Approaches Towards Revising Instrumentation-Based Reliability Coefficients

Abstract

This study addresses negative Cronbach's alpha reliability coefficients observed in two constructs of a newly developed questionnaire assessing agricultural educators' adoption of artificial intelligence. A field test with ten participants revealed negative reliability in the "perceived benefits and drawbacks" (α = -.01) and "access and infrastructure" (α = -.56) constructs, indicating potential multidimensionality and reverse-coded items. Through detailed inter-item correlation matrix analysis, specific statement inconsistencies were identified. The study emphasizes the importance of transparency in reporting such irregularities to improve instrumentation. It highlights the need for careful consideration of item phrasing, reverse coding, and construct unidimensionality in early-stage instrument development, advocating for open discussion of methodological challenges and subsequent corrective actions. This research contributes to establishing guidelines for reporting and addressing instrumentation errors in social science research.