Agentic education · Learning science

Actionable companion · Feedback & scaffolding

Learning with agency: Assessing the capabilities of AI-based large language models (AI-LLMs) in interpreting histopathological slides and scientific figures: Performance evaluation study

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): Age / stage not established. Verify the participant population and task before transferring the findings.

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

Background Integrating artificial intelligence-based large language models (AI-LLMs) into medical and other scientific domains is increasingly recognized as a tool to support complex tasks, such as interpreting histopathology slides and scientific figures. AI-LLMs can simplify these processes by providing clearer explanations. By improving accessibility and comprehension, AI-LLMs can significantly assist healthcare professionals in diagnosing and therapy determination. Students and the public also find it easier to understand complex scientific concepts and images. Objectives This study explores the capability of AI-LLMs in interpreting histopathological slides and scientific images. This study aims to evaluate the performance of AI-LLMs in supporting diagnostics and improving comprehension in biomolecular sciences. Methods The study was divided into two parts: interpreting histopathology slides and scientific figures. Twelve histopathology images and twelve scientific figures were tested on each of the three most frequently used chatbots (ChatGPT-4, Gemini Advanced, and Copilot). Responses from the chatbots were coded and blindly examined by expert raters using five parameters—relevance, clarity, depth, focus, and coherence—on a 5-point Likert scale. Statistical analysis included one-way ANOVA and multiple linear regression. Results ChatGPT-4 outperformed Gemini Advanced and Copilot in histopathology and scientific image interpretation (P < 0.001) with significantly higher scores across all parameters (relevance, clarity, depth, focus, and coherence). ChatGPT-4’s superior performance may be due to its advanced algorithms, extensive training data, specialized modules, and user feedback. Conclusions ChatGPT-4 excels in interpreting histopathology and scientific images, which may lead to improving diagnostic accuracy, clinical decision-making, and reducing pathologists’ workload. It also benefits education by enhancing students’ understanding of complex images and promoting interactive learning. ChatGPT-4 shows a significant potential to improve patient care and enrich student learning.

What kind of evidence is this?

Conversational & pedagogical agents · Design requires full-text appraisal

Stage: Age / stage not established. 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

In accordance with local legislation and institutional requirements, ethical review and approval were not required for the study of human participants.

Limitations and uncertainty · source passage

Although AI-LLMs demonstrate clear potential as educational tools for doctors and medical students, their application in diagnostic support remains in its early stages. The use of histopathology images in this study served as a simplified approach that does not fully capture the complexity of whole slide images (WSIs) encountered in routine pathology practice.

Design the system before the pilot

Define whether the agent asks questions, gives hints, explains or assesses. Test hallucinated feedback, excessive praise, misleading confidence and responses when the learner is confused. A conversational interface alone does not establish autonomous agency.

  • Give feedback on the learner’s reasoning, then fade hints. Record whether the learner can complete the next task without that scaffold.
  • Use common rubrics, blind scoring where feasible and a fresh unaided reasoning task. Document permissible assistance and avoid treating AI-detector scores as proof of misconduct.
  • Co-design accommodations with learners and specialists. Check access and error burden across groups without inferring subgroup effectiveness from small or unreported samples.

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. Treat the paper as background evidence until its population is verified in the full text. Do not assign an age-specific intervention or extend adult results to children. 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. Extract the participant ages or education stage, recruitment, task and comparator before selecting a local pilot. If these cannot be established, retain the paper for design discussion rather than age-band efficacy claims. 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

Khanisyah E. Gumilar; Grace Ariani; Priangga A. Wiratama; Rimbun; Tri H. Yuliawati; Hong Chen; Ibrahim H. Ibrahim; Cheng-Han Lin; Tai-Yu Hung; Dewanti Anggrahini; Arya S. Rajanagara; Khaled E. Omran; Zih-Ying Yu; Yu-Cheng Hsu; Erry G. Dachlan; Jer-Yen Yang; Li-Na Liao; Ming Tan. Assessing the capabilities of AI-based large language models (AI-LLMs) in interpreting histopathological slides and scientific figures: Performance evaluation study. BioMedicine (2026-03-01). https://doi.org/10.37796/2211-8039.1698

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.