The study, published in SN Social Sciences by researchers at the University of Fortaleza, focused on seven children aged nine to eleven, all with a documented clinical diagnosis of Autism Spectrum Disorder, enrolled in the fourth and fifth grades of elementary school. Each child had just taken part in the Young Scientists Project, a structured visit to the university’s Morphofunctional Sciences Laboratory designed specifically to accommodate the sensory, communicative, and learning profiles of students with autism. During the visit, the children handled non-toxic anatomical models of bones, muscles, and organs, explored interactive software, watched educational videos, and tried simple experiments, all under the guidance of university professors and researchers wearing white lab coats, masks, and gloves.

Crucially, the drawing task was not administered on the spot. To capture consolidated mental representations rather than immediate sensory recall, the researchers waited approximately twenty-four hours before asking the children, in a quiet and familiar classroom, to follow the standard DAST instruction: draw a scientist. No additional guidance was given, and any text that appeared in the drawings was included spontaneously by the children. The researchers then analyzed the seven images using an adapted version of the Draw-a-Scientist Test Checklist, a standardized coding framework that scores features such as lab coats, glasses, laboratory instruments, chemical formulas, computers, and gender cues.

The coding process itself was unusual. As an exploratory screening step, the team fed high-resolution digital copies of the drawings to ChatGPT-4o, a multimodal vision-language model, with a standardized prompt asking it to evaluate the images against the core DAST checklist indicators. Two human researchers then independently audited the model’s textual descriptions against the original physical drawings, correcting misclassifications and adding contextual interpretation. In line with Springer Nature’s guidelines on responsible artificial intelligence use, the AI was never treated as an autonomous codifier; all final coding decisions remained with the human researchers, who explicitly acknowledge that no formal inter-rater reliability testing, such as Cohen’s kappa, was performed in this pilot phase.

The results paint a picture of remarkable persistence. Six of the seven drawings—85.7 percent—contained male visual cues such as short hair or traditionally male clothing, while only one drawing featured a female-coded scientist, identifiable by long hair and a headband accessory. Four drawings placed the scientist in a laboratory environment, and three included specific laboratory instruments such as glassware and experimental apparatus. Just one drawing showed a lab coat. None of the children drew glasses, disheveled hair, books or papers, chemical formulas, or computer technology. Representations of scientists working in the field, using digital tools, or engaging with the broader modern scientific enterprise were entirely absent from the cohort’s output.

One drawing stood out from the conventional pattern. Rather than a laboratory, it depicted what the researchers interpreted as an office or clinical setting, complete with chairs, a desk, and a wall-mounted height measurement scale, accompanied by the spontaneously written word METICO—a phonetic misspelling of the Portuguese word for doctor. The researchers suggest this child’s representation may reflect the healthcare and morphology elements of the previous day’s laboratory experience, where anatomical models and researchers in clinical attire featured prominently. Without post-drawing interviews, however, the team is careful to frame such interpretations as contextual readings rather than conclusions about the child’s inner mental model, noting that categorizing such features strictly as conventional laboratory equipment risks oversimplifying the child’s intent.

The authors are equally careful about what the gender findings do and do not mean. The predominance of male-coded figures, they argue, should not be read as proof that these children consciously hold fixed gender stereotypes. Instead, it reflects how historically consolidated images of scientists serve as readily accessible graphical templates for young learners, a pattern documented repeatedly since Chambers first introduced the Draw-a-Scientist Test in 1983. The visual presence of male indicators and laboratory equipment is a descriptive finding about graphic tropes, not a definitive window into underlying beliefs—a distinction the study maintains rigorously throughout its analysis.

The pilot’s limitations are considerable and openly acknowledged. With only seven participants, no statistical analysis is possible, and the sample cannot represent the enormous heterogeneity of neurodivergent learners, whose cognitive and communicative profiles vary widely. There was no neurotypical control group for comparison, no categorization of participants’ granular clinical profiles, and no draw-and-tell interviews to clarify ambiguous graphic elements. The twenty-four-hour interval between the laboratory visit and the drawing task means recent experiential retention cannot be ruled out as an influence. The researchers also note that the classic prompt itself tends to elicit cultural stereotypes rather than nuanced personal views, and they recommend that future studies try alternative instructions, such as asking children to draw a scientist they admire, alongside structured verbal interviews and formal dual-coding frameworks for evaluating AI-assisted analysis.

