Giora Alexandron

Licol Zeinfeld, Alona Strugatski, Ziva Bar-Dov et al.

The rapid adoption of large language models (LLMs) in education raises profound challenges for assessment design. To adapt assessments to the presence of LLM-based tools, it is crucial to characterize the strengths and weaknesses of LLMs in a generalizable, valid and reliable manner. However, current LLM evaluations often rely on descriptive statistics derived from benchmarks, and little research applies theory-grounded measurement methods to characterize LLM capabilities relative to human learners in ways that directly support assessment design. Here, by combining educational data mining and psychometric theory, we introduce a statistically principled approach for identifying items on which humans and LLMs show systematic response differences, pinpointing where assessments may be most vulnerable to AI misuse, and which task dimensions make problems particularly easy or difficult for generative AI. The method is based on Differential Item Functioning (DIF) analysis -- traditionally used to detect bias across demographic groups -- together with negative control analysis and item-total correlation discrimination analysis. It is evaluated on responses from human learners and six leading chatbots (ChatGPT-4o \& 5.2, Gemini 1.5 \& 3 Pro, Claude 3.5 \& 4.5 Sonnet) to two instruments: a high school chemistry diagnostic test and a university entrance exam. Subject-matter experts then analyzed DIF-flagged items to characterize task dimensions associated with chatbot over- or under-performance. Results show that DIF-informed analytics provide a robust framework for understanding where LLM and human capabilities diverge, and highlight their value for improving the design of valid, reliable, and fair assessment in the AI era.

Abigail Victoria Gurin Schleifer, Moriah Ariely, Beata Beigman Klebanov et al.

Automated short answer scoring (ASAS) is shifting from discriminative, fine-tuned models to large language models (LLMs) used in few-shot settings. This paradigm leverages LLMs broad world knowledge and ease of deployment, but limited task-specific data may reduce alignment on complex scoring tasks. In particular, its impact on scoring partially correct responses that require nuanced interpretation remains underexplored. We investigate the relationship between the degree of task-specific adaptation of different models and quality-conditioned scoring agreement. We compare three LLMs (GPT-5.2, GPT-4o, Claude Opus 4.5) in few-shot mode, a fine-tuned BERT-based encoder, and a human expert on two open-ended biology items, using several hundred student responses and ground truth scores provided by a biology education expert. The results show that human-human agreement is highest and stable across the full quality spectrum. All AI models perform well on fully correct and fully incorrect responses, but exhibit substantial degradation on mid-range responses. This mid-range degradation is conditioned on task-specific adaptation: It is most severe in few-shot LLMs with few examples and decreases as task-specific data increases, with fine-tuned encoder models performing best. This mid-range degradation may lead to inequitable evaluation of responses produced by students with developing understanding. Our findings highlight the importance of quality-conditioned fairness, with particular attention to mid-range responses.

Alona Strugatski, Giora Alexandron

Generative AI is transforming the educational landscape, raising significant concerns about cheating. Despite the widespread use of multiple-choice questions in assessments, the detection of AI cheating in MCQ-based tests has been almost unexplored, in contrast to the focus on detecting AI-cheating on text-rich student outputs. In this paper, we propose a method based on the application of Item Response Theory to address this gap. Our approach operates on the assumption that artificial and human intelligence exhibit different response patterns, with AI cheating manifesting as deviations from the expected patterns of human responses. These deviations are modeled using Person-Fit Statistics. We demonstrate that this method effectively highlights the differences between human responses and those generated by premium versions of leading chatbots (ChatGPT, Claude, and Gemini), but that it is also sensitive to the amount of AI cheating in the data. Furthermore, we show that the chatbots differ in their reasoning profiles. Our work provides both a theoretical foundation and empirical evidence for the application of IRT to identify AI cheating in MCQ-based assessments.

Abigail Gurin Schleifer, Beata Beigman Klebanov, Moriah Ariely et al.

Unsupervised clustering of student responses to open-ended questions into behavioral and cognitive profiles using pre-trained LLM embeddings is an emerging technique, but little is known about how well this captures pedagogically meaningful information. We investigate this in the context of student responses to open-ended questions in biology, which were previously analyzed and clustered by experts into theory-driven Knowledge Profiles (KPs). Comparing these KPs to ones discovered by purely data-driven clustering techniques, we report poor discoverability of most KPs, except for the ones including the correct answers. We trace this "discoverability bias" to the representations of KPs in the pre-trained LLM embeddings space.

Tanya Nazaretsky, Sara Hershkovitz, Giora Alexandron

Sequencing items in adaptive learning systems typically relies on a large pool of interactive assessment items (questions) that are analyzed into a hierarchy of skills or Knowledge Components (KCs). Educational data mining techniques can be used to analyze students performance data in order to optimize the mapping of items to KCs. Standard methods that map items into KCs using item-similarity measures make the implicit assumption that students performance on items that depend on the same skill should be similar. This assumption holds if the latent trait (mastery of the underlying skill) is relatively fixed during students activity, as in the context of testing, which is the primary context in which these measures were developed and applied. However, in adaptive learning systems that aim for learning, and address subject matters such as K6 Math that consist of multiple sub-skills, this assumption does not hold. In this paper we propose a new item-similarity measure, termed Kappa Learning (KL), which aims to address this gap. KL identifies similarity between items under the assumption of learning, namely, that learners mastery of the underlying skills changes as they progress through the items. We evaluate Kappa Learning on data from a computerized tutor that teaches Fractions for 4th grade, with experts tagging as ground truth, and on simulated data. Our results show that clustering that is based on Kappa Learning outperforms clustering that is based on commonly used similarity measures (Cohen Kappa, Yule, and Pearson).