Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 2
Encontrada(s) 1 página(s)

DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Otros (Other)

On autonomy, governance and values; an AGV approach to value engineering

  • Noriega, Pablo
  • Plaza, Enric
First International Workshop, VALE 2023, Krakow, Poland, September 30, 2023, Proceedings., This book constitutes the post-conference proceedings of the First International Workshop on Value Engineering in Artificial Intelligence, VALE 2023, held in Krakow, Poland, on September 30, 2023., In this paper we show how to approach the engineering of values in social human-AI systems by looking into the interplay of three notions: autonomy, governance and value (AGV). We propose a particular characterisation of the Value Alignment Problem based on this approach. We use an example to illustrate how the values can be engineered in order to solve the VAP in this example. Based on these elements we advance some arguments in favour of framing a theory of values that may apply to the governance of social systems that involve artificial autonomous entities as well as humans., Authors wish to acknowledge the contributions and fruitful discussions with Mark d’Inverno, Julian Padget and Harko Verhagen. Research for his paper is supported by EU (Horizon-EIC-2021-Pathfinderchallenges-01) Project VALAWAI 101070930; the EU (NextGenerationEU/PRTR program) and the Spanish (MCIN/AEI-10.13039-501100011033 program) project VAE TED2021-131295B-C31; and CSIC’s (Bilateral Collaboration Initiative i-LINK-TEC) project DESAFIA2030 BILTC22005., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación Otros (Other)

An Approximate Embedding for Designing Ethical Reinforcement Learning Environments

  • Mayoral-Macau, Arnau
  • Rodríguez-Soto, Manel
  • Marchesini, Enrico
  • Sánchez-Fibla, Martí
  • López-Sánchez, Maite
  • Rodríguez-Aguilar, Juan Antonio
  • Farinelli, Alessandro
28th European Conference on Artificial Intelligence, 25-30 October 2025, Bologna, Italy – Including 14th Conference on Prestigious Applications of Intelligent Systems (PAIS 2025). Organized by the European Association for Artificial Intelligence (EurAI) and the Italian Association for Artificial Intelligence (AIxIA). Hosted by the University of Bologna., This publication presents the proceedings of ECAI-2025, the 28th European Conference on Artificial Intelligence, held from 25 to 30 October 2025 in Bologna, Italy. Originally organized as a biennial event, ECAI is now held annually, and continuing a long-standing tradition, this edition of ECAI was held alongside the 14th Conference on Prestigious Applications of Intelligent Systems (PAIS). Since its inception in 1974, ECAI has become the premier venue for presenting AI research in Europe, where researchers and practitioners gather to discuss the latest trends and challenges and showcase innovative applications and advances in AI technology. A record total of 3,266 abstract submissions were received for this year’s main track, 2,667 of which were followed by full papers (acceptance rate 23%), together with 81 submissions to the demo track (acceptance rate 33%). Topics cover the full breadth of AI, with many papers exploring interdisciplinary themes. PAIS, an important forum for the sharing of insights into the applicability, development, and deployment of intelligent systems, received 82 submissions (acceptance rate 45%). As well as the papers presented at ECAI and PAIS, this volume also includes eight invited papers on Frontiers in AI corresponding to short invited talks highlighting important new results, techniques, and trends, and the abstracts of 15 recently published papers not previously included in an archival proceedings. This volume of proceedings will be of interest to all those working in the field of Artificial Intelligence., Conference paper., This paper introduces the Approximate Ethical Embedding Process, an algorithm for automating the design of ethical environments for learning agents. Our algorithm helps build environments wherein multiple agents learn policies that align with an ethical (moral) value while simultaneously pursuing their individual objectives. Therefore, we contribute to endowing environment designers with algorithmic tools for building ethical environments. We demonstrate the ethical design process for two different settings of an environment where agents have to adhere to beneficence to promote the collective survival of the population. Our experiments show that our approximate embedding process successfully generates environments that incentivise the learning of value-aligned policies., The research presented in this paper was supported by the EU-funded VALAWAI (# 101070930) project, and the Spanish-funded VAE (# TED2021-131295B-C31), EMOROBCARE (# IASOMMA2024), and ACISUD (PID2022-136787NB-I00) projects., Peer reviewed