Why does this small study matter beyond its sample size? Because neurodivergent individuals may be among the most under-tapped talent pools in science. Research cited by the authors indicates that people with autism often display above-average abilities in pattern recognition, memorization, and mathematics, and that neurodiverse thinkers frequently perceive problems and solutions in ways that drive innovation. Yet these same individuals face systemic barriers to science engagement and identity formation, and they often lack equitable access to STEM learning opportunities. If the figure of the scientist that children carry in their minds shapes their motivation and their sense of who belongs in science, then understanding how neurodivergent children visualize that figure is a first step toward building pedagogical strategies that help them see themselves in it.

The study’s working hypotheses point toward concrete educational experiments. Science curricula, the authors suggest, could move beyond the traditional chemistry laboratory to showcase field research, computational modeling, and healthcare innovation. Exposure to diverse role models—including neurodivergent researchers, women, and other underrepresented groups—could help students build personal identification with STEM roles. Multimodal, sensory-adapted learning tools such as interactive anatomical models, visual schedules, and accessible digital platforms offer a promising route for engaging neurodivergent learners on their own terms. The experience of Drawing 3, with its clinical setting and phonetically spelled doctor, hints that even a single sensory-rich laboratory visit can begin to broaden a child’s mental model of scientific practice beyond media-driven stereotypes. Whether that broadening can be sustained, generalized, and translated into lasting scientific identity among neurodivergent children is precisely the question this modest but methodologically inventive pilot study was designed to raise—and future research with larger cohorts, longitudinal designs, and direct measures of student engagement will now have to answer it.

Subject of Research: How neurodivergent children with autism represent scientists through drawings, assessed with the Draw-a-Scientist Test and AI-assisted visual coding

Article Title: The image of scientists among neurodivergent children: a pilot study

Article References: The image of scientists among neurodivergent children: a pilot study. (n.d.). https://doi.org/10.1007/s43545-026-01842-6

Image Credits: AI Generated

DOI: 10.1007/s43545-026-01842-6

Keywords: Draw-a-Scientist Test, neurodivergence, autism spectrum disorder, scientist stereotypes, STEM education, inclusive education, science identity, ChatGPT-4o, multimodal AI, elementary school, Brazil, pilot study

The study, published in Discover Artificial Intelligence, was designed to mirror how ordinary people actually use these tools. On a single day, 30 June 2025, the team submitted all fifteen questions to each model through standard consumer web interfaces, exactly as a patient would, with no API access, no custom settings, and no prompt engineering. The questions themselves were drawn exhaustively from the official NHS page of common questions about pregabalin, a resource compiled and validated by clinical pharmacists for the general public. Each query came wrapped in a standardized instruction asking for a clear, factual, patient-appropriate answer that must include a warning if any potential danger existed. That explicit safety instruction is what makes the study’s central finding so striking.

Pregabalin was a deliberate and timely choice. The drug is a structural analogue of the neurotransmitter GABA, but it does not act on GABA receptors directly; instead it binds selectively to the alpha-2-delta-1 subunit of voltage-gated calcium channels, dampening the release of excitatory neurotransmitters. Its clinical footprint has expanded dramatically over the past decade. Global annual sales of gabapentinoids, the drug class that includes pregabalin, rose by 114.5 percent between 2012 and 2022, and by a staggering 180.9 percent in low- and middle-income countries, where specialist care is scarce and patients are especially likely to seek answers from digital platforms. Prescribing rates more than doubled over ten years, reaching 7.2 prescriptions per 100 patients in 2019, with the highest volumes among elderly patients juggling multiple medications. Off-label and recreational use have broadened the population asking questions even further.

To judge the answers, the researchers assembled a multidisciplinary panel of a rheumatologist, a neurologist, and a pharmacologist, each with at least five years of clinical experience, reflecting the three main specialties that prescribe and monitor pregabalin. The evaluators worked independently and blind to which model had produced each response, after a calibration session in which they jointly scored three practice answers to align their criteria. Six dimensions were assessed: warning compliance, accuracy, completeness, content safety, readability, and clinical relevance. The study followed the TRIPOD-LLM reporting guideline, an emerging standard for research involving large language models, and the team calculated inter-rater reliability using Fleiss’ kappa and intraclass correlation coefficients.

