New dataset explores moral valence in AI ethics research

Researchers have proposed a new dataset for annotating moral valence in natural language, aiming to better align AI with human ethics by incorporating affective considerations. The dataset, comprising 500 annotations across action/judgement and consequence valence, was derived from text-presented scenarios within the Commonsense Norm Bank. Preliminary results show significant relationships between these valence features and moral classification, with a noteworthy Matthew's correlation coefficient of 0.764 achieved for binary classification using regularised logistic regression. This work suggests that incorporating valenced consequences could lead to more human-morally aligned AI. AI

IMPACT This research could lead to AI systems that better understand and align with human moral reasoning by incorporating affective elements.

RANK_REASON The item is an academic paper proposing a new dataset and methodology for AI ethics research. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• CatalyzeX Code Finder for Papers
• Commonsense Norm Bank
• Consequentialist ethics
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• Kantian Deontological ethics
• Matthew's correlation coefficient
• Moral Foundations Theory
• regularised logistic regression
• ScienceCast
• virtue ethics

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →