United Kingdom

We study the problem of zero-shot coordination (ZSC), where agents must independently produce strategies for a collaborative game that are compatible with novel partners not seen during training. In particular, our first contribution is to consider the need for diversity in generating such agents. Because self-play agents control their own trajectory distribution during training, their policy only performs well on this exact distribution. As a result, they achieve low scores in ZSC, since playing with another agent is likely to put them in situations they have not encountered during training. To address this issue, we train a common best response (BR) to a population of agents, which we regulate to be as diverse as possible. For that purpose, we introduce Trajectory Diversity (TrajeDi) - a differentiable objective for generating diverse reinforcement learning (RL) policies. We present TrajeDi as a generalization of the Jensen-Shannon divergence (JSD) between policies and motivate it experimentally in a simple matrix game, where it allows to find the unique ZSC-optimal solution.

AAMAS 2021

Trajectory Diversity for Zero-Shot Coordination

AAMAS is the largest and most influential conference in the area of agents and multiagent systems, bringing together researchers and practitioners in all areas of agent technology and providing and internationally renowned high-profile forum for publishing and finding out about the latest developments in the field. AAMAS is the flagship conference of the non-profit International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS).  

![](https://assets.underline.io/uploads/markdown_image/1/image/ce4604bd54468a006164d16e080204fb.jpg)

Please click on the button below to register for this event.

If you are registered and see this page please check the account email you are logged in with. It must be the same as for the registration.

If you still have trouble accessing the contact please contact support at aamas2021@underline.io 

poster

In many community health settings, it is crucial to have a systematic monitoring and intervention process to ensure that the patients adhere to healthcare programs, such as periodic health checks or taking medications. When these interventions are expensive, they can be provided to only a fixed small fraction of the patients at any period of time. Hence, it is important to carefully choose the beneficiaries who should be provided with interventions and when. We model this scenario as a restless multi-armed bandit (RMAB) problem, where each beneficiary is assumed to transition from one state to another depending on the intervention provided to them. In practice, the transition probabilities are unknown a priori, and hence, we propose a mechanism for the problem of balancing the explore-exploit trade-off. Empirically, we find that our proposed mechanism outperforms the baseline intervention scheme maternal healthcare dataset.

Learning Index Policies for Restless Bandits with Application to Maternal Healthcare

We study the problem of learning helpful behavior, specifically, learning to cooperate with differently-skilled and diverse partners in the context of two-player, cooperative Atari games. We show robust performance of these so-called Helper-AIs when paired with different kinds of partners (both human and artificial agents), including partners that they have not previously encountered during training. In particular, while pairing an expert AI with a non-expert AI leads to performance that is worse than when pairing the non-expert AI with a copy of itself, these Helper-AIs provide a substantial boost in joint performance.

Learning Robust Helpful Behaviors in Two-Player Cooperative Atari Environments

Recent impossibility results have shown that strategyproofness is difficult to obtain for multiwinner voting rules, especially in combination with proportionality. In this paper, we attempt to identify cases where strategyproofness can be established by considering manipulation on party-list profiles. We distinguish between three types of manipulation  subset-manipulation, superset-manipulation, and disjoint-set-manipulation. Our focus is the class of irresolute Thiele rules. For all three types of manipulation, we are able to establish that Thiele rules are strategyproof on party-list profiles for several well-known preference extensions. For superset- and disjoint-set-strategyproofness, we can extend this result to all preference extensions. We are also able to show that Thiele rules are fully strategyproof for optimistic agents on these profiles.

Manipulability of Thiele Methods on Party-List Profiles

Egalitarian considerations play a central role in many areas of social choice theory. Applications of egalitarian principles range from ensuring everyone gets an equal share of a cake when deciding how to divide it, to guaranteeing balance with respect to gender or ethnicity in committee elections. Yet, the egalitarian approach has received little attention in judgment aggregation  a powerful framework for aggregating logically interconnected issues. We make the first steps towards filling that gap. We introduce axioms capturing two classical interpretations of egalitarianism in judgment aggregation and situate these within the context of existing axioms in the pertinent framework of belief merging. We then explore the relationship between these axioms and several notions of strategyproofness from social choice theory at large. Finally, a novel egalitarian judgment aggregation rule stems from our analysis; we present complexity results concerning both outcome determination and strategic manipulation for that rule.