The headline numbers on accuracy look encouraging. ChatGPT-4o, Gemini 2.0 Flash, and Copilot each scored 0.933 for factual correctness, while Claude 3.5 Sonnet scored 0.867. Clinical relevance ranged from 0.867 to 1.000 across the four models. But the confidence intervals overlapped so extensively that no model could claim reliable superiority over any other, a point the authors emphasize repeatedly. With only fifteen questions and a single query per model, the differences fall well within the statistical noise. The researchers are candid about this: the point estimates are best read as an illustration of the evaluation framework rather than stable performance benchmarks, and the single-query design means the stochastic variability of these models, which can produce substantively different answers to identical prompts, was never characterized.

The safety findings are where the study bites. Every one of the fifteen questions was judged by panel consensus to carry a clinically relevant safety concern warranting a warning, and the prompt explicitly demanded one. Yet warning compliance scores ranged only from 0.467 for ChatGPT-4o and Gemini 2.0 Flash to 0.733 for Claude 3.5 Sonnet. In plain terms, the models ignored a direct safety instruction in roughly 27 to 53 percent of cases. The authors argue this is arguably more alarming than a simple failure to volunteer warnings, because it demonstrates that even explicit prompting cannot guarantee reliable safety communication. For a drug with dependence potential, meaningful interaction risks, and a heavy presence among polypharmacy patients, that gap between instruction and behavior is the study’s most clinically consequential result.

When the researchers dissected the errors, a clear pattern emerged. Oversimplification dominated, accounting for 56.3 percent of all recorded errors, followed by omission of essential information at 34.8 percent, with outright hallucinations making up 8.9 percent. Gemini 2.0 Flash logged the most errors at 32, while Claude 3.5 Sonnet logged the fewest at 25. The authors connect these patterns to plausible, though unproven, architectural explanations: token probability maximization may favor statistically common, simplified phrasing over clinically nuanced detail; attention-based architectures may underweight rare but critical content; and reinforcement learning from human feedback may optimize for fluency and consensus at the cost of completeness. Hallucinations, though rarest, remain the most dangerous failure mode, since a fabricated dosing recommendation or a fictitious drug interaction could translate directly into harm.

The study is notable as much for its transparency as for its findings. In an unusual move, the authors disclose that the rater-by-item scoring matrix and the original response corpus were not archived after analysis, that the rule used to collapse three raters’ scores into one per-question value was never recorded, and that formal paired statistical tests such as McNemar’s test therefore became impossible. Copilot’s characteristic inline citations may also have compromised the blinding protocol. Rather than presenting plausible reconstructions as fact, the team reports these gaps openly, framing the work as an exploratory evaluation whose primary contribution is the multidimensional framework itself and a model for honest documentation of methodological limitations in early-stage LLM research.

What should be done with these results? The authors recommend hybrid architectures that ground chatbot outputs in structured, validated pharmaceutical databases rather than relying on parametric knowledge alone, along with dedicated fine-tuning on pharmacovigilance corpora so that safety alerts are generated reliably rather than left to general instruction-following. They also propose uncertainty detection mechanisms that would signal a model’s limits and steer users toward professional consultation, and even a continuous error-monitoring system modeled on existing pharmacovigilance infrastructure. The limitations are real: one medication, fifteen questions, one query per model, English only, and no testing with actual patients. The team, based in Francophone West and Central Africa, already plans a follow-up in French, where the stakes may be even higher. For now, the message to patients is clear: chatbots can be a reasonable starting point for basic medication facts, but when it comes to the warnings that keep you safe, they still cannot be trusted to speak up.

