Agentic education · Learning science

Actionable companion · Feedback & scaffolding

Learning with agency: Navigating AI feedback in translation training: how text type, proficiency, and attitude shape students’ acceptance behaviors

AI-assisted editorial implementation guide, separate from the original research. Proposed actions are not claims or recommendations attributed to the source authors. No independent peer review of this guide is asserted.

Read original paperSearch paper agentCC BY 4.0

Evidence and context

Title and abstract screening; stage labels describe reported settings, not verified numerical ages. Design labels require full-text appraisal. Reported stage(s): Secondary school, Higher education. Verify the participant population and task before transferring the findings.

Original authors’ abstract, reproduced under the linked source license:

This study investigates how undergraduate and graduate translation students in post-secondary education engage with and evaluate Large Language Model (LLM)-generated feedback through a mixed-methods approach, analyzing acceptance rates, influencing factors, decision rationales, and perceived limitations. 78 students majoring in translation (55 undergraduates, 23 postgraduates) completed translation tasks spanning six text types and received ChatGPT-3.5-generated feedback. Participants made binary accept/reject decisions with immediate written rationales, followed by semi-structured interviews to explore evaluative criteria and perceived deficiencies. Quantitatively, participants accepted an average of 68.2% of LLM suggestions, demonstrating receptive yet selective engagement, with no student accepting or rejecting all suggestions. Acceptance was most strongly shaped by text type, with technical and news texts receiving the highest approval and literary and tourism texts the lowest. Baseline AI attitude and proficiency further moderated engagement, as optimists accepted more suggestions than skeptics, higher-proficiency students within each academic level demonstrated greater criticality than their lower-proficiency peers, and postgraduates overall exhibited more selective evaluation than undergraduates. Qualitatively, students accepted feedback that corrected objective errors, improved fluency, or resolved uncertainty, but rejected suggestions due to cultural or contextual misunderstandings, preservation of personal style, unconvincing justifications, or risk aversion in high-stakes texts. Rejection often triggered deeper engagement, including self-revision, external verification, and dialogue with the AI. Thematic analysis revealed key deficiencies in LLM feedback, such as cultural blind spots, stylistic flattening, and contextual myopia. These findings highlight that students’ engagement with LLM feedback is shaped by the interplay of task characteristics, individual dispositions, and domain expertise, underscoring the need for translator training programs to develop feedback systems that are contextually aware, stylistically adaptive, and dialogic in post-secondary translation education.

What kind of evidence is this?

Supporting AI learning science · Qualitative / mixed methods

Stage: Secondary school; Higher education. Screening labels describe reported context and study-design language. They are not a risk-of-bias rating or proof of effectiveness for an age band.

Study methods · source passage

The present study adopted an explanatory sequential mixed-methods research design, in which quantitative behavioral data were collected and analyzed first to identify patterns of students’ acceptance of AI-generated translation feedback, followed by qualitative data to explain the underlying mechanisms driving these patterns. This design was selected to align explicitly with the three research questions, which required both (a) systematic measurement of acceptance behaviors and (b) in-depth interpretation of learners’ reasoning, perceptions, and decision-making processes. In the quantitative phase, students’ accept/reject decisions, baseline AI attitudes, and interaction logs were analyzed to examine overall acceptance trends and group differences. In the subsequent qualitative phase, written rationales and semi-structured interviews were used to interpret how students evaluated AI feedback in relation to translation quality, task demands, and self-perceived competence. This study attempts to uncover how and why translation students, defined as students formally enrolled in undergraduate or postgraduate programs in Translation or Translation and Interpreting, engage with AI-generated feedback in pedagogical contexts.

Limitations and uncertainty · source passage

This study offers a comprehensive account of how translation students engage with feedback generated by LLM. Participants demonstrated a pattern of selective rather than unconditional trust in the system. Importantly, acceptance decisions were shaped by a hierarchy of influences: text genre emerged as the most decisive factor, with factual and objective genres eliciting greater reliance, followed by students’ baseline attitudes toward AI, where more optimistic learners were more receptive, and finally by proficiency level, with more experienced students showing heightened criticality. The qualitative analysis further illuminated these dynamics. Students tended to accept corrections that clearly resolved linguistic errors, enhanced fluency, or were accompanied by convincing explanations. Conversely, they rejected suggestions perceived as culturally insensitive, stylistically reductive, inadequately justified, or unnecessarily risky. Notably, rejection did not signify disengagement: in many cases, students pursued additional cognitive strategies such as self-revision or independent research, suggesting that even disregarded feedback can catalyze deeper learning. Taken together, these results highlight the dynamic interplay between AI-generated suggestions, student judgment, and task context. While students valued LLMs as reliable tools for formal accuracy checks, they remained skeptical of their capacity to address creativity and cultural nuance. This has clear pedagogical implications: translation educators should equip students with the critical literacy needed to engage with AI judiciously, while developers of feedback systems should prioritize explainability, contextual sensitivity, and adaptability. Ultimately, the integration of LLMs into translator training will depend on cultivating calibrated trust, enabling learners to discern when to rely on AI support and when to exercise independent expertise, thereby maximizing the benefits of human, AI collaboration without undermining educational or translational quality.

Design the system before the pilot

Use this study to inform a learning mechanism or implementation question. It is not direct evidence that autonomous agents improve learning. Keep the proposed agent behavior separate from what the source actually tested.

  • Give feedback on the learner’s reasoning, then fade hints. Record whether the learner can complete the next task without that scaffold.
  • Measure engagement separately from achievement. Check whether novelty-driven participation persists and whether it translates to independent performance.

A 90-day district pilot

  1. Days 1–15 · Appraise. Have the curriculum lead and a teacher read the methods and limitations. Record the study design, sample, comparator, outcomes and evidence gaps. Define a local question that the study can help inform.
  2. Days 16–30 · Design. Ask learners to attempt a problem unaided, compare an agent hint, explain a revision and verify the final reasoning against a trusted source. Make the permitted use explicit for each assignment. Name an accountable owner and set the baseline, success threshold and budget before implementation.
  3. Days 31–60 · Test. Start with a small voluntary cohort, preserve an accessible non-AI route and require teacher review. Use synthetic or de-identified examples; approve data handling and tool terms through district processes.
  4. Days 61–90 · Decide. Measure unaided transfer and delayed retention, calibration of confidence and the ability to identify an incorrect AI explanation. Record whether help-seeking becomes answer copying. Compare with baseline, document uncertainty and stop or revise if learning worsens, access is unequal or critical errors persist.

The board decision brief

Report the learning problem, evidence quality, participating population, total cost including review time, measured benefit, errors, access differences and remaining uncertainties. Make a dated recommendation to stop, revise or expand. A positive perception survey alone does not establish learning improvement.

Attribution and reuse

Shiyue Chen; Jie Lou. Navigating AI feedback in translation training: how text type, proficiency, and attitude shape students’ acceptance behaviors. Frontiers in Artificial Intelligence (2026-03-12). https://doi.org/10.3389/frai.2026.1727544

Companion prepared 2026-10-30. Original authorship and DOI are preserved. The abstract is an attributed source extract; the implementation plan is new AI-assisted editorial material, offered under the same CC BY 4.0 license. This guide has not undergone independent peer review.