Interdisciplinary Expert Pool for NLU

Important results from the project

** maskinöversatt ** 1. Project Consortium: Established a consortium of 36 experts from various fields for AI analysis and discussion. 2. Proof of Concept: - Conducted an experiment to evaluate LLM´s behavior. - Developed a small dataset for detection of misogyny in Swedish. 3. Expert training: Erbjöd AI-utbildning to experts, enabling their research and knowledge dissemination, diversified the AI pipeline. 4. Long-term results: Laid the foundation for scalable expert involvement in AI development, fostered interdisciplinary collaboration and addressed research issues.

Expected long term effects

** Denna text är maskinöversatt ** The project explored how best to integrate a variety of skills, knowledge and experience in the production and deployment of AI models. Experiments demonstrated the need for interdisciplinary expertise to successfully analyze data, which is critical to building AI. This approach has potential to provide significant assistance to organizations that need to ensure transparency and public trust in AI systems (e.g. public sector).

Approach and implementation

** Denna text är maskinöversatt ** Key components of the implementation included: 1. Regular workshops, training sessions and seminars to promote collaboration and knowledge sharing. 2. Proof of Concept experiment. 3. Training and Capacity Building. 4. Feedback sessions with industry and public sector representatives that guided project adjustments and ensured practical relevance and alignment with stakeholder needs.

External links

Lista över medlemmar i den tvärvetenskapliga expertgruppen, projektets omfattning och utbildningsmaterial Dataset för detektion av misogyni, annoterat av experterna och utvecklat för att hantera fördomar i språkmodeller.