Subject of Research: Evaluation of large language models answering patient questions about the medication pregabalin

Article Title: A structured exploratory multidisciplinary evaluation of four large language models responding to frequently asked patient questions about pregabalin

Article References: A structured exploratory multidisciplinary evaluation of four large language models responding to frequently asked patient questions about pregabalin. (n.d.). https://doi.org/10.1007/s44163-026-02394-7

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02394-7

Keywords: large language models, ChatGPT-4o, Copilot, Gemini, Claude, pregabalin, patient safety, medication information, warning compliance, hallucination, pharmacovigilance, TRIPOD-LLM

The research team, led by Jussi S. Jauhiainen together with Agustín Garagorry Guerra, did not test the AI on toy examples or classroom quizzes. Instead, they turned to the Finnish matriculation examination, the national test that concludes upper secondary education and is enshrined in Finnish law. The dataset covered the open-ended written responses of 1,016 students, graded on the examination’s official 0-to-15 scale. This scale is narrow enough that even a two-point deviation can matter, and it demands nuanced judgment about argumentation, subject knowledge, and contextual use of concepts—precisely the qualities that have historically made automated essay scoring so difficult.

Technically, the researchers did not simply paste student answers into a chatbot and hope for the best. They deployed ChatGPT-4o within a reranked retrieval-augmented generation, or RAG, framework. In a RAG pipeline, the model first retrieves relevant reference material—in this case, grading criteria and model answers associated with each examination question—then reranks the retrieved content so the most pertinent information sits closest to the prompt, and finally generates its assessment grounded in that curated context. This architecture matters because it anchors the model’s judgment in the official grading instructions rather than relying purely on the statistical patterns encoded in its training data. The approach builds on the authors’ earlier work applying RAG-based grading to models such as GPT-3.5, GPT-4, Claude-3, and Mistral-Large, but the new study scales the method up and subjects it to far stricter comparison against human ground truth.

The headline result is arresting. Measured with descriptive statistics and correlation analyses, GPT-4o’s scores landed within two points of the official human grades in 75.00 percent of cases. Severe outliers—grades that diverged so far as to raise questions about fundamental misunderstanding—accounted for only 3.00 percent of responses. For a machine assigning integer scores on a 16-point scale to free-form, argumentative prose in a national exam, that level of agreement represents a genuine milestone. Earlier automated scoring systems, from latent semantic analysis in the late 1990s to modern BERT-based essay scorers, typically required extensive task-specific training on thousands of pre-graded examples. Here, a general-purpose language model, guided by retrieved grading criteria, approached human-level agreement largely out of the box.

Yet the study’s second experiment may prove the most consequential for the billions of people who do not speak English natively. Finnish is a low-resource language in the world of large language models: it appears far less frequently in training corpora than English, and its rich morphology and flexible word order pose additional challenges for statistical language processing. When the researchers translated the students’ Finnish responses into English and had the model grade the translations, alignment with human examiners jumped from 75.00 percent to 85.00 percent. The implication is sobering and empowering at once. The same AI system performs meaningfully better in a high-resource language than in a low-resource one, but a relatively simple translation step can recover much of the lost accuracy. For education systems in smaller language communities, that could determine whether AI-assisted grading is a dangerous shortcut or a viable supplement.

The study also probed a subtler question: not just whether the machine’s final scores matched human scores, but whether the machine was grading for the right reasons. The researchers examined whether GPT-4o recognized grading-relevant keywords—the technical terms and concepts that official criteria require examiners to spot. Broadly, it did. The model identified relevant keywords in the great majority of responses, which helps explain its strong score alignment. But the failures were revealing. In a minority of cases, the model misinterpreted the contextual usage of a term, registering a keyword as present when a student had used it incorrectly or in a way that did not satisfy the criterion, or occasionally overlooking legitimate conceptual content expressed in unusual phrasing. Those misreadings translated directly into grading errors, underscoring that surface-level keyword matching is not the same as genuine comprehension.

This distinction between matching scores and matching understanding sits at the heart of the paper’s conclusions. The authors are explicit that the findings demonstrate both the promise and the current limitations of LLM-based assessment. A model that agrees with human graders three-quarters or more of the time is a powerful tool, but the residual disagreements are not random noise; they cluster exactly where human interpretive depth matters most—irony, implicit reasoning, unconventional but valid arguments, and nuanced use of domain concepts. High-stakes examinations leave no room for such errors to be dismissed as acceptable variance when a student’s university admission or graduation certificate hangs in the balance.

