Accepted Tutorials

Explainability, Trust and Ethics for Robots and Autonomous Systems

Presenter: Jim Torresen, University of Oslo

Abstract: Explainability, trust and ethics for robots and autonomous systems are getting increased attention. This is partly as a result of the currently limited deployment of such systems in human support and interaction settings. The tutorial will give an overview of the most commonly expressed ethical challenges and ways being undertaken to reduce their impact using the findings in an earlier undertaken review (https://www.frontiersin.org/articles/10.3389/frobt.2017.00075/full) supplemented with recent work and initiatives. The presentation will exemplify the challenges related to privacy, security and safety through several examples from own and others’ work.

Duration: Half day

Recent Advances in Fair Resource Allocation

Presenters: Rupert Freeman (MSR New York) and Nisarg Shah (University of Toronto)

Abstract: This tutorial will survey results on fairness in resource allocation. Starting from the classical setting of cake-cutting, the tutorial will explore the allocation of divisible and indivisible goods, public decision-making, and online resource allocation. The tutorial will include various formal definitions of fairness, relations among these definitions, and algorithms for (and complexity of) satisfying such definitions.

The focus will be on recent advances in this line of work; however, no prior background of the fair division field will be necessary. The tutorial is suitable for everyone from undergraduate students interested in learning about this research area to researchers actively working on it. The attendees can expect to walk away with a deeper understanding of formal definitions of fairness and existing algorithms to achieve them. The tutorial will also point out a broad variety of open problems in the field, which researchers interested in the field can work on.

Duration: Half day

Ethics in Sociotechnical Systems

Presenters: Pradeep Murukannaiah (Delft University of Technology), Nirav Ajmeri (NC State University), Munindar P. Singh (NC State University)

Abstract: The surprising capabilities demonstrated by AI technologies 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.

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 2015 and IJCAI 2016) on engineering decentralized MAS, which were well attended. However, we extend the previous tutorials substantially, including ideas on ethics and values. Attendees will learn the theoretical foundations as well as how to apply those foundations to systematically engineer an ethical STS.

Duration: Half day

Automated Verification of Autonomous Agents: Why, What, and Especially: How?

Presenters: Wojciech Jamroga, Wojciech Penczek, and Damian Kurpiewski, Institute of Computer Science, Polish Academy of Sciences

Abstract: The goal of this tutorial is to introduce the latest developments in automated verification of multi agent systems. We will present the ideas and results in a lightweight manner, and aim especially at researchers outside of the formal methods community.

Automated verification of discrete-state systems has been a hot topic in computer science for over 35 years. The idea found its way into AI and multi-agent systems in late 1990’s, and techniques for verification of such systems have been in constant development since then.
Model checking of temporal, epistemic, and strategic properties is one of the most prominent and most successful approaches here. In this tutorial, we present a brief introduction to the topic, and mention relevant properties that one might like to verify this way. Then, we describe some very recent results on incomplete model checking algorithms and model reductions, which can potentially lead to practical solutions for the notoriously hard problem. We conclude by a presentation of the prototype tool STV (StraTegic Verifier) for verification of strategic ability, developed recently by our group.

Duration: Half day

Virtual Human ToolKit

Presenters: Arno Hartholt, Sharon Mozgai, Albert “Skip” Rizzo, University of Southern California Institute for Creative Technologies

Abstract: The Toolkit is “a collection of modules, tools, and libraries designed to aid and support researchers and developers with the creation of virtual human conversational characters.” The Toolkit is developed at the University of Southern California Institute for Creative Technologies, is freely available for the academic research community, and supports speech recognition, natural language processing, nonverbal behavior generation, nonverbal behavior realization, text-to-speech generation, and rendering. This half-day tutorial focuses on three areas: 1) overview of the main technologies, 2) overview of the overall architecture, and 3) hands-on creation of a basic interactive character. The target audience is interested researchers within the field of interactive agents and related areas with an affinity for technology.

Duration: Half day

Computational Game Theory and Its Applications

Presenter: Hau Chan, University of Nebraska-Lincoln

Abstract: In this tutorial, the audiences will be :

1. introduced to fundamental game-theoretic decision-making tools for modeling and understanding the strategic interaction of self-interested and strategic agents
2. exposed to the modeling tools’ solution concepts and how they can be used to predict decision-making behavior of agents
3. introduced to computational aspects of computing these solution concepts
4. exposed to some of the main applications of game theory to security and social science domains

In addition, this tutorial will summarize recent and modern advances in developing efficient algorithms for games with complex strategy spaces, including the use of marginal probabilities and general framework for representing and solving games with structured strategy spaces. We will cover application domains ranging from infrastructure security to environmental and wildlife protection.

