Marija Kaplar

This paper presents SPARK_AI, a prompt-orchestrated system architecture for governing how large language models (LLMs) conduct structured and adaptive reasoning in human–AI interaction. The framework mitigates ad hoc LLM use by replacing direct answer generation with a process-oriented, step-by-step reasoning workflow. We focus on SPARK_AI_MATH, a domain module that supports learners in solving non-routine problem-solving tasks by operationalizing well-established problem-solving phases and guided questioning dialog strategies (Socratic-style prompts), with an optional tool-mediated visualization layer (e.g., GeoGebra). The module implements a five-phase conversational protocol consisting of problem interpretation, analysis of givens, planning, execution, and reflection, together with a controlled hint policy. This design is realized through a stateful system architecture in which each problem instance is maintained as an independent interaction track with a persistent reasoning state. User acceptance was evaluated by first-year mechanical engineering students (N = 108) using an expanded Technology Acceptance Model instrument, and the results were analyzed via PLS-SEM. The findings indicate overall favorable perceptions, with perceived usefulness and learning support emerging as key predictors of intention for continued use. Beyond this specific domain, the SPARK_AI framework enables efficient domain adaptation through localized prompt strategies while preserving a shared cognitive control layer for reasoning-centered human–LLM interaction.

Large language model-based educational chatbots are increasingly used to support learning, yet the ways in which their pedagogical behavior is specified remain insufficiently structured from a teacher-centered perspective. This paper presents a concept built on SPARK_AI (Socratic–Polya Adaptive Reasoning Kit), a stateful adaptive architecture for AI-supported learning, to enable teacher-driven authoring of educational chatbots. The main idea is to allow teachers to configure a chatbot through a structured specification form, without requiring prompt-writing or programming skills. Instead of manually composing prompts, teachers define key pedagogical dimensions such as guidance mode (step-by-step support, direct answer, or solution followed by explanation), instructional methodology, level of reflection, practical contextualization, and support escalation policy, including the number and type of hints provided before revealing a solution. These inputs are transformed into a structured pedagogical profile, which is then used to generate the chatbots behavioral configuration within the SPARK_AI system. In this way, the proposed approach translates teacher intent into consistent, reusable, and adaptable chatbot behavior. The paper positions this idea as a step toward more transparent, teacher-centered, and configurable AI-supported learning environments, in which domain experts can shape chatbot behavior in accordance with their own educational goals and teaching strategies.

In contemporary society, the ability to understand and interpret data has become a fundamental skill. Regardless of STEM or non-STEM fields, all individuals is expected the minimum level of stochastic literacy. This study examined Non-STEM students’ base-rate reasoning, their attitudes toward statistics, and the potential of artificial intelligence (AI) to support the analysis of students’ explanations. There participated 105 students, from the University of Novi Sad (non-STEM: Law, Economics) during regular classes. They solved a base-rate task, provided written explanations, and completed an attitude questionnaire. Responses were coded into six categories reflecting correctness and explanation type. Overall, 41% of students answered the task correctly, with no significant differences by faculty or gender. However, almost half of the participants chose the same wrong option. Students of the Faculty of Economics were more likely to provide explanations (70%) compared to law students (25%). When considering only responses with explanations, a significant gender difference was found, with male students more likely to provide correct probability based reasoning, while female students more often relied on equiprobability explanations. Attitude measures showed generally positive orientations although self-efficacy was weaker. AI-based (ChatGPT) coding of explanations gives results comparable to human classification and shows its potential for identifying misconceptions, albeit with certain limitations. Findings emphasize the need to strengthen probabilistic reasoning and statistical self-efficacy among Non STEM students, while also pointing to AI’s potential as a research and pedagogical tool.

