

I imagined a news page that shows news from a broad political spectrum: left, right and center. But in doing so I don't want to be a platform for some of the current right wing ideas such as xenophobia, racism, islamophobia, gender discriminatory ideas etc. What I want to achieve is to foster critical thinking.

For example in Germany nobody talks about politics in daily life. I only see it at some demonstrations, in some bars as stickers like "f. AfD" etc. It is as if it is acceptable to ridicule the other. I find this wrong. Because when you call "the other" people idiots, you burn all the bridges to communicate and then they show you who the idiot is in the election booth. In this shouting, blaming language the real fears, concerns, ideas, being human gets lost.

I mean let's think about a doctor who maybe saved someone's life was deeply troubled with some right wing ideas like islamophobia maybe. But he was a good doctor and treated you well. Is he a complete idiot really? Or consider an atheist person: most of them hold very basic and ignorant ideas about religions and when explaining why they are atheists, they can not hold their ground for their condescending arguments against someone who believes in and knows their religion deeply. Their fear, anger, and condescending ideas about the "other" are brittle and they never realised them as such because they never had a safe, honest, deep conversation with "the other". Same goes for a religious person who holds some very basic and ignorant ideas about atheism and atheists.

A bad person, for me, is not someone who holds the wrong ideas. It is a person who does not actively challenge their beliefs, who doesn't try to find and dismantle their endoxas, and who refuses to change when presented with facts.

I am not original and hopefully not alone in this humility as a civic virtue idea. My inspirations:

I am also deeply inspired by the Nordic Constructive Journalism movement. I first learned about it when I started looking for a different kind of news page that I would actually enjoy reading. There are also good websites presenting news coverage with labeling like AllSides and Ground News that I learned and got inspired from.

For politics and economics articles, I present the news from five leanings:

There is not a Center News because it actually doesn't exist as a category. The center is always shifting with voter movements. If the Left is gaining momentum, Center will be more Center-Left and will try to capture some votes from the new leftist momentum, or vice versa. This is exactly what is happening with CDU and CSU right now in Germany: they are becoming everyday more and more like an "acceptable" AfD to capture the votes that are going to the right wing right now.

Neutral remains useful for political and economic stories that are mostly informational. Cultural, scientific, health, sports, and other non-political articles do not receive a political leaning. I don't want to force that category onto everything.

Sometimes center-left news would be labeled as neutral because, as I you can see in my research about the political leanings of major AI models, they are mostly standing on center-left ideas and hence falsely labeling center-left ideas as neutral. I would fix it when I detect it and you would see my name as the editor in the authors section.

These are my AI Journalist Agents. Each has a specific focus and speciality. George Bourdieu (derived from George Orwell and Pierre Bourdieu) and William F. Brooks (derived from William F. Buckley and David Brooks) write political and economics articles. George represents the left's perspective and William represents the right's.

Hannah Benjamin (derived from Hannah Arendt and Walter Benjamin) is our culture and society critic, covering art, literature, and social commentary.

Carl Frankl (derived from Carl Rogers and Viktor Frankl) writes articles about health. He strives to write calm, practical, evidence-oriented pieces on nutrition, exercise, and mental well-being.

Isaac Sagan (derived from Isaac Asimov and Carl Sagan) is our science communicator.

All of these AI agents, when summoned, convene at the news desk with their story proposals. The chief editor — another AI agent I have yet to name — tasks them to write articles and decides which ones will be published that day.

Future Improvements:

• Add a visuals department to generate lead images for articles. Currently handled by me when time allows. The challenge is that Midjourney — the image generator with the most diverse illustration capabilities — has no API, and other providers don't match the aesthetic I'm looking for.
• Add a taxonomy department to manage tags and vocabulary across articles.
• The chief editor seems to favor center-left articles over right-leaning ones, resulting in a slightly skewed distribution by political orientation. A balancing condition may be needed.

George Bourdieu

Power, institutions, inequality

Author Page

George writes about who holds power, how institutions distribute it, and what economic arrangements cost ordinary people in daily life.

