[Paper Review] One Formalization of Virtue Ethics via Learning

Naveen Sundar Govindarajulu, Selmer Bringjsord+1 more|arXiv (Cornell University)|May 20, 2018

Logic, Reasoning, and Knowledge8 references3 citations

TL;DR

This paper presents a formalization of exemplarist virtue ethics using a cognitive calculus that integrates learning, emotions (especially admiration), and utility-based reasoning. It models how agents learn virtuous traits by observing and internalizing exemplars, using a quantified modal logic framework to automate ethical reasoning in machines, marking a foundational step toward machine ethics grounded in virtue theory rather than deontology or consequentialism.

ABSTRACT

Given that there exist many different formal and precise treatments of deontologi- cal and consequentialist ethics, we turn to virtue ethics and consider what could be a formalization of virtue ethics that makes it amenable to automation. We present an embroyonic formalization in a cognitive calculus (which subsumes a quantified first-order logic) that has been previously used to model robust ethical principles, in both the deontological and consequentialist traditions.

Motivation & Objective

• To address the lack of formal, computable treatments of virtue ethics compared to deontological and consequentialist ethics.
• To enable automation of virtue ethics in AI and robotics by formalizing a specific version—exemplarist virtue theory—using a cognitive calculus.
• To model the role of emotions, particularly admiration, as foundational to moral understanding and trait learning.
• To formalize the process of learning virtuous traits from exemplars, moving beyond isolated actions to internalized character dispositions.
• To bridge the gap between informal virtue ethics and machine-computable ethical reasoning.

Proposed method

• Uses the deontic cognitive event calculus (DCEC), a quantified first-order modal logic, to formalize ethical reasoning with temporal and intentional fluents.
• Introduces a utility function μ(f,t) that aggregates total utility across time and agents, derived from agent-specific utility functions ν(a,f,t).
• Models emotions via the OCC (Ortony, Cohen, and Collins) model, formalized in a quantified modal logic to represent emotions like joy, distress, and pity based on event outcomes and agents involved.
• Defines admiration as a key emotional response to exemplars, triggering the learning of virtuous traits through observation and study.
• Introduces a formal mechanism for updating beliefs and traits over time based on continuous evaluation of exemplars and their consequences.
• Uses a counterfactual conditional framework to evaluate what-if scenarios in ethical reasoning, grounded in proof-based semantics.

Experimental results

Research questions

• RQ1 How can exemplarist virtue ethics be formalized in a way that enables automation in artificial agents?
• RQ2 What role do emotions—particularly admiration—play in the acquisition of virtuous traits, and how can they be modeled computationally?
• RQ3 How can learning of complex character traits be formalized rather than just isolated actions?
• RQ4 Can a cognitive calculus that supports deontological and consequentialist reasoning also support virtue ethics through learning?
• RQ5 What formal mechanisms are needed to represent the dynamic, evolving nature of exemplars and their moral status over time?

Key findings

• The paper successfully formalizes exemplarist virtue ethics (Vz) as Vzf within a cognitive calculus, enabling machine reasoning about virtuous behavior.
• Emotions such as admiration are modeled as fluents that trigger the learning of virtuous traits, with formal definitions grounded in the OCC model.
• A utility-based framework is established where total utility is computed over time and agents, allowing for consequentialist evaluation within a virtue ethics context.
• The formalization supports dynamic updating of beliefs and trait learning, reflecting the evolving nature of exemplars and moral understanding.
• The approach demonstrates that virtue ethics can be formalized and automated using a logic that already supports deontological and consequentialist reasoning.
• The work provides a foundation for integrating virtue ethics into moral AI, showing that character-based ethics is computationally tractable when grounded in learning and emotion.