Scales to measure the perceived beneficence and non-maleficence of generative artificial intelligence systems

Purpose The rapid adoption of generative artificial intelligence (GAI) offers significant opportunities but also raises ethical concerns. Across institutional guidelines and theoretical frameworks in the field of AI ethics, two principles – beneficence and non-maleficence – aim to direct the development and use of GAI to ensure maximum value and minimum harm. However, understanding how the general population perceives these principles in the context of GAI remains limited. This study aims to develop and validate scales for two new constructs in the context of GAI: perceived beneficence and perceived non-maleficence of GAI systems. Design/methodology/approach The authors propose items for two scales. Two independent samples were used for scale development. The first group comprised young adults from the general population and was analysed using exploratory factor analysis (EFA). The second group comprised used adults residing in the European Union and was analysed using confirmatory factor analysis (CFA). Findings EFA revealed a two-factor solution, explaining approximately 64 % of the total variance. CFA confirmed the expected two-factor structure, with all standardised loadings exceeding 0.70. Originality/value The proposed scales provide a practical and research-relevant means for further behavioural analysis within structural models to address various contexts of GAI use. The proposed scales can help policymakers better understand GAI user perceptions, thereby enabling more robust policies to govern the development and use of GAI.