Actual Causality and Responsibility Attribution in Decentralized Partially Observable Markov Decision Processes

  title={Actual Causality and Responsibility Attribution in Decentralized Partially Observable Markov Decision Processes},
  author={Stelios Triantafyllou and Adish Kumar Singla and Goran Radanovic},
  journal={Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society},
Actual causality and a closely related concept of responsibility attribution are central to accountable decision making. Actual causality focuses on specific outcomes and aims to identify decisions (actions) that were critical in realizing an outcome of interest. Responsibility attribution is complementary and aims to identify the extent to which decision makers (agents) are responsible for this outcome. In this paper, we study these concepts under a widely used framework for multi-agent… 

Figures and Tables from this paper



Blameworthiness in Multi-Agent Settings

Get a good notion of blameworthiness in a group setting will be critical for designing autonomous agents that behave in a moral manner and this work shows how this can be done using the Shapley value.

Counterfactual Off-Policy Evaluation with Gumbel-Max Structural Causal Models

An off-policy evaluation procedure for highlighting episodes where applying a reinforcement learned policy is likely to have produced a substantially different outcome than the observed policy, and a class of structural causal models for generating counterfactual trajectories in finite partially observable Markov Decision Processes (POMDPs).

Woulda, Coulda, Shoulda: Counterfactually-Guided Policy Search

The Counterfactually-Guided Policy Search (CF-GPS) algorithm is proposed, which leverages structural causal models for counterfactual evaluation of arbitrary policies on individual off-policy episodes and can improve on vanilla model-based RL algorithms by making use of available logged data to de-bias model predictions.

A Concise Introduction to Decentralized POMDPs

This book introduces multiagent planning under uncertainty as formalized by decentralized partially observable Markov decision processes (Dec-POMDPs). The intended audience is researchers and

A Modification of the Halpern-Pearl Definition of Causality

The original Halpern-Pearl definition of causality was updated in the journal version of the paper and was modified yet again to deal with some problems pointed out by Hopkins and Pearl [2003].

Graded Causation and Defaults

A flexible formal framework for incorporating defaults, typicality, and normality into an account of actual causation is developed and the resulting account takes actual causation to be both graded and comparative.

Responsibility and Blame: A Structural-Model Approach

The definition of causality is extended to take into account the degree of responsibility of A for B, and a notion of degree of blame is defined, which takes into account an agent's epistemic state.

A Game-Theoretic Account of Responsibility Allocation

This work model strategic multi-agent interaction as an extensive form game of imperfect information and defines notions of forward and backward responsibility, where the former captures the epistemic knowledge of players along a play, while the latter formalizes which players – possibly unknowingly – caused the outcome.

Towards Formal Definitions of Blameworthiness, Intention, and Moral Responsibility

These formal definitions of degree of blameworthiness and intention relative to an epistemic state and actual causality provide the key ingredients for moral responsibility judgments and give insight into commonsense intuitions in a variety of puzzling cases from the literature.

On the Definition of Actual Cause (draft Copy { Comments Are Welcome)

1 Background This note proposes a formal explication of the notion of \actual cause" as in, for example, \Socrates drinking hemlock was the actual cause of Socrates death." The philosophical