Confident Use of AI in Media

The rapid change in media consumption, the development of the media landscape, and the associated challenges for democracy are at the heart of the project FAIRmedia – Fair and Trusted Data Sets for Media Computing.

The project concentrates on the interface of media, artificial intelligence (AI), and data sovereignty and aims to support media producers in overcoming these challenges and strengthening the essential role of media in a democracy.

Toolbox for Trustworthy Datasets

The focus of the project is on developing a set of tools for the efficient creation of trustworthy datasets with both current and archived contents by Austrian media organisations that are used for training AI models.

Aid for Media Producers – Data from Europe

Georg Thallinger, Project Coordinator of JOANNEUM RESEARCH: “The results of the project will enable media producers to independently create balanced datasets based on existing archive material, thereby implementing AI models that live up to the requirements of their present task.”

At the moment, most datasets used in the training of AI are derived from organisations outside the EU, especially in the USA and China. This does not only lead to dependence on global technology companies but also harbours the risk that critical datasets cannot be withdrawn again at any time.

Moreover, there is the danger that AI tools developed outside of Europe may not meet the European standards and legal framework conditions such as the GDPR and future AI regulations.

Creating Fair and Trustworthy Datasets

The project focuses on the current and archived contents of Austrian media organisations in order to ensure a locally sourced and trustworthy base. Its establishment is supported by tools for the analysis of the balance of datasets.

The project team will create so-called no-/low-code tools. These enable the training of AI models without profound programming knowledge. The target group consists of various actors in the media sector including journalists, chief editors, audience managers, and archivists. Thanks to training at the local level, the data remains under the media companies’ own control.

Human-Computer Interaction

“The human-computer interaction team at the St. Pölten UAS ensures the good and intuitive use of the tool developed in the project. We want to enable users to engage in dialogue-like and result-oriented interaction”, explains Peter Judmaier, deputy head of the Media Computing research group at the St. Pölten UAS’ Institute of Creative\Media/Technologies. 

Another goal of the project is to develop AI tools that can be explained in retrospect. They serve to improve users’ understanding of the trained models and provide comprehensible explanations and hints regarding the balance of datasets.

Reinforcing Data Sovereignty

FAIRmedia has the potential to improve data sovereignty in Europe by creating datasets that are fully under the control of Austrian media organisations. The project thus opens up new paths for media organisations to effectively use AI technologies and contribute to the European media data space.

The project is borne by the partners APA, ORF, the St. Pölten UAS, Redlink, the Department of Innovation and Digitalisation in Law (University of Vienna), and the Österreichische Mediathek. The Institut DIGITAL of JOANNEUM RESEARCH is in charge of coordination.

FAIRmedia receives funding from the Federal Ministry for Climate Action, Environment, Energy, Mobility, Innovation and Technology (BMK) as represented by the Austrian Research Promotion Agency (FFG).

• Project FAIRMedia