Transitivity in Human and AI-generated Arabic Translation of The Power of Positive Thinking

Abstract

Since the advent of Artificial Intelligence tools, translation researchers endeavored to examine how human translations are in comparison to those generated by AI tools, using corpora from various genres. Many genres have been investigated, yet self-help remains underexplored. Within Translation Studies, research of this type comes in as an attempt to delineate limits and potentials of AI-generated translations, compared to human translations. This study comes in the same vein to investigate whether AI translations preserve of flatten transitivity patterns, and how AI-generated translation realizes experiential meaning relative to human translation. The Power of Positive Thinking by Norman V. Peale was selected for its prominence in the self-help genre and the sub-genre of positivity. In this study, transitivity analysis framework as per Halliday and Matthiessen (2014) was implemented on 119 units of a purposive sample. The analysis was done across the source text, Arabic human translation and another Arabic translation generated by Gemini 2.5 Flash. ST verses AITT resulting in an observed agreement of 98.3%. the 98.3% versus 90.4% result, the contrasting structural-shift profiles. The findings reveal that output translations of AI tend to preserve transitivity pattern more than human translations. However, more agency-reducing structural shifts are produced, such as depersonalization and nominalization, by AI than by the human. The study concluded that categorical preservation and structural realization are separable dimensions of translation accuracy.