Quantifying Responsibility with Probabilistic Causation -- The Case of Climate Action

TL;DR

This paper introduces probabilistic metrics for quantifying moral responsibility in complex, uncertain climate decision scenarios, moving beyond binary causation models to more nuanced responsibility assessments.

Contribution

It develops a set of quantitative responsibility metrics using game trees, applicable to real-world climate action scenarios with uncertainty.

Findings

Metrics effectively quantify responsibility degrees in probabilistic terms.

Proposed methods outperform existing causation-based approaches.

Application to real-world scenarios yields meaningful responsibility scores.

Abstract

Many real-world situations of ethical and economic relevance, such as collective (in)action with respect to the climate crisis, involve not only diverse agents whose decisions interact in complicated ways, but also various forms of uncertainty, including both quantifiable risk and unquantifiable ambiguity. In such cases, an assessment of moral responsibility for ethically undesired outcomes or of the responsibility to avoid these is challenging and prone to the risk of under- or over determination. In contrast to existing approaches that employ notions of causation based on combinations of necessity and sufficiency or certain logics that focus on a binary classification of `responsible' vs `not responsible', we present a set of quantitative metrics that assess responsibility degrees in units of probability. To this end, we adapt extensive-form game trees as the framework for representing decision scenarios and evaluate the proposed responsibility functions based on the correct representation of a set of analytically assessed paradigmatic example scenarios. We test the best performing metrics on a reduced representation of a real-world decision scenario and are able to compute meaningful responsibility scores.