Duration: Half day

Agent Technologies in Power and Energy Systems

Presenters: Archie Chapman (University of Queensland), Valentin Robu (Heriot-Watt University), Sarvapali Ramchurn (University of Southampton)

Abstract: This tutorial will cover emerging applications of agent technologies in power and energy systems. Key agent technologies covered will include: automated home energy management, group buying and collective tariff switching, virtual power plants and aggregated demand response, network-aware distributed energy resource coordination, electric vehicle integration, peer-to-peer energy trading and strategic asset investment games.
Starting from these energy application areas, we will discuss the main agent methodologies employed in this domain, including: sequential decision making under uncertainty and reinforcement learning frameworks; large-scale distributed optimisation, non-cooperative games (e.g. congestion and Stackelberg games); approaches to cost-sharing and cooperative games; and market and mechanism design problems.
These methods have shown great potential to solve emerging technical challenges in power systems (such a intermittent renewable generation; electrification of heating/cooling and transport) by harnessing the flexibility of the demand side. This will facilitate a transition to a low-carbon energy future in which loads to follow the intermittent supply of power from renewable generators, and will help make future power systems truly “smart” power grids. Our intention is to review and illustrate, for a broad audience of multi-agent researchers and practitioners (not necessarily having prior experience in this application area), how theory and algorithms for agent-based systems can be applied and translated into useful tools and products, in the energy domain.

Duration: Half day

Network Economics, Variational Inequalities, and Deep Learning

Presenter: Sridhar Mahadevan, Adobe Research

Abstract: Increasingly, much work in AI – from machine learning and natural language processing to planning, perception, and robotics – is based on classical (continuous) optimization. While this foundation has proved to be of considerable practical utility, the increasingly decentralized and networked nature of computation in the 21st century, implies that classical optimization may prove increasingly restrictive, as AI tackles applications that involve massively large networked environments, where data is stored heterogeneously in the “cloud”, and computation involves balancing multiple competing objectives, including cost, privacy, reliability, and security. The aim of this tutorial is to present an elegant mathematical formalism for solving large games that has been extensively studied in network economics, but so far, has not yet played a major role in multiagent AI. The framework is based on variational inequalities, a formalism that extends classical optimization to vector fields. We use a variety of real-world problems, from modeling traffic flow to content distribution on the Internet and green supply chains in sustainable manufacturing, to illustrate the power of this formalism. We also show that work in deep learning on generative adversarial networks results in complex network dynamics, and can be profitably studied this framework. The tutorial will introduce all the necessary mathematics, and should be of interest to AAMAS researchers from a wide variety of backgrounds.

Duration: Half day

Behavior Informatics: Methods and Applications

Presenters: Longbing Cao (University of Technology, Sydney), Can Wang (Griffith University)

Abstract: Complex behaviors are widely seen in artificial and natural intelligent systems, on the internet, social and online networks, multi‐agent systems, and brain systems. The in‐depth understanding of complex behaviors has been increasingly recognized as a crucial means for disclosing interior driving forces, causes and impact on businesses in handling many challenging issues. However, traditional behavior modeling mainly relies on qualitative methods from behavioral science and social science perspectives. The so‐called behavior analysis in data analytics and learning often focuses on human demographic and business usage data, in which behavior‐oriented elements are hidden in routinely collected transactional data. As a result, it is ineffective or even impossible to deeply scrutinize native behavior intention, lifecycle, dynamics and impact on complex problems and business issues. In this tutorial, we will present an overview of behavior analytics, review and discuss state‐of‐the‐art and newly emerged techniques for complex behavior analytics, which cover high impact behavior sequence analysis, impact‐oriented combined behavior analysis, high utility behavior analysis, non‐occurring behavior analysis, coupled/group/collective behavior analysis, statistical modeling of coupled behaviors, probabilistic modeling of sparse rating behaviors, understanding behavior choice and attraction, behavior analysis with recurrent networks, behavior analysis in visual data, behavior learning from demonstrations. We will show that in‐depth behavior analytics creates new opportunities, directions and means for, learning and analysis of complex behaviors in both physical and virtual organizations.

Duration: Half day

How to create and execute online human behavior experiments for informing agent-based models

Presenters: Kiran Lakkaraju, Sandia National Labs

Abstract: The ubiquity of the internet and the popularization of online crowdsourcing tools such as Amazon Mechanical Turk and prolific.ac have made capturing data from human subjects easy, fast and cheap; opening the door for data driven calibration and validation of agent-based models. However, there are many parts of designing and obtaining approval of a human subject study that can be foreign to the agent based modeling community.
In this tutorial we will go over the benefits to gathering data from an online experiment, how to design an online experiment, the intricacies of writing an institutional review board (IRB) protocol, and how to deploy and collect data from your experiment.

Duration: Half day

Reinforcement Learning and Imitation Learning

Presenter: Neda Navidi, AI Redefined

Abstract: The purpose of this tutorial is to provide an introduction to Deep RL and IL at a level easily understood by students and researchers in a wide range of disciplines. Also, we will overview different RL and IL techniques and methods as long as discuss how they are utilized to improve the performance of different tasks. Also, we cover the usage of RL and IL in CV, NLP and time-series tasks. First, the introduction of value-based and policy-based learning methods in RL will be presented. We will discuss the challenging concepts of RL: independence/ collaborative multi-agent learning, distribution learning, high dimensionality and continuousness of real environment, long-horizon problems. Second, we will cover different tasks solved by RL: autonomous driving, recommendation system, robotics, and games. Third, we will discuss different IL concepts: curriculum learning, interactive teaching, active/passive learning, meta-learning, transfer learning and Arcade Learning Environment. Finally, we will present the different tasks which are solved by IL or integration of RL and IL.

Duration: Half day