Egalitarian Judgment Aggregation

Multi-agent reinforcement learning (RL) often struggles to ensure the safe behaviours of the learning agents, and therefore it is generally not adapted to safety-critical applications. To address this issue, we present a methodology that combines formal verification with (deep) RL algorithms to guarantee the satisfaction of formally-specified safety constraints both in training and testing. The approach we propose expresses the constraints to verify in \textit{Probabilistic Computation Tree Logic} (PCTL) and builds an abstract representation of the system to reduce the complexity of the verification step. This abstract model allows for model checking techniques to identify a set of abstract policies that meet the safety constraints expressed in PCTL. Then, the agents' behaviours are restricted according to these safe abstract policies.We provide formal guarantees that by using this method, the actions of the agents always meet the safety constraints, and provide a procedure to generate an abstract model automatically. We empirically evaluate and show the effectiveness of our method in a multi-agent environment.

An Abstraction-based Method to Check Multi-Agent Deep Reinforcement-Learning Behaviors

Traditional multi-agent reinforcement learning algorithms are not scalable to environments with more than a few agents, since these algorithms are exponential in the number of agents. Recent research has introduced successful methods to scale multi-agent reinforcement learning algorithms to many agent scenarios using mean field theory. Previous work in this field assumes that an agent has access to exact cumulative metrics regarding the mean field behaviour of the system, which it can then use to take its actions. In this paper, we relax this assumption and maintain a distribution to model the uncertainty regarding the mean field of the system. We consider two different settings for this problem. In the first setting, only agents in a fixed neighbourhood are visible, while in the second setting, the visibility of agents is determined at random based on distances. For each of these settings, we introduce a Q-learning based algorithm that can learn effectively. We prove that this Q-learning estimate stays very close to the Nash Q-value (under a common set of assumptions) for the first setting. We also empirically show our algorithms outperform multiple baselines in three different games in the MAgents framework, which supports large environments with many agents learning simultaneously to achieve possibly distinct goals.

Partially Observable Mean Field Reinforcement Learning

We consider the problem of committee selection from a fixed set of candidates where each candidate has multiple quantifiable attributes. Instead of voting for a candidate, a voter is allowed to approve the preferred attributes of a given candidate. Though attribute-based preferences capture several important real-life scenarios, committee selection problem with attribute approval of voters has not properly formalized or studied. We present a detailed study of axioms, rules, algorithms, and computational complexity for the setting. Apart from extending previous axioms and rules, we design two new algorithms that are especially appropriate for our model.

Committee Selection using Attribute Approvals

We model interpretive blindness, a type of epistemic bias that poses a problem for learning from testimony, in which one acquires information from text or conversation but lacks direct access to ground truth. Interpretive blindness arises when a co-dependence between background beliefs and interpretation leads to a dynamic process of bias hardening that impedes or precludes learning. We argue that when bodies of data are argumentatively complete, even constraints from hierarchical Bayesian learning designed to promote good epistemic practices will fail to stop interpretive blindness.

Interpretive Blindness and the Impossibility of Learning from Testimony

We investigate a repeated two-player game setting where the column player is also a designer of the system, and has full control over payoff matrices. In addition, we assume that the row player uses a no-regret algorithm to efficiently learn how to adapt their strategy to the column player's behaviour over time. The goal of the column player is to guide her opponent into picking a mixed strategy which is preferred by the system designer. Therefore, she needs to: (i) design appropriate payoffs for both players; and (ii) strategically interact with the row player during a sequence of plays in order to guide her opponent to converge to the desired mixed strategy. To design appropriate payoffs, we propose a novel zero-sum game construction whose unique minimax solution contains the desired behaviour. We also propose another construction in which only the minimax strategy of the row player is unique. Finally, we propose a new game playing algorithm for the system designer and show that it can guide the row player to its minimax strategy, under the assumption that the row player adopts a stable  no-regret algorithm.

How to Guide a Non-Cooperative Learner to Cooperate: Exploiting No-Regret Algorithms in System Design

We present the first explicit-state method for analysing and ensuring the safety of DRL agents for Atari games. Our method only requires access to the emulator. We give a suite of 42 properties that characterise "safe behaviour" for 31 games. We evaluate the safety of the best available DRL agents which, as our experiments show, violate most of our properties. We propose a countermeasure that implements shielding using bounded explicit-state exploration. Our method improved their overall safety, producing the safest DRL agents for Atari games currently available.

Next from AAMAS 2021

Learning Index Policies for Restless Bandits with Application to Maternal Healthcare

Similar lecture

Trajectory Diversity for Zero-Shot Coordination