Oral communication is a crucial skill in modern society. Nevertheless, it requires sustained practice and constructive feedback. Consequently, several studies have explored the development of oral communication trainers powered by Artificial Intelligence (AI). However, what characterizes expert speech remains unclear, especially given the need to adapt speech to contextual factors. In instructional environments, the speaker’s communication proficiency is a key determinant of audience learning outcomes. For this reason, we have analyzed 1250 speeches from five types of instructional discourses: in-person college classes ( Lectures ), online learning lessons ( Online Courses ), instructional animations ( Animated Lessons ), supplementary materials for school and high school ( Supplementary Lessons ), and public presentations ( Public Talks ). We extracted 16 speech metrics, including six additional multiple-participant metrics for Lectures . We obtained 250 videos of each discourse type, ensuring a minimum length of five minutes. Our analysis revealed expert values for each speech metric and showed how speech metrics vary across discourse types. We also developed an AI speech classifier that achieved an F1 score of 0.78. The model struggled to identify Online Courses , which is consistent with the Uniform Manifold Approximation and Projection analysis, showing that Online Courses are closely interjected with the speech of other instructional discourses. Furthermore, we identified distinct speech profiles in Lectures, Public Talks , and Online Courses , highlighting variations in speaking styles. This research provides valuable insights into expert speech in instructional discourses by offering reference values that can help speakers refine their delivery and support researchers in developing more effective speech training systems.

ABSTRACT The quality of education hinges on the proficiency and training of educators. Due to the importance of teacher training, the innovative platform Teacher Moments creates simulated classroom scenarios. In this scenario‐based learning, confusion is an important indicator to detect users who struggle with the simulations. Through Teacher Moments, we gathered 7975 audio recording responses from participants who self‐labelled their recordings according to whether they sounded confused. Our dataset stands out for its size, for not including actor‐generated audio, and for measuring confusion, a neglected emotion in artificial intelligence (AI). Our experiments tested unimodal approaches and feature‐level, model‐level and decision‐level fusion. Feature‐level fusion demonstrated superior performance to unimodal methods, achieving a balanced accuracy of 0.6607 on the test set. This outcome highlights the necessity for further investigation in the overlooked area of confusion detection, particularly employing realistic datasets like the one used in this study and exploring new methods. Beyond teacher training, the insights of this research also extend to other directions, such as other professionals making critical decisions, user interface design or adaptive learning systems.

ABSTRACT In recent years, serious games (SGs) have emerged as a powerful tool in education by combining pedagogy and entertainment, facilitating the acquisition of knowledge and skills in engaging environments. SGs enable the collection of valuable interaction data from students, allowing for the analysis of student performance, with artificial intelligence (AI) playing a key role in processing this data to make informed inferences about their knowledge and skills. However, the lack of explainability in AI models represents a significant challenge. This research aims to develop an interpretable model for predicting students' performance in real‐time while playing an SG by: (1) calculating the performance of an interpretable prediction model of task completion in an SG and (2) demonstrating the application of the interpretable model for just‐in‐time (JIT) classroom interventions. Our results show that we are able to predict students' task completion in real‐time with a balanced accuracy result of 77.21% after a short playtime has elapsed. In addition, an explainable artificial intelligence (XAI) approach has been applied to ensure the interpretability of the developed models. This approach supports personalised learning experiences, unlocks AI benefits for non‐technical users, and maintains transparency in education.

Serious Games (SGs) have gained attention as powerful educational tools because of their potential to provide reliable assessments and evaluate hard-to-measure constructs and competencies that are difficult to capture using traditional forms of assessment. Specifically, this study presents a human-centered approach to model and detect persistence—a key component of successful learning outcomes—in the context of SGs. With this purpose in mind, we developed a comprehensive rubric to characterize persistence behaviors in SGs. To design the rubric, we identified a set of persistence profiles and characteristics from previous literature and elaborated a general rubric for identifying persistence behaviors at the level of individual attempts. These characteristics were then mapped onto measurable features within Shadowspect, the SG used for data collection. Following this rubric, two annotators manually labeled 1,374 level attempts from 64 students using two visualization methods: in-game and text replays. With a comprehensive dataset of 2,748 labeled attempts, we trained and evaluated Machine Learning (ML) models for each type of replay to classify persistence behaviors across four categories: Persistence , Non-persistence , Unproductive persistence , and No behavior . Our results indicate that while text-based replays enable efficient annotation with promising performance, in-game replays may provide finer detail for certain complex behaviors, highlighting the strengths and limitations of each visualization method. This work contributes the use of SGs for assessment, illustrating a transparent and adaptable AI-driven approach that enhances reliability and user-centered insights, highlighting the complementary role of human input in optimizing AI-based models to achieve meaningful, user-centered assessments in education.

As cyber threats continue to evolve, the need for highly skilled cyber defence operators becomes increasingly critical. In this work, we aim to provide a multidisciplinary exploration into the current educational landscape, focusing on the following pivotal areas: cyber defence educational initiatives, digital skills, technological enablers, and ethical considerations. First, we present the current landscape of cyber defence educational initiatives. Then, we examine the required digital skills, virtual reality and augmented reality initiatives for immersive learning experiences and delve into the advantages of game-based learning for skill acquisition in order to finally provide a holistic evaluation of the complexities involved in cyber defence training. We underscore the importance of standardising training modules tailored to diverse roles within cyber defence. The discussion emphasises the need for ethical and legal guidelines, especially concerning privacy and bias in AI-driven educational tools. Finally, we highlight the importance of dynamic curricula that include technical, legal, and soft skills, along with hands-on training through simulations to prepare operators for real-world cyber threats. This work will serve as a foundation for academics, industry professionals, and policy-makers interested in elevating the standards and effectiveness of cyber defence training suggesting ideas for more specialised, adaptable, and ethically responsible programs.

When using simulations to design and implement novice teacher practice, a teacher educator may be concerned about if what is technically possible in terms of generating feedback to novice teachers' responses is educationally purposeful to support their learning. This paper details the design of infrastructure to incorporate user feedback within the Teacher Moments platform that is generated by an AI agent, and how we designed to sustain and scale the expertise of mathematics teacher educators when training a large language model. To best support the learning of novice mathematics teacher users to enact ambitious and equitable mathematics teaching, this paper explains the research design of training a large language model by collaborating with mathematics teacher educators to edit or endorse generated feedback across multiple training cycles. This paper also describes the UI design to explore potential of hosting such processes all within the Teacher Moments platform.

Managing stress is a crucial soft skill that affects cognitive performance and health. Stress detection through biometrics can be used to improve and evaluate stress management. However, measuring the effects of stress on biometrics and executive functions is difficult and dependent on the individual. Despite these challenges, this paper presents a case study that collects a comprehensive multimodal dataset with two stress metrics, four biometric signals, and twenty-two executive function metrics from Game-based Assessment (GBA) trace data specifically designed for this purpose. The experiments suggest that biometrics, especially the heart rate and skin temperature, are effective predictors of stress. Additionally, noteworthy correlations were observed between heart rate and certain executive function variables. The levels of GBA that measured shifting and processing speed showed a higher heart rate than the response inhibition levels. This case study, together with the developed stress detectors, enables the detection of persons who struggle to manage stress and measure their executive function performance under stressful situations.