My work focuses on turning responsible-AI research into technologies, evidence and organisational practices that can be evaluated in real settings. My distinctive contribution lies in building the research environments, teams and industry partnerships through which this translation becomes possible.
As founder and director of SFI MediaFutures, I initiated and lead a national research-based innovation ecosystem connecting the University of Bergen with leading Norwegian media organisations, technology companies and research institutions. The centre develops and evaluates responsible technologies for recommendation, journalism, fact-checking, media production, accessibility and democratic resilience.
~NOK 300mMediaFutures funding and in-kind contributions
25+public MediaFutures tools and demonstrations
150+peer-reviewed publications
NOK 650m+competitive funding secured or co-secured
Selected innovation and impact cases
Innovation ecosystem
SFI MediaFutures
MediaFutures is Norway’s first research centre dedicated to responsible media technology. It connects academic research with the operational needs of broadcasters, news organisations, fact-checkers and media-technology companies.
Long-term research-industry collaboration under the Research Council of Norway’s SFI scheme.
Research and innovation across personalisation, content production, language technology, media forensics and accessibility.
A growing portfolio of demonstrators, pilots, datasets, models and industry-oriented studies.
MediaFutures researchers and TV 2 have evaluated recommendation approaches in industry-relevant and live environments, including calibrated popularity-bias mitigation and sequential recommendation.
Methods designed to improve exposure to less-popular catalogue content while preserving relevance.
Evaluation that goes beyond offline accuracy to include clicks, consumption and real-user behaviour.
Editor-facing tools for discovering relevant and related content.
Work with Faktisk.no and other partners has produced tools and methods aimed at reducing the time and specialist knowledge needed to assess audiovisual material circulating online.
Tank Classifier: recognition of tanks and artillery vehicles in conflict imagery.
Language Checker: spoken-language identification, transcription, translation and context support.
Research resources for detecting and analysing synthetic or manipulated visual content.
AIModerator demonstrates how a multimodal conversational agent can help audiences verify, compare and contextualise political claims while a live-streamed discussion is unfolding.
The case shows how responsible AI can be treated not only as a governance principle, but as a concrete design and evaluation problem in a democratic setting.
Research through RE-AIMED has examined a prototype combining recommender functionality with structured documentation for calls to medical communication centres.
The work moves recommender-system research into a high-consequence public-service environment and evaluates operator performance, workflow integration and human control.
MediaFutures teams have developed tools for related-content discovery, event extraction, sports commentary, knowledge-supported journalism and accessible media experiences.
These projects are intended to augment professional judgement and improve workflows rather than replace journalists and editors.