Validation of RSDA Model in Moral Decision-Making of Artificial Moral Agent or AI Robots

Abstract

The objective of this research is to validate the RSDA model for forecasting decision-making processes in Artificial Moral Agents (AMAs) or AI Robots. The RSDA model delineates the Rubric evaluation, Scenario development, and Data collection phases, ultimately culminating in the creation of an Algorithm capable of predicting robot or AI responses that closely align with human decision-making. This investigation demonstrates the feasibility of a hybrid AMA model that can emulate human ethical judgment by discerning ethical scores through the analysis of human decision-making patterns in real-life scenarios. Data for this study were gathered from a cohort of elementary and college students who responded to four ethical dilemma scenarios involving domestic, medical, and educational robots, as well as autonomous vehicles. The Univariate Dynamic Encoding Algorithm for Searches (uDEAS) was subsequently employed to construct a statistical model that conforms to the decision-making patterns observed in human groups under consideration. According to the results of the RSDA model, the absolute mean ethics score for ethical principle 1 is 0.49 for elementary school students and 1.53 for university students, indicating that ethical awareness of human rights develops as students grow older. In addition, the average standard deviation of the ethics scores of the five principles is 1.02 for elementary school students and 0.67 for college students, indicating that ethical judgment narrows with age and ethical consensus is formed. The findings of this study affirm that the RSDA model, while intuitive, systematically elucidates each step and holds substantial promise for deployment in scenarios wherein intelligent agents necessitate human-like decision-making capabilities. Moreover, the RSDA model is anticipated to enhance the credibility of AMAs by augmenting the transparency and explainability of decisions made by social robots or AI.