The practical prescription that emerges is not replacement but supervision. The researchers argue that large language models hold significant potential as supplementary grading tools, particularly in high-resource languages, while cautioning that they do not yet match the consistency or interpretative depth of trained human evaluators. Deploying them responsibly, the study suggests, requires human oversight of every consequential grade, rigorous validation against official outcomes before any deployment, and careful attention to language effects—especially the systematic penalty that low-resource languages appear to suffer. The authors also frame the technology within the broader debate on trust and transparency in AI: teachers’ willingness to accept AI-powered tools has been shown to depend on demonstrated reliability, and the so-called black-box problem remains a psychological and institutional barrier even when statistical performance is strong.

The context of the study amplifies its significance. Finland’s matriculation examination is a legally regulated national institution, with data access restricted by the Finnish National Agency for Education, which makes any benchmark against its official grades unusually authoritative. Open-ended questions on such exams have long been the last bastion of exclusively human assessment; multiple-choice testing was automated decades ago, but essays, structured arguments, and analytical responses resisted automation because grading them is an act of interpretation. If a general-purpose model with a well-designed retrieval framework can now approach human agreement on that terrain—and if translation can substantially close the gap for smaller languages—then the question facing ministries of education is no longer whether AI can grade, but under what safeguards it should be allowed to try.

What comes next, the researchers suggest, is continued benchmarking across subjects, question types, and languages, alongside the development of hybrid workflows in which the machine proposes and the human disposes. The vision is not an examination hall emptied of teachers, but one in which AI handles the first pass at enormous scale, flags uncertain cases, and frees human experts to focus their judgment where it counts. The Finnish results show that such workflows are technically plausible today, with measured agreement rates that would have seemed fanciful only a few years ago. They also show, with equal clarity, that the last few percentage points of agreement—the difference between a machine that assists grading and a machine that can be trusted to grade alone—remain stubbornly human territory. For now, the examiners of Finland, and everywhere else, can breathe a measured sigh of relief: the machine is close, but not close enough, and the study that proved it may shape how the world’s schools adopt AI for years to come.

Beyond the headline agreement rates, the study’s methodology reflects a broader shift in how automated assessment is evaluated. Rather than relying on a single aggregate statistic, the researchers combined descriptive statistics with correlation analyses and keyword-recognition checks, giving a multidimensional picture of where the model succeeded and failed. This kind of layered evaluation is increasingly seen as essential, because two grading systems can produce identical score distributions while reasoning about student answers in fundamentally different ways.

The translation experiment also carries implications for how multilingual capability is measured in language models. Performance gaps between high- and low-resource languages are well documented in the machine learning literature, but this study quantifies the gap in a concrete, consequential setting: a legally regulated national examination. The finding that translating responses into English improved alignment by ten percentage points suggests that language itself, not the difficulty of the exam content, was a major source of grading error.

For institutions considering similar deployments, the Finnish data restrictions highlight a practical constraint. Because the examination data could not be shared publicly, independent replication remains limited, reinforcing the authors’ call for rigorous local validation before any AI-assisted grading system is adopted in high-stakes contexts.

Subject of Research: Benchmarking ChatGPT-4o against human expert grading of open-ended responses in Finland's national high-stakes matriculation examination across high- and low-resource languages

Article Title: Evaluating Open-Ended High-Stakes Examinations with LLMs: Alignment Between ChatGPT-4o and Human Grading in High- and Low-Resource Languages

Article References: Jauhiainen, J. S., & Guerra, A. G. (2026). Evaluating Open-Ended High-Stakes Examinations with LLMs: Alignment Between ChatGPT-4o and Human Grading in High- and Low-Resource Languages. Frontiers of Digital Education, 3(2), Article 17. https://doi.org/10.1007/s44366-026-0091-1

Image Credits: AI Generated

DOI: 10.1007/s44366-026-0091-1

Keywords: ChatGPT-4o, large language models, automated grading, high-stakes examination, Finnish matriculation examination, retrieval-augmented generation, open-ended assessment, low-resource languages, educational assessment, human-AI agreement, AI in education, automated essay scoring