{"id":24783,"date":"2025-06-24T10:59:36","date_gmt":"2025-06-24T08:59:36","guid":{"rendered":"https:\/\/hub.hslu.ch\/informatik\/?p=24783"},"modified":"2026-02-09T13:58:23","modified_gmt":"2026-02-09T12:58:23","slug":"young-voices-for-fair-ai-research-that-connects","status":"publish","type":"post","link":"https:\/\/hub.hslu.ch\/informatik\/young-voices-for-fair-ai-research-that-connects\/","title":{"rendered":"Young Voices for Fair AI: Research That Connects"},"content":{"rendered":"\n

<\/h2>\n\n\n\n

Deutsche Version<\/em><\/a><\/p>\n\n\n\n

By Andreas Waldis<\/a><\/p>\n\n\n\n

Fog, rain, snow. And in the middle of it: 25 young researchers from across Europe spending a week exploring societal biases in AI systems.
Instead of conference routines, the program was shaped by workshops, personal conversations, and shared activities.<\/p>\n\n\n\n

KnitTogether \u2013 A Research Week in the Mountains<\/strong><\/strong><\/summary>
\n

KnitTogether<\/a> is a format for early-career researchers that fosters exchange on socially relevant questions related to artificial intelligence. The first edition took place in May 2025 in the Appenzell village of Wildhaus \u2013 organized by Lucerne University of Applied Sciences and Arts, the University of Stuttgart, and the University of Hamburg.<\/p>\n\n\n\n

At its core was the topic of bias in language models \u2013 societal distortions that can be reflected in AI systems, such as gender inequality. Twenty-five participants from across Europe discussed how fair and transparent modeling can be achieved.<\/p>\n\n\n\n

Instead of lectures, the program featured workshops, peer reviews, mini-talks \u2013 along with shared cooking, games, and hiking. This made the exchange not only productive but also personal and connecting.<\/p>\n\n\n\n

The week was made possible thanks to the support of Lucerne University of Applied Sciences and Arts, the University of Stuttgart, the University of Hamburg, and the Society for Language Technology and Computational Linguistics.<\/p>\n<\/div><\/details><\/div>\n\n\n\n


Three Takeaways from a Week of Pressure-Free Research<\/strong><\/strong><\/h3>\n\n\n\n

<\/p>\n\n\n\n

1. Trust before Data<\/strong><\/strong><\/h2>\n\n\n\n

Good research begins with relationships: with open exchange, shared experiences, and the freedom to show vulnerability. KnitTogether<\/a> made clear how essential safe spaces are \u2013 especially for young researchers often caught between pressure and high expectations. Many participants used this protected environment to figure out what they truly want to focus on in their research and where they want to grow personally.<\/p>\n\n\n\n

\n
\"\"<\/a><\/figure>\n\n\n\n
\"\"<\/a><\/figure>\n\n\n\n
\"\"<\/a><\/figure>\n\n\n\n
\"\"<\/a><\/figure>\n\n\n\n
\"\"<\/a><\/figure>\n<\/figure>\n\n\n\n


2. AI Needs Critical Awareness<\/strong><\/strong><\/h2>\n\n\n\n

At the center of the week was the topic of bias in NLP (Natural Language Processing) \u2013 biases in language models that reflect or even reinforce societal inequalities. Whether it\u2019s gender-neutral language, cultural stereotypes, or the dominance of English: These distortions are not merely technical errors; they reflect societal power structures embedded in the training data. And these structures are constantly changing. The participants agreed: Research must account for this \u2013 and above all, make bias visible so users can engage with these models responsibly and confidently.<\/p>\n\n\n\n

Making Bias in AI Visible Means Taking Responsibility<\/strong><\/strong><\/summary>
\n

Language models are based on massive amounts of text and in the process also adopt societal biases present in these training data. These biases can relate to gender, origin, language, or social role models. They are not merely technical side effects, but reflections of societal power structures \u2013 and can be further amplified by AI systems.<\/p>\n\n\n\n

\n