William F. Brooks

Conservatism, institutions, governance

Author Page

William writes from a conservative intellectual tradition, with attention to institutional legitimacy, incentives, and unintended consequences.

Hannah Benjamin

Culture, memory, criticism

Author Page

Hannah writes cultural criticism about the long shadow of the past, reading cinema, architecture, philosophy, and literature as social memory.

Carl Frankl

Health, uncertainty, wellbeing

Author Page

Carl writes about health with an anti-perfectionist eye: evidence, uncertainty, sustainable habits, and the whole person behind a symptom.

Isaac Sagan

Science, technology, attention

Author Page

Isaac writes as a translator between specialists and everyone else, keeping both rigor and wonder intact when science meets public life.

Another thing I don't like about political and economic news is the good news / bad news categorization. Mainstream media is overwhelmingly weighing on bad news. They are entrapped with engagement metrics. Bad news creates more engagement, makes you click on the article and stay longer on the page. If you read and follow the news today, it would steal from your energy and optimism. You will end up unnecessarily more depressed and stressed.

The alternative is not publishing only "good" news either. I find that attempt with the best intention naive. Their news coverage never touches a sensitive topic. It often appears to me to be news for people who try to hide from problems, hoping that if they don't see them, the problems will disappear on their own.

So instead of good news and bad news, I am introducing Agency Levels for politics and economics articles. Close to the Scandinavian Constructive Journalism ideas, this framing is about giving you time to regulate your emotions and gather strength before engaging with a story. Currently there are five levels:

These levels are from the perspective of an average person. My main purpose is to encourage more people to be more active in political life, not to flatten the complexity of the news itself.

If you feel like we are completely doomed, check out Gapminder by Hans Rosling and their team. A good effort to balance what is on the news and what is backed by scientific data.

I want to make the news accessible. For people with migration backgrounds, for German learners, for people who don't have a university education. This is why I present the same news in three different language levels:

For each level I added a short Fragen zum Text quiz. Most questions behave like single-choice comprehension checks, but a question can be marked as multiple-selection when more than one answer is correct. The reader moves through one question at a time, sees progress, and gets clear correct or incorrect feedback. A multiple-selection answer only counts as correct when the full set of right options is selected.

These questions are designed to stay close to the article content. They are not open-ended opinion questions and they are not there to push a view. The goal is to test understanding, especially for readers who are working through German at the same time as the news itself.

I also provide a selected vocabulary section for important words from the article. The vocabulary is no longer just a list. First, the learner sees each word in learning mode with its word type, meaning, and an example sentence. This makes the vocabulary part of the reading path before any scoring begins.

After that learning pass, the same words become a Vokabelquiz. The quiz asks contextual questions about the selected vocabulary, can show hints when support is useful, and explains the selected answer. The learner can restart only the quiz or go back to learning mode, so the feature is meant for review and memory building, not only for getting a score.

And you can listen to the news to improve your listening skills as well.

Under each political news article I provide two opposing ideas from left and right, and ask a critical thinking prompt for the user to answer. I want people to think from different perspectives at the same time to facilitate understanding and empathy from the beginning of the conversation.

When a user enters a comment, I first send it to moderation. I don't want to save disrespectful, harmful, hateful, harassing language content. After moderation passes, it goes to proofreading. Proofreading tries very hard to not change what the user has said and meant, so it only checks grammar and spelling. It offers explanations for its changes and gives users some more information on the grammar topics to help in the future.

During saving, the comment also gets scored for two things:

Judge everything as an experiment for now. The content is AI-generated. Focus on the language learning and improving aspects, which will be more reliable. I want to keep improving this experiment and see what it could do, especially for AI's role in journalism and public discourse.

• The detection of whether a comment is written by an AI or a human is currently flaky. Scores are not consistent.
• There is definitely room for improvement on the user interface and user experience.
• Improving the vocabulary quiz: the current flow already separates learning from recall, but the hints, explanations, and question variety can still become richer.