SCIENCE@home

Can Machines Communicate Psychotrauma? Affective and Cognitive Shifts in AI-Translated Russia-Ukraine War Narratives

Serhii Zasiekin,Diana Kalishchuk

Published 2025 in PSYCHOLINGUISTICS

ABSTRACT

Aim. This paper aims to examine the extent to which AI-based translation tools can capture the psychological and linguistic features of trauma discourse by identifying discrepancies between original and machine-translated texts. Methods. The dataset comprises 299 narratives written by young Ukrainians sharing their personal experiences of the Russia-Ukraine war. These war testimonies were translated into English using DeepL and Google Translate. Using the Ukrainian and English versions of the LIWC2015 (Pennebaker et al., 2015), the authors analysed emotional and stylistic variation and calculated the Categorical-Dynamic Index (CDI) (Jordan & Pennebaker, 2017) of the selected original and two translated sets of texts. A one-way ANOVA was employed to determine statistically significant differences among the three groups. Results. NMT systems’ target versions intensified the narratives’ affective language. Both NMT tools distorted the original disruptive pattern of trauma discourse, producing more dynamic and personalised translations and transforming the original ‘analytical’ style to a more ‘narrative’ one. Although more coherent at the local level, the target versions were deficient in terms of thematic structure and global discourse coherence due to a lack of linguistic fillers. No statistically significant differences were found between the two NMT tools across all psycholinguistic and content dimensions, suggesting comparable performance in translation quality. The findings indicate that the NMT systems under consideration are still far from keeping the cognitive, emotional, and stylistic complexity of war-related trauma narratives. Therefore, it remains premature to consider NMT a substitute for human translators, particularly when translating emotionally sensitive or trauma-related content.

PUBLICATION RECORD

CITATION MAP

EXTRACTION MAP

CLAIMS

  • No claims are published for this paper.

CONCEPTS

  • No concepts are published for this paper.

REFERENCES

Showing 1-41 of 41 references · Page 1 of 1

CITED BY