MMMohammad Moein

Mohammad Moein

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Mohammad Moein has a total of 27 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Designing Teacher-Centered Interfaces to Elicit Pedagogical Context for LLM-Based Content Generation

    by Abdolali Faraji, Mohammadreza Tavakoli, Mohammad Moein, Mohammadreza Molavi, Gábor Kismihók

    Large Language Models (LLMs) can help teachers generate and adapt educational materials, but most interactions still rely on free-form prompting, making instructional intent difficult to express and reuse. In this paper, we study how teacher-facing interfaces can elicit pedagogically meaningful contextual information and reduce ad hoc prompt engineering. We first conducted a requirement analysis with $$N=100$$ N = 100 teachers across 17 contextual fields. We measured importance and introduced fill deficit to identify “friction points” (i.e., information teachers consider important but are less likely to provide). Based on this analysis, we implemented a context-rich interface with targeted supports for important and friction-point fields. In a separate user study with a different sample of $$N=100$$ N = 100 teachers, the interface achieved high usability (SUS mean $$=80.33$$ = 80.33 ), low perceived workload, and increased willingness to provide contextual information, particularly for friction-point fields. For further validation, we conducted a comparative pilot with $$N=10$$ N = 10 experienced teachers; participants reported a greater tendency to use content generated with our context-rich interface (72.2%) than content generated with a baseline GPT interface (37.0%).

  • Open Access

    Embedding-Based Rankings of Educational Resources based on Learning Outcome Alignment: Benchmarking, Expert Validation, and Learner Performance

    by Mohammadreza Molavi Hajiagha, Mohammad Moein, Mohammadreza Tavakoli, Abdolali Faraji, Gábor Kismihók, Stefan T. Mol

    As the online learning landscape evolves, the need for personalization is increasingly evident. Although educational resources are burgeoning, educators face challenges selecting materials that both align with intended learning outcomes and address diverse learner needs. Large Language Models (LLMs) are attracting growing interest for their potential to create learning resources that better support personalization, but verifying coverage of intended outcomes still requires human alignment review, which is costly and limits scalability. We propose a framework that supports the cost-effective automation of evaluating alignment between educational resources and intended learning outcomes. Using human-generated materials, we benchmarked LLM-based text-embedding models and found that the most accurate model (Voyage) achieved 79% accuracy in detecting alignment. We then applied the optimal model to LLM-generated resources and, via expert evaluation, confirmed that it reliably assessed correspondence to intended outcomes (83% accuracy). Finally, in a three-group experiment with 360 learners, higher alignment scores were positively related to greater learning performance, χ2(2, N = 360) = 15.39, p <.001. These findings show that embedding-based alignment scores can facilitate scalable personalization by confirming alignment with learning outcomes, which allows teachers to focus on tailoring content to diverse learner needs.

  • Open Access

    Embedding-Based Rankings of Educational Resources based on Learning Outcome Alignment: Benchmarking, Expert Validation, and Learner Performance

    by Mohammadreza Molavi, Mohammad Moein, Mohammadreza Tavakoli, Abdolali Faraji, Stefan T. Mol, Gábor Kismihók

    As the online learning landscape evolves, the need for personalization is increasingly evident. Although educational resources are burgeoning, educators face challenges selecting materials that both align with intended learning outcomes and address diverse learner needs. Large Language Models (LLMs) are attracting growing interest for their potential to create learning resources that better support personalization, but verifying coverage of intended outcomes still requires human alignment review, which is costly and limits scalability. We propose a framework that supports the cost-effective automation of evaluating alignment between educational resources and intended learning outcomes. Using human-generated materials, we benchmarked LLM-based text-embedding models and found that the most accurate model (Voyage) achieved 79% accuracy in detecting alignment. We then applied the optimal model to LLM-generated resources and, via expert evaluation, confirmed that it reliably assessed correspondence to intended outcomes (83% accuracy). Finally, in a three-group experiment with 360 learners, higher alignment scores were positively related to greater learning performance, chi-squared(2, N = 360) = 15.39, p < 0.001. These findings show that embedding-based alignment scores can facilitate scalable personalization by confirming alignment with learning outcomes, which allows teachers to focus on tailoring content to diverse learner needs.

