Actionable companion · Learning design & outcomes
From AI evidence to responsible practice: Automation bias in teachers’ evaluation of student writing: effects of algorithmic warnings and visual risk cues in AI detection reports
AI-assisted editorial implementation guide, separate from the original research. Proposed actions are not findings or recommendations attributed to the source authors. No independent peer review of this guide is asserted.
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): Higher education. Verify the participants, intervention and comparator before applying this study in a new setting.
Original authors’ abstract, reproduced under the linked source license:
As generative artificial intelligence becomes integrated into higher education, teachers increasingly rely on AI text-detection reports to support judgments about authorship, writing quality, and academic integrity. Existing research has mainly examined detector accuracy, false positives, fairness, and policy; less is known about whether report design itself shapes teachers’ evaluations when the judged text is unchanged. This gap matters because numerical scores and visual warnings may frame interpretation, anchor suspicion, and encourage confirmatory reading under uncertainty. Here, we tested how algorithmic warning strength and visual risk cues affect teachers’ evaluations of student writing in a controlled single-stimulus experiment. In a 2 × 2 between-subjects experiment, 214 university teachers evaluated the same medium-quality Chinese social-science course paper accompanied by a fictitious AI-detection report that varied by AI detection rate (7% vs. 87%) and red-highlighting/report-presentation package (absent vs. present). A high detection rate increased perceived AI authorship likelihood and risk and lowered overall quality evaluations, percentage-based scores, originality, language expression, and logical structure. Red highlighting also influenced report perception, language-expression judgments, and self-reported intervention tendency. Significant warning × highlighting interactions emerged for percentage-based scoring, originality, language expression, logical structure, overall multidimensional quality, and intervention tendency, but not for the 1–10 overall rating or report perception. These preliminary and context-specific findings suggest that AI detection reports may function not merely as technical outputs but as socio-technical judgment environments under controlled evaluative conditions. Numerical warnings may anchor teachers’ evaluations, while visual risk cues may selectively amplify suspicion and intervention-oriented responses. Responsible use of AI detection therefore requires neutral report design, independent teacher judgment, human oversight, and training on automation bias.
What kind of evidence is this?
Supporting AI learning science · Design requires full-text appraisal
Stage: 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
This study used a 2 × 2 between-subjects design. The first independent variable was algorithmic warning strength, operationalized as a low AI detection rate (7%) versus a high AI detection rate (87%). The second independent variable was visual risk cue, operationalized as the absence versus presence of a red-highlighting/report-presentation package. The four conditions were Group A (low/no highlighting), Group B (low/red), Group C (high/no highlighting), and Group D (high/red). Factorial between-subjects designs are appropriate when researchers need to estimate main effects and interactions while preserving the temporal priority of experimentally assigned manipulations (Lakens, 2013; Richardson, 2011; Simmons et al., 2011).
Limitations and uncertainty · source passage
This study has several limitations. First, it used a fictitious detection report, an online experimental setting, and a single medium-quality Chinese social-science paper. Although this design improved experimental control, it limits ecological and external validity. Moreover, the medium-quality level of the stimulus paper was not independently validated through a pilot study or external rater calibration. The effects of detection reports might differ for very high-quality or very poor papers, where teachers’ prior quality expectations could moderate the influence of algorithmic warnings and visual risk cues. Future research should replicate the findings with real detection interfaces, real grading procedures, independently validated stimulus materials, and papers varying in discipline, language, topic, and quality level.
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.
- Specify a measurable learning objective, a comparator and a task without tool assistance before adopting a system.
A 90-day district pilot
- 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.
- Days 16–30 · Design. Choose one disciplinary task. Require a first attempt, a documented AI consultation, a critique of the output and an independent explanation. Align the comparison group and scoring rubric with the same objective. Name an accountable owner and set the baseline, success threshold and budget before implementation.
- 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.
- Days 61–90 · Decide. Blind-score reasoning and accuracy on a new task without AI. Distinguish model examination scores, student perceptions and actual learner gains; only the last directly tests educational benefit. 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
Peitao Du; Tingting Liu; Xujin Xian. Automation bias in teachers’ evaluation of student writing: effects of algorithmic warnings and visual risk cues in AI detection reports. Frontiers in Psychology (2026-07-07). https://doi.org/10.3389/fpsyg.2026.1889402
Companion prepared 2026-11-01. 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.