Future-oriented AI risk management

Gotcheva, N. & Wessberg, N., (2024)

Integrating foresight and risk assessment traditions is beneficial in studies of new emerging risks, including the risks arising from Artificial Intelligence (AI). However, potential future AI risks, including social harms, may require a long timeframe for the risk implications on society to manifest. Further, riskinformed decision-making related to AI technologies is challenging because an understanding of both direct and latent risks is needed to identify hazards and capture the full risk landscape in great uncertainty. In this paper, we propose a new method for continuous horizon scanning of AI technologies development and societal implications – Signal post series. The goal is to capture and analyze weak signals and to co-create knowledge for anticipating risks and opportunities related to AI and its various applications in the future. The Signal posts are essentially thematic future-oriented short texts, based on weak signals, identified, and interpreted in joint researchers’ workshops and from media. In terms of impact, Signal posts include focus on ethical and societal considerations and thus contribute to strengthening future oriented responsibility and resilience of AI systems. In addition, we address potential opportunities related to issues of acceptability and desirability of the AI systems. There are pressing concerns for a harmonized global approach and regulation for assessing and managing AI-related risks, especially in terms of human values, norms, and overall culture. The novelty of Signal post method for integrating horizon scanning and risk management is that it holds strong emphasis on safety, ethics and responsibility while considering risks and opportunities related to AI technologies in the future. This approach has the potential to complement traditional risk management methods to enhance capacities across sectors, value chains and layers of uncertainty to prepare for the future and for co-create sustainable futures.

Future-oriented AI risk management
International Association for Probabilistic Safety Assessment and Management (IAPSAM) Topical Conference: Artificial Intelligence (AI) and Risk Analysis, PSAM2023, October 23-25, 2023. University of Illinois at Urbana-Champaign, 10 p.