Ethics in Sociotechnical Systems

Tutorial at AAAI 2022 (Vancouver, BC)

February 23, 2022 (10:00 to 14:00 GMT -8)

Tutorial AH4: AAAI Virtual Venue -- Room Red 1

Overview

There is an increasing interest in Ethics and AI and rightly so. Normative and sociotechnical systems have been key topics of interests in the AI literature, specifically, multiagent systems. This tutorial demonstrates that multiagent systems research has much to offer in making AI systems ethical, not just at a single-agent level but at a societal level.

The surprising capabilities of AI overlaid on detailed data and fine-grained control give cause for concern that agents can wield enormous power over human welfare, drawing increasing attention to ethics in AI. This tutorial introduces ethics as a sociotechnical construct, demonstrating how ethics can be modeled and analyzed, and requirements on ethics (values) can be elicited, in a sociotechnical system.

Target Audience

This tutorial is presented at a senior undergraduate student level. It is accessible to developers from industry and to students. Typical attendees for our past tutorials have been researchers and practitioners from industry and government, developers, graduate and senior undergraduate students, and university faculty.

Relevance

Realizing the full potential of AI systems for supporting ethics requires educating a new generation of students and researchers in the relevant concepts. We bring together the latest research including theoretical underpinnings and practical approaches valuable to both researchers and practitioners. Our tutorial intends to serve the following objectives.

  • Motivate and explain emerging topics relevant to engineering ethical AI and MAS

  • Introduce novices to major topics within AI and MAS

  • Introduce expert non-specialists to an AI and MAS subarea

Prerequisites

The tutorial is designed for a broad audience, with an undergraduate-level understanding of AI. The topics are self-contained---foundations are introduced before diving into research challenges and emerging topics. A familiarity with multiagent systems is useful but not assumed. The participants familiar with MAS will benefit by learning research challenges and emerging topics. Those not familiar with MAS will gain a new perspective on engineering AI systems and why the MAS perspective facilitates ethics.

Interaction Style and Delivery

In order to engage the attendees, we will use a variety of instruction materials, including slides, specifications, demos, examples, and problem sets. Assuming that the conference will be face-to-face and few travel restrictions, the tutorial is planned for a hybrid mode allowing both face-to-face and virtual interactions.

Content

Ethics is inherently a multiagent concern---an amalgam of (1) one party's concern for another and (2) a notion of justice. To capture the multiagent conception, this tutorial introduces ethics as a sociotechnical construct. Specifically, we demonstrate how ethics can be modeled and analyzed, and requirements on ethics (value preferences) can be elicited, in a sociotechnical system (STS). An STS comprises of autonomous social entities (principals, i.e., people and organizations), and technical entities (agents, who help principals), and resources (e.g., data, services, sensors, and actuators).

This tutorial includes three key elements.

  1. Specifying a decentralized STS, representing ethical postures of individual agents as well as the systemic (STS level) ethical posture.

  2. Reasoning about ethics, including how individual agents can select actions that align with the ethical postures of all concerned principals.

  3. Eliciting value preferences (which capture ethical requirements) of stakeholders using a value-based negotiation technique.

We build upon our earlier tutorials (e.g., at AAMAS 2020, ACSOS 2020, IJCAI 2020, and AAMAS 2021) on engineering ethics in sociotechnical systems and (e.g., at AAMAS 2015 and IJCAI 2016) on engineering a decentralized multiagent system. However, we extend the previous tutorials substantially, including ideas on ethics and values applied to AI. Attendees will learn the theoretical foundations as well as how to apply those foundations to systematically engineer an ethical STS.

Outline

A brief outline of the tutorial annotated with the number of minutes for each section is below.

  • [10] Motivation: Ethical STS as a decentralized multiagent systems

Foundations

  • [10] Philosophy: Background on ethics (virtue, utilitarianism, Rawls ...)

  • [15] Law, Political Science: Sociotechnical systems

  • [15] Psychology: Preferences and values (Rokeach, Schwartz)

  • [20] Q &A

  • [15] Break

Techniques

  • [20] Software Engineering: Value sensitive design

  • [15] Artificial Intelligence & N & Specifying an ethical STS

  • [15] Operations Research & N & Reasoning about ethics (balance self and others)

  • [20] Q&A

  • [15] Break

Research Directions

  • [5] Formal methods: Verification and simulation

  • [5] Psychology: Emotions and equity

  • [5] Machine Learning: Elicitation (surveys; active value learning; inverse RL)

  • [5] Artificial Intelligence: Uniting individual and societal perspectives

  • [5] Law: Law and consent

Synthesis

  • [20] Summary and concluding remarks

  • [20] Q&A

Presenters

Nirav Ajmeri is a Lecturer in Artificial Intelligence at the University of Bristol in the United Kingdom. His research interests are in intelligent agents and multiagent systems with an emphasis on ethics, privacy, and security. Nirav's research seeks to facilitate engineering a society of socially intelligent agents that act ethically and yield a satisfactory experience to their users. His research has appeared in top AI and computing venues including AAMAS, AAAI, IJCAI, IEEE Computer, IEEE Intelligent Systems, IEEE Internet Computing, and ACM TOSEM. Contact Nirav at nirav.ajmeri@bristol.ac.uk.

Pradeep Murukanniah is an Assistant Professor in the Interactive Intelligence group at the Delft University of Technology in the Netherlands. Prior to TU Delft, Pradeep was an Assistant Professor in the Software Engineering department at Rochester Institute of Technology. Engineering socially intelligent agents is the overarching theme of Pradeep's research. His works facilitate sociotechnical systems that support rich interactions between humans and computational agents, enabling a variety of individual and societal applications. Pradeep's research has appeared in prestigious AI conferences (including AAMAS, IJCAI, ACL, and RecSys) and computing journals (including ACM TOSEM, and ACM TOCHI). Pradeep delivered an invited talk on Privacy Analytics at the 2017 International Conference on Privacy, Security, and Trust. Contact Pradeep at p.k.murukannaiah@tudelft.nl.

Munindar P. Singh is a Professor in Computer Science and a co-director of the Science of Security Lablet at North Carolina State University. His research interests include multiagent systems and software engineering with a special emphasis on the engineering of systems consisting of autonomous parties. He coauthored the text Service-Oriented Computing in 2005. Munindar is an IEEE Fellow, a AAAI Fellow, a AAAS Fellow, an ACM Fellow, and a former Editor-in-Chief of IEEE Internet Computing and ACM Transactions on Internet Technology. His current editorial service includes IEEE Internet Computing, IEEE Transactions on Services Computing, Autonomous Agents and Multiagent Systems, and the ACM Transactions on Intelligent Systems and Technology. His prior editorial service includes Journal of Artificial Intelligence Research and Journal of Web Semantics. Munindar was general cochair of AAMAS 2005 and program cochair of CoopIS 1997, ICWS 2008, and several other events. He was a member of the founding IFAAMAS Board of Directors. Contact Munindar at singh@ncsu.edu.

Material