The paper presents an empirical study that examines the academic performance in mathematics of fifteen-year-old students at the end of elementary school in Serbia. In conducting the analysis, the student results of the national test and their mathematics grades were utilized. The study covers a seven-year period 2013-19 and over 440 thousand students. Empirical findings affirm that girls exhibit superior math grades and higher achievements in national testing. However, the occurrence of Simpson?s paradox indicates a grading bias favoring girls. The results further highlight that girls exhibit a notably higher interest in attempting to solve open-ended tasks, even when unsuccessful, in contrast to boys, who leave those tasks unanswered during tests more frequently. The qualitative component of the research involved a focus group comprising teachers. It was conducted to gain insights into the teachers? perspectives and experiences regarding the gender grading gap.

The use of infographics for presenting data in the media and on the Internet has become a widespread phenomenon. This study examines how well undergraduates interpret the mean when presented in infographic, their attitudes towards the misuse of statistics, and their perceptions of their knowledge and use of statistics. The infographic was intended for the general public. The questionnaire was developed in a pilot study and then it was presented to a sample of 270 students from universities in Serbia and the USA. The study showed that STEM undergraduates and those who attended a course in statistics interpreted the infographic better. However, between 46% and 65% of those misinterpreted the infographic which indicates that changes are needed in statistics education to improve statistical literacy. Concerning demographic characteristics, somewhat better results were achieved by the American students, while gender had weak significance. Students whose interpretations were more successful perceived their knowledge and use of statistics as better. An extremely low number of students, 7.5%, disagreed that statistics is often misused, while only 14.2% trusted the research results presented in the media. The results from this study can be useful for curriculum developers, teachers, and researchers on statistical education. Keywords: infographics, mean interpretation, non-STEM students, statistical literacy, STEM students, students’ attitudes

Background: In the rapidly changing industrial environment and job market, engineering profession requires a vast body of skills, one of them being decision making under uncertainty. Knowing that misunderstanding of probability concepts can lead to wrong decisions, the main objective of this study is to investigate the presence of probability misconceptions among undergraduate students of electrical engineering. Five misconceptions were investigated:insensitivity to sample size,base rate neglected,misconception of chance,illusory correlation, andbiases in the evaluation of conjunctive and disjunctive events.The study was conducted with 587 students who attended bachelor schools of electrical engineering at two universities in Serbia. The presence of misconceptions was tested using multiple-choice tasks. This study also introduces a novel perspective, which is reflected in examination of the correlation between students’ explanations of given answers and their test scores.Results: The results of this study show that electrical engineering students are, susceptible to misconceptions in probability reasoning. Although future engineers from the sample population were most successful in avoiding misconceptions of chance, only 35% of examinees were able to provide a meaningful explanation. Analysis of students’ explanations, revealed that in many cases majority of students were prone to common misconceptions. Among the sample population, significant percentage of students were unable to justify their own answers even when they selected the correct option. The results also indicate that formal education in probability and statistics did not significantly influence the test score.Conclusions: Results of the present study indicate a need for further development of students’ deep understanding of probability concepts, as well as the need for the development of competencies that enable students to validate their answers. The study emphasizes the importance of answer explanations, since they allow us to discover whether students who mark the correct answer have some misconceptions or may be prone to some other kind of error. We found that the examinees who failed to explain their choices had much lower test scores than those who provided some explanation.

The purpose of this study is to analyse the effects of the Interactive Learning Materials Triangle (iLMT) on the learning and knowledge retention of 12-year-old students. The iLMT is a digital version of the standard school learning materials in Serbia, and is characterized by a high degree of interactivity and immediate feedback during the learning process. We conducted an experiment to explore whether iLMT influences student success in solving mathematical tasks that require different types of mathematical reasoning. Based on previous extensive research by Lithner, 4 types of tasks are discussed: high relatedness answer, high relatedness algorithm, local low relatedness, and global low relatedness. The study involved 633 students and 13 teachers of mathematics, equally distributed in control and test groups. The main findings indicate that student success on a knowledge test for high relatedness answer and local low relatedness tasks for the test group was significantly higher than for the control group. On the knowledge retention test, students in the test group outperformed students in the control group at high relatedness algorithm and local low relatedness tasks. Our results also suggest that, even when learning materials are carefully digitalized with the use of available technological advantages, student success in global low relatedness tasks may still be lacking.