AI Ed Researcher

Authors: Weiyue Li, Minda Zhao, Weixuan Dong, Jiahui Cai, Yuze Wei, Michael Pocress, Yi Li, Wanyan Yuan, Xiaoyue Wang, Ruoyu Hou, Kaiyuan Lou, Wenqi Zeng, Yutong Yang, Yilun Du, Mengyu Wang

Abstract: Large language models (LLMs) are increasingly used as automated evaluators, yet prior works demonstrate that these LLM judges often lack consistency in scoring when the prompt is altered. However, the effect of the grading scale itself remains underexplored. We study the LLM-as-a-judge problem by comparing two kinds of raters: humans and LLMs. We collect ratings from both groups on three scales and across six benchmarks that include objective, open-ended subjective, and mixed tasks. Using intraclass correlation coefficients (ICC) to measure absolute agreement, we find that LLM judgments are not perfectly consistent across scales on subjective benchmarks, and that the choice of scale substantially shifts human-LLM agreement, even when within-group panel reliability is high. Aggregated over tasks, the grading scale of 0-5 yields the strongest human-LLM alignment. We further demonstrate that pooled reliability can mask benchmark heterogeneity and reveal systematic subgroup differences in alignment across gender groups, strengthening the importance of scale design and sub-level diagnostics as essential components of LLM-as-a-judge protocols.

Link: https://arxiv.org/abs/2601.03444

Authors: Joslyn Orgill, Andra Rice, Max Fowler, Seth Poulsen

Abstract: The development of effective autograders is key for scaling assessment and feedback. While NLP based autograding systems for open-ended response questions have been found to be beneficial for providing immediate feedback, autograders are not always liked, understood, or trusted by students. Our research tested the effect of transparency on students’ attitudes towards autograders. Transparent autograders increased students’ perceptions of autograder accuracy and willingness to discuss autograders in survey comments, but did not improve other related attitudes — such as willingness to be graded by them on a test — relative to the control without transparency. However, this lack of impact may be due to higher measured student trust towards autograders in this study than in prior work in the field. We briefly discuss possible reasons for this trend.

Link: https://arxiv.org/abs/2601.00765

Authors: Mario Angelelli, Morena Oliva, Serena Arima, Enrico Ciavolino

Abstract: Artificial Intelligence (AI) and large language models (LLMs) are increasingly used in social and psychological research. Among potential applications, LLMs can be used to generate, customise, or adapt measurement instruments. This study presents a preliminary investigation of AI-generated questionnaires by comparing two ChatGPT-based adaptations of the Body Awareness Questionnaire (BAQ) with the validated human-developed version. The AI instruments were designed with different levels of explicitness in content and instructions on construct facets, and their psychometric properties were assessed using a Bayesian Graded Response Model. Results show that although surface wording between AI and original items was similar, differences emerged in dimensionality and in the distribution of item and test information across latent traits. These findings illustrate the importance of applying statistical measures of accuracy to ensure the validity and interpretability of AI-driven tools.

Link: https://arxiv.org/abs/2510.24739

Authors: Bruno Florentino, Camila Sestito, Wellington Cruz, André de Carvalho, Robson Bonidia

Abstract: This study examines the perceptions of Brazilian K-12 education teachers regarding the use of AI in education, specifically General Purpose AI. This investigation employs a quantitative analysis approach, extracting information from a questionnaire completed by 346 educators from various regions of Brazil regarding their AI literacy and use. Educators vary in their educational level, years of experience, and type of educational institution. The analysis of the questionnaires shows that although most educators had only basic or limited knowledge of AI (80.3\%), they showed a strong interest in its application, particularly for the creation of interactive content (80.6%), lesson planning (80.2%), and personalized assessment (68.6%). The potential of AI to promote inclusion and personalized learning is also widely recognized (65.5%). The participants emphasized the importance of discussing ethics and digital citizenship, reflecting on technological dependence, biases, transparency, and responsible use of AI, aligning with critical education and the development of conscious students. Despite enthusiasm for the pedagogical potential of AI, significant structural challenges were identified, including a lack of training (43.4%), technical support (41.9%), and limitations of infrastructure, such as low access to computers, reliable Internet connections, and multimedia resources in schools. The study shows that Brazil is still in a bottom-up model for AI integration, missing official curricula to guide its implementation and structured training for teachers and students. Furthermore, effective implementation of AI depends on integrated public policies, adequate teacher training, and equitable access to technology, promoting ethical, inclusive, and contextually grounded adoption of AI in Brazilian K-12 education.

Link: https://arxiv.org/abs/2512.23834

Authors: Ting Wang, Caroline Prendergast, Susan Lottridge

Abstract: Large language models (LLMs) have emerged as powerful tools for generating domain-specific multiple-choice questions (MCQs), offering efficiency gains for certification boards but raising new concerns about examination security. This study investigated whether LLM-generated items created with proprietary guidance differ meaningfully from those generated using only publicly available resources. Four representative clinical activities from the American Board of Family Medicine (ABFM) blueprint were mapped to corresponding Entrustable Professional Activities (EPAs), and three LLMs (GPT-4o, Claude 4 Sonnet, Gemini 2.5 Flash) produced items under a naive strategy using only public EPA descriptors, while GPT-4o additionally produced items under a guided strategy that incorporated proprietary blueprints, item-writing guidelines, and exemplar items, yielding 160 total items. Question stems and options were encoded using PubMedBERT and BioBERT, and intra- and inter-strategy cosine similarity coefficients were calculated. Results showed high internal consistency within each prompting strategy, while cross-strategy similarity was lower overall. However, several domain model pairs, particularly in narrowly defined areas such as viral pneumonia and hypertension, exceeded the 0.65 threshold, indicating convergence between naive and guided pipelines. These findings suggest that while proprietary resources impart distinctiveness, LLMs prompted only with public information can still generate items closely resembling guided outputs in constrained clinical domains, thereby heightening risks of item exposure. Safeguarding the integrity of high stakes examinations will require human-first, AI-assisted item development, strict separation of formative and summative item pools, and systematic similarity surveillance to balance innovation with security.

Link: https://arxiv.org/abs/2512.23729