  • Designing Effective LLM-Assisted Interfaces for Curriculum Development

    by Abdolali Faraji, Mohammadreza Tavakoli, Mohammad Moein, Mohammadreza Molavi, Gábor Kismihók


  • Open Access

    LLM-Driven Personalized Answer Generation and Evaluation

    by Mohammadreza Molavi, Mohammadreza Tavakoli, Mohammad Moein, Abdolali Faraji, Gábor Kismihók

    Online learning has experienced rapid growth due to its flexibility and accessibility. Personalization, adapted to the needs of individual learners, is crucial for enhancing the learning experience, particularly in online settings. A key aspect of personalization is providing learners with answers customized to their specific questions. This paper therefore explores the potential of Large Language Models (LLMs) to generate personalized answers to learners' questions, thereby enhancing engagement and reducing the workload on educators. To evaluate the effectiveness of LLMs in this context, we conducted a comprehensive study using the StackExchange platform in two distinct areas: language learning and programming. We developed a framework and a dataset for validating automatically generated personalized answers. Subsequently, we generated personalized answers using different strategies, including 0-shot, 1-shot, and few-shot scenarios. The generated answers were evaluated using three methods: 1. BERTScore, 2. LLM evaluation, and 3. human evaluation. Our findings indicated that providing LLMs with examples of desired answers (from the learner or similar learners) can significantly enhance the LLMs' ability to tailor responses to individual learners' needs.

  • Open Access

    Chemistry-Based Modeling on Phenotype-Based Drug-Induced Liver Injury Annotation: From Public to Proprietary Data

    by Mohammad Moein, Markus Heinonen, Natalie Mesens, Ronnie Chamanza, Chidozie Amuzie, Yvonne Will, Hugo Ceulemans, Samuel Kaski, Dorota Herman

    Drug-induced liverinjury (DILI) is an important safety concernand a major reason to remove a drug from the market. Advancementsin recent machine learning methods have led to a wide range of insilico models for DILI predictive methods based on molecule chemicalstructures (fingerprints). Existing publicly available DILI data setsused for model building are based on the interpretation of drug labelsor patient case reports, resulting in a typical binary clinical DILIannotation. We developed a novel phenotype-based annotation to processhepatotoxicity information extracted from repeated dose in vivo preclinicaltoxicology studies using INHAND annotation to provide a more informativeand reliable data set for machine learning algorithms. This work resultedin a data set of 430 unique compounds covering diverse liver pathologyfindings which were utilized to develop multiple DILI prediction modelstrained on the publicly available data (TG-GATEs) using the compound'sfingerprint. We demonstrate that the TG-GATEs compounds DILI labelscan be predicted well and how the differences between TG-GATEs andthe external test compounds (Johnson & Johnson) impact the modelgeneralization performance.


  • Open Access

    Quantification of the purinergic P2X7 receptor with [11C]SMW139 improves through correction for brain-penetrating radiometabolites

    by Joachim Brumberg, Richard Aarnio, Anton Forsberg, Päivi Marjamäki, Vera Kerstens, Mohammad M Moein, Sangram Nag, Saara Wahlroos, Michael Kassiou, Albert D Windhorst, et al.

    The membrane-based purinergic 7 receptor (P2X7R) is expressed on activated microglia and the target of the radioligand [11C]SMW139 for in vivo assessment of neuroinflammation. This study investigated the contribution of radiolabelled metabolites which potentially affect its quantification. Ex vivo high-performance liquid chromatography with a radio detector (radioHPLC) was used to evaluate the parent and radiometabolite fractions of [11C]SMW139 in the brain and plasma of eleven mice. Twelve healthy humans underwent 90-min [11C]SMW139 brain PET with arterial blood sampling and radiometabolite analysis. The volume of distribution was estimated by using one- and two- tissue compartment (TCM) modeling with single ( VT) and dual ( VTp) input functions. RadioHPLC showed three major groups of radiometabolite peaks with increasing concentrations in the plasma of all mice and humans. Two radiometabolite peaks were also visible in mice brain homogenates and therefore considered for dual input modeling in humans. 2TCM with single input function provided VT estimates with a wide range (0.10–10.74) and high coefficient of variation (COV: 159.9%), whereas dual input function model showed a narrow range of VTp estimates (0.04–0.24; COV: 33.3%). In conclusion, compartment modeling with correction for brain-penetrant radiometabolites improves the in vivo quantification of [11C]SMW139 binding to P2X7R in the human brain.