by Slavkovik, Marija
Abstract:
To have ethical AI two questions need to be answered: i) what is the ethical impact that an AI system can have, and, ii) what does it mean for an AI system to behave ethically. The lack of answers to both of these questions hinder the identification of what are the values or principles that we want upheld by AI and for AI. Identifying these principles is not enough, we also want to define them so that they can be operational, or at least understand what operational means here. There is a gap between moral philosophy and ethically behaving AI. The tutorial contributes towards closing this gap, by motivating researchers to interpret an abstract principle from moral philosophy into an algorithmic property that can be formally specified and measured or computationally implemented. The tutorial uses recent articles in AI ethics that attempt to define and identify pertinent ethical principles, as well as ethically motivated desirable algorithmic properties.
Reference:
Mythical Ethical Principles for AI and how to attain them (Slavkovik, Marija), Chapter in (Chetouani, Mohamed, Dignum, Virginia, Lukowicz, Paul, Sierra, Carles, eds.), Springer International Publishing, 2023.
Bibtex Entry:
@Inbook{Slavkovik2023,
author="Slavkovik, Marija",
editor="Chetouani, Mohamed
and Dignum, Virginia
and Lukowicz, Paul
and Sierra, Carles",
title="Mythical Ethical Principles for AI and how to attain them",
bookTitle="Human-Centered Artificial Intelligence: Advanced Lectures",
year="2023",
publisher="Springer International Publishing",
address="Cham",
pages="275--303",
abstract="To have ethical AI two questions need to be answered: i) what is the ethical impact that an AI system can have, and, ii) what does it mean for an AI system to behave ethically. The lack of answers to both of these questions hinder the identification of what are the values or principles that we want upheld by AI and for AI. Identifying these principles is not enough, we also want to define them so that they can be operational, or at least understand what operational means here. There is a gap between moral philosophy and ethically behaving AI. The tutorial contributes towards closing this gap, by motivating researchers to interpret an abstract principle from moral philosophy into an algorithmic property that can be formally specified and measured or computationally implemented. The tutorial uses recent articles in AI ethics that attempt to define and identify pertinent ethical principles, as well as ethically motivated desirable algorithmic properties.",
isbn="978-3-031-24349-3",
doi="10.1007/978-3-031-24349-3_15",
url="https://doi.org/10.1007/978-3-031-24349-3_15"
}