Synthetic disinformation detection among German information elites – Strategies in politics, administration, journalism, and business [Erkennung synthetischer Desinformation unter deutschen Informationseliten – Strategien in Politik, Verwaltung, Journalismus und Wirtschaft]

Vief, Nils; Bösch, Marcus; Unger, Saïd; Klapproth, Johanna; Boberg, Svenja; Quandt, Thorsten; Stöcker, Christian;

Abstract

Since the technology for generating synthetic media content became available to a wider audience in 2022, the social and communication sciences face the urgent question of how these technologies can be used to spread disinformation and how well recipients are equipped to deal with this risk. Research so far has focused primarily on the phenom- enon of deepfakes, which mostly refers to visual media generated or modified by artificial intelligence. Most studies aim to test how well recipients can detect such deepfakes, and they generally conclude that recipients are rather poor at detecting them. In contrast, this analysis focuses on the broader concept of synthetic disinformation, which includes all forms of AI-generated content for the purpose of deception. We investigate the process of how actors with professional expertise in the field of disinformation try to detect AI-gener- ated disinformation in text, visual and audio content and which strategies and resources they employ. To gauge an upper bound for societal preparedness, we conducted guided interviews with 41 actors in elite positions from four sectors of German society (politics, corporations, media and administration) and asked them about their strategies for detect- ing synthetic disinformation in text, visual and audio content. The respondents apply dif- ferent detection strategies for the three media formats. The data shows substantial differ- ences between the four groups when it comes to detection strategies. Only the media professionals consistently describe analytical, rather than simply intuitive, methods for verification.