Data di Pubblicazione:
2020
Abstract:
Explainability is becoming a key requirement of AI applications. Theavailability of meaningful explanations of decisions is seen as crucial to ensure awide range of system properties such as trustability, transparency, robustness, andinnovation. Our claim is that thisneed for explanationis part of a broader problemrelated to the fact that most of the current architectures lack properly devisedchannels for collecting and for propagating feedback about decisions and actions:that is, they do not envisage nor supportaccountability. The aim of this paper is toclarify the differences between the concepts of explainability and accountability,which are often (and wrongly) used interchangeably. We draw a line of thoughtseeing in accountability a key factor for innovation in AI applications, and wesuggest a paradigm shift from aneed for explanationto aneed for accountability.
Tipologia CRIS:
04A-Conference paper in volume
Elenco autori:
Matteo Baldoni, Cristina Baroglio, Roberto Micalizio, Stefano Tedeschi
Link alla scheda completa:
Link al Full Text:
Titolo del libro:
Proceedings of the Italian Workshop on Explainable Artificial Intelligence
Pubblicato in: