Answers are grounded in this paper and cite their sources.
This longitudinal study investigates how postgraduate translation students appraise formative feedback delivered by human instructors versus generative artificial intelligence (DeepSeek‐V2.5) across an 8‐month academic curriculum. Utilizing an authentic classroom repeated‐measures design across eight instructional cycles, we tracked multi‐wave evaluations across eleven perception items grouped under four descriptive domains: Objectivity, Usefulness, Genuineness, and Credibility. Paired comparisons and linear mixed‐effects modeling revealed that while students acknowledged the automated system's competitive technical precision, human teacher feedback maintained a consistent advantage in perceived factual grounding, fairness, and relational trustworthiness. Crucially, item‐level and longitudinal trajectory analyses identified a targeted temporal divergence: instructor ratings remained stably high, whereas GenAI feedback showed a progressive decline in perceived genuineness driven specifically by perceived sincerity rather than authenticity. Text‐derived sentiment modeling further indicated an exploratory contrast within each source's score range: for human instructors, highly corrective, less positive commentary was appraised as more objective; conversely, elevated algorithmic sentiment positivity was associated with reduced perceived genuineness. Overall, these findings show that students in this cohort consistently rated detailed instructor feedback more favorably than default DeepSeek‐V2.5 commentary—particularly on credibility‐ and sincerity‐related items over time—suggesting that translation training should pair AI's surface‐level error screening with human instructors’ diagnostic guidance rather than relying on unmonitored automated feedback.
No takes yet. Share an insight, caveat, or question.
Ye et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: