campus a: Journalistic feedback for Austria's first nationwide school newspaper.
At a glanceStarting point
Thousands of texts, one small editorial team.
The task was complex: in a time of democratic decline, Bernhard Salomon wanted to bring the ethical and craft values of journalism closer to schoolchildren. Unlike earlier school newspapers, which each school produced on its own, a shared nationwide platform was to be created. The central challenge: a small editorial team cannot possibly give professional feedback on thousands of articles a day. „Wir brauchen einen Chatbot, der dieses Feedback professionell, glaubwürdig und belastbar an die Zielgruppe der 12- bis 18-Jährigen transportieren kann", explains Bernhard Salomon. On top of this came the difficulty of simplifying the basic principles of journalism so that they are understandable even for young people without prior knowledge, without losing quality in the process.
Solution
Feedback that 12 to 18 year olds accept.
Bernhard Salomon had a sobering experience with traditional AI consulting firms: „Nachdem ich mich vier Wochen intensiv damit beschäftigt hatte, wusste ich mehr über KI als alle professionellen KI-Consulting-Unternehmen." The breakthrough came through working with Niclas Barthélemy (CEO and Co-Founder of scoreprise.AI), who took a different approach: „Die haben gleich gesagt: Wir kennen uns auch nicht aus, aber wir lernen jetzt gemeinsam mit dir, wie wir dein Problem lösen könnten. Das war wirklich einer der drei großen Lichtmomente."
Together they developed a tailor-made chatbot with two main functions: on the one hand it gives detailed feedback on schoolchildren's texts, for example pointing out that the opposing view is missing, that too many passive sentences are used or that facts and opinion are mixed. On the other hand it revises texts journalistically on the spot, to a quality that goes beyond standard AI tools such as ChatGPT.
Task
Feedback for thousands of texts
A small editorial team cannot review thousands of articles a day to a professional standard. The feedback had to reach 12 to 18 year olds in a way that was credible and reliable.
Feedback
What the text is missing
The chatbot names concretely what is missing: the opposing view, too many passive sentences, mixed facts and opinions.
Revision
Journalistic, not just linguistic
On request it revises the text journalistically, to a quality that goes beyond standard tools.
The aha moment
„Der Durchbruch war, als ich verstanden habe, ich muss die KI wie einen begabten Schüler behandeln und auch wie einen Menschen, nämlich erkennen, wo steht die im Bereich Text, welche klassischen Fehler macht sie, und eben darauf aufbauen und nicht versuchen, ihr was aufs Auge zu drücken.“
Two insights came first: that most AI consultants were not really any help, and that AI does not work linearly like people but does everything at once. From this came the decision to build its own tools.
Result
Journalistic feedback in three quality levels.
The project is currently in a pilot with one school with a journalism focus. Bernhard Salomon observes that the rewrite function is used considerably more often than the feedback function: a sign of convenience, but also an opportunity to learn. The articles created with the tailor-made chatbot are now barely distinguishable from manually written texts, whereas articles created with standard AI tools remain recognisable.
„Ich bin teilweise tief in meiner Seele verblüfft, was die KI alles liefert, aber auch dann wieder am Punkt, wo ich mir denke: Wie ist so ein Schwachsinn möglich", Bernhard Salomon describes the current state honestly. The development is a long-term process: „Ich hoffe, dass wir in drei Jahren so weit sind, dass wir ein sinnvolles Instrument haben, das an die Zielgruppe angepasst ist und Spaß macht."
„Die sind nicht belehrend an die Sache herangegangen, um zu demonstrieren, was sie alles wissen, sondern sie sind lernend daran herangegangen. Jeder, der mir erzählt, er weiß alles über KI, das glaube ich einfach nicht.“
More use cases.
All use cases
M&A advisory
KERN Unternehmensnachfolge: invoices, proposals and NDAs from a Teams message.
Networking and professional development · Project
Leaders of AI: 60 specialists build five specialised AI teams in one day.
Higher education and research · Project
Leuphana: How human may a chatbot appear? A study with 500 students.
Next step
Your use case, shown on a comparable one.
Cancellable monthly · Operated in Germany