Author name cluster

Edmond Awad

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

6 papers

1 author row

AAMAS Conference 2025 Conference Paper

When Is It Acceptable to Break the Rules? Knowledge Representation of Moral Judgements Based on Empirical Data (Extended Abstract)

Edmond Awad
Sydney Levine
Andrea Loreggia
Nicholas Mattei
Iyad Rahwan
Francesca Rossi
Kartik Talamadupula
Joshua Tenenbaum

This paper explores how humans make contextual moral judgments to inform the development of AI systems capable of balancing rulefollowing with flexibility. We investigate the limitations of rigid constraints in AI, which can hinder morally acceptable actions in specific contexts, unlike humans who can override rules when appropriate. We propose a preference-based graphical model inspired by dual-process theories of moral judgment and conduct a study on human decisions about breaking the social norm of "no cutting in line. " Our model outperforms standard machine learning methods in predicting human judgments and offers a generalizable framework for modeling moral decision-making across various contexts. This short paper summarizes the main findings of our paper published in the journal Autonomous Agents and Multi-Agent Systems. [2]

PDF

JAAMAS Journal 2024 Journal Article

When is it acceptable to break the rules? Knowledge representation of moral judgements based on empirical data

Edmond Awad
Sydney Levine
Max Kleiman-Weiner

Abstract Constraining the actions of AI systems is one promising way to ensure that these systems behave in a way that is morally acceptable to humans. But constraints alone come with drawbacks as in many AI systems, they are not flexible. If these constraints are too rigid, they can preclude actions that are actually acceptable in certain, contextual situations. Humans, on the other hand, can often decide when a simple and seemingly inflexible rule should actually be overridden based on the context. In this paper, we empirically investigate the way humans make these contextual moral judgements, with the goal of building AI systems that understand when to follow and when to override constraints. We propose a novel and general preference-based graphical model that captures a modification of standard dual process theories of moral judgment. We then detail the design, implementation, and results of a study of human participants who judge whether it is acceptable to break a well-established rule: no cutting in line. We then develop an instance of our model and compare its performance to that of standard machine learning approaches on the task of predicting the behavior of human participants in the study, showing that our preference-based approach more accurately captures the judgments of human decision-makers. It also provides a flexible method to model the relationship between variables for moral decision-making tasks that can be generalized to other settings.

Details DOI

AAMAS Conference 2023 Conference Paper

Should My Agent Lie for Me? A Study on Attitudes of US-basedParticipants Towards Deceptive AI in Selected Future-of-work

StefanXXX Sarkadi
Peidong Mei
Edmond Awad

Artificial Intelligence (AI) advancements might deliver autonomous agents capable of human-like deception. Such capabilities have mostly been negatively perceived in HCI design, as they can have serious ethical implications. However, AI deception might be beneficial in some situations. Previous research has shown that machines designed with some level of dishonesty can elicit increased cooperation with humans. This raises several questions: Are there future-of-work situations where deception by machines can be an acceptable behaviour? Is this different from human deceptive behaviour? How does AI deception influence human trust and the adoption of deceptive machines? In this paper, we describe a user study to answer these questions by considering different contexts and job roles. We report differences and similarities with the perception of humans behaving deceptively in the same roles. Our findings provide insights and lessons that will be crucial in understanding what factors shape the social attitudes and adoption of AI systems that may be required to exhibit dishonest behaviour as part of their jobs.

PDF

AIJ Journal 2020 Journal Article

An approach for combining ethical principles with public opinion to guide public policy

Edmond Awad
Michael Anderson
Susan Leigh Anderson
Beishui Liao

We propose a framework for incorporating public opinion into policy making in situations where values are in conflict. This framework advocates creating vignettes representing value choices, eliciting the public's opinion on these choices, and using machine learning to extract principles that can serve as succinct statements of the policies implied by these choices and rules to guide the behavior of autonomous systems.

Details DOI

AAAI Conference 2018 Conference Paper

A Voting-Based System for Ethical Decision Making

Ritesh Noothigattu
Snehalkumar Gaikwad
Edmond Awad
Sohan Dsouza
Iyad Rahwan
Pradeep Ravikumar
Ariel Procaccia

We present a general approach to automating ethical decisions, drawing on machine learning and computational social choice. In a nutshell, we propose to learn a model of societal preferences, and, when faced with a speciﬁc ethical dilemma at runtime, efﬁciently aggregate those preferences to identify a desirable choice. We provide a concrete algorithm that instantiates our approach; some of its crucial steps are informed by a new theory of swap-dominance efﬁcient voting rules. Finally, we implement and evaluate a system for ethical decision making in the autonomous vehicle domain, using preference data collected from 1. 3 million people through the Moral Machine website.

PDF Details

KR Conference 2014 Short Paper

Interval Methods for Judgment Aggregation in Argumentation

Richard Booth
Edmond Awad
Iyad Rahwan

In the present paper, we embark on a broader study of JA in argumentation. We define a general family of aggregation operators called interval methods and show that they contain existing operators as instances. Interval methods always satisfy a strong version of Independence, but will usually fail Collective Rationality. But despite this important barrier, we are able to fully axiomatize interval methods in terms of a set of fundamental postulates. Then, building on Caminada and Pigozzi’s down-admissible + up-complete (DAUC) construction, we present an approach to transform any interval method into one satisfying Collective Rationality while preserving a weaker and more reasonable form of independence known as Directionality. Given a set of conflicting arguments, there can exist multiple plausible opinions about which arguments should be accepted, rejected, or deemed undecided. Recent work explored some operators for deciding how multiple such judgments should be aggregated. Here, we generalize this line of study by introducing a family of operators called interval aggregation methods, which contain existing operators as instances. While these methods fail to output a complete labelling in general, we show that it is possible to transform a given aggregation method into one that does always yield collectively rational labellings. This employs the downadmissible and up-complete constructions of Caminada and Pigozzi. For interval methods, collective rationality is attained at the expense of a strong Independence postulate, but we show that an interesting weakening of the Independence postulate is retained. Preliminaries We assume a countably infinite set U of argument names, from which all possible argumentation frameworks are built. Definition 1 An argumentation framework (AF for short) A “ pArgs, áq is a pair consisting of a finite set Args Ď U of arguments and an attack relation áĎ Args ˆ Args. Sometimes we use Args A and áA to denote the arguments and attack relation of a given AF A.