Schlagwort: Bias

Veröffentlicht am

Revisiting the Trolley Problem

On May 28, 2025, the „Proceedings of the 2025 AAAI Spring Symposium Series“ (Vol. 5 No. 1) were published. Oliver Bendel was involved in two papers at the symposium „Human-Compatible AI for Well-being: Harnessing Potential of GenAI for AI-Powered Science“. The paper „Revisiting the Trolley Problem for AI: Biases and Stereotypes in Large Language Models and their Impact on Ethical Decision-Making“ by Sahan Hatemo, Christof Weickhardt, Luca Gisler, and Oliver Bendel is summarized as follows: „The trolley problem has long served as a lens for exploring moral decision-making, now gaining renewed significance in the context of artificial intelligence (AI). This study investigates ethical reasoning in three open-source large language models (LLMs) – LLaMA, Mistral and Qwen – through variants of the trolley problem. By introducing demographic prompts (age, nationality and gender) into three scenarios (switch, loop and footbridge), we systematically evaluate LLM responses against human survey data from the Moral Machine experiment. Our findings reveal notable differences: Mistral exhibits a consistent tendency to over-intervene, while Qwen chooses to intervene less and LLaMA balances between the two. Notably, demographic attributes, particularly nationality, significantly influence LLM decisions, exposing potential biases in AI ethical reasoning. These insights underscore the necessity of refining LLMs to ensure fairness and ethical alignment, leading the way for more trustworthy AI systems.“ The renowned and traditional conference took place from March 31 to April 2, 2025 in San Francisco. The proceedings are available at ojs.aaai.org/index.php/AAAI-SS/issue/view/654.

Fig.: Revisiting the Trolley Problem

Veröffentlicht am

AI for Social Good V

Am 28. März 2017 begann um 9.00 Uhr der zweite Tag des AAAI Spring Symposium „AI for Social Good“. Den Session Chair für die Talk Session 3 („Urban Planning“) hatte Virginia Dignum (Delft University of Technology) inne. Sie führte in das Thema ein und postulierte: „Artefacts have politics“. Sie erwähnte Anliegen der Maschinenethik und unterschiedliche Modelle der normativen Ethik, die man auf autonome Systeme anwenden kann. „Smart-City Planning with Constrained Crowd Judgment Analysis“ war der Titel des ersten Vortrags von Sujoy Chatterjee (Paper mit Anirban Mukhopadhyay und Malay Bhattacharyya). Syed Ali Asad Rizvi (Paper mit Elmarie van Heerden, Arnold Salas u.a.) trug zu „Identifying Sources of Discrimination Risk in the Life Cycle of Machine Intelligence Applications under New European Union Regulations“ vor. Er erwähnte einen „discrimination bias“ und fragte danach, wie man diesem begegnen kann. Daniela Rosu präsentierte das Paper „Knowledge-based Provisioning of Goods and Services: Towards a Virtual Social Needs Marketplace“, das zusammen mit Dionne M. Aleman, Christopher Beck u.a. erstellt wurde. Ausgangspunkt war, dass immer mehr Menschen Produkte und Dienstleistungen benötigen, die sie sich nicht leisten können. Es geht nun darum, diese in effizienter Weise von der Anbieter- zur Nachfragerseite zu bringen. Amin Ghafouri (Paper mit Aron Laszka, Abhishek Dubey und Xenofon Koutsoukos) referierte zu „Optimal Detection of Faulty Traffic Sensors Used in Route Planning“, Karen Judd Smith – den Fokus wieder stark erweiternd – zu „The AI Community and the United Nations: A Missing Global Conversation“.