PuSH - Publikationsserver des Helmholtz Zentrums München

Zhou, L.* ; Casares, P.A.M.* ; Yang, X.* ; Burden, J.* ; Burnell, R.* ; Cheke, L.G.* ; Ferri, C.s.* ; Marcoci, A.* ; Mehrbakhsh, B.* ; Moros-Daval, Y.* ; Sean Ó hÉigeartaigh,* ; Rutar, D.* ; Schellaert, W.* ; Voudouris, K. ; Hernández-Orallo, J.*

Predictable artificial intelligence.

Artif. Intell. 353:104491 (2026)
Verlagsversion DOI
Open Access Hybrid
Creative Commons Lizenzvertrag
Many areas of artificial intelligence, and machine learning in particular, aim at being probably correct, i.e., valid on average, rather than pursuing the idealistic goal of being provably valid for all inputs. However, AI systems could still be predictably valid, such as an imperfect robot deliverer for which we can reliably and precisely predict the task instances for which it is correct and safe, its valid operating range. "Predictable AI" is a nascent research area that explores ways of anticipating key validity indicators (e.g., performance, safety) of present and future AI ecosystems. We argue that achieving predictability is crucial for fostering trust, liability, control, alignment and safety of AI, and thus should be prioritised over performance. We formally characterise predictability, explore its most relevant components, illustrate what can be predicted, describe alternative candidates for predictors, as well as the trade-offs between maximising validity and predictability. To illustrate these concepts, we bring an array of illustrative examples covering diverse ecosystem configurations. "Predictable AI" is related to other areas of technical and non-technical AI research, but have distinctive questions, hypotheses, techniques and challenges. This paper aims to elucidate them, calls for identifying paths towards a landscape of predictably valid AI systems and outlines the potential impact of this emergent field.
Altmetric
Weitere Metriken?
[➜Einloggen]
Zusatzinfos bearbeiten [➜Einloggen]
Publikationstyp Artikel: Journalartikel
Dokumenttyp Review
Schlagwörter Ai Evaluation ; Aleatoric Uncertainty ; Epistemic Uncertainty ; Operating Condition ; Predictability ; Predictable Ai; Ai; Future
ISSN (print) / ISBN 0004-3702
e-ISSN 1872-7921
Zeitschrift Artificial Intelligence
Quellenangaben Band: 353, Heft: , Seiten: , Artikelnummer: 104491 Supplement: ,
Verlag Elsevier
Verlagsort Radarweg 29, 1043 Nx Amsterdam, Netherlands
Begutachtungsstatus Peer reviewed
Institut(e) Institute of AI for Health (AIH)
Förderungen Leverhulme Centre for the Future of Intelligence at the University of Cambridge
The "Spanish Europe Excelencia" 2024
EU-NextGenerationEU through the Spanish government's Plan de Recuperacion, Transformacion y Resiliencia
Catedra ENIA-UPV in Sustainable AI Development
ERDF A way of making Europe
MCIN/AEI
Generalitat Valenciana