Multimodal learning analytics provides researchers new tools and techniques to capture different types of data from complex learning activities in dynamic learning environments. This paper investigates the use of diverse sensors, including computer vision, user‐generated content, and data from the learning objects (physical computing components), to record high‐fidelity synchronised multimodal recordings of small groups of learners interacting. We processed and extracted different aspects of the students' interactions to answer the following question: Which features of student group work are good predictors of team success in open‐ended tasks with physical computing? To answer this question, we have explored different supervised machine learning approaches (traditional and deep learning techniques) to analyse the data coming from multiple sources. The results illustrate that state‐of‐the‐art computational techniques can be used to generate insights into the "black box" of learning in students' project‐based activities. The features identified from the analysis show that distance between learners' hands and faces is a strong predictor of students' artefact quality, which can indicate the value of student collaboration. Our research shows that new and promising approaches such as neural networks, and more traditional regression approaches can both be used to classify multimodal learning analytics data, and both have advantages and disadvantages depending on the research questions and contexts being investigated. The work presented here is a significant contribution towards developing techniques to automatically identify the key aspects of students success in project‐based learning environments, and to ultimately help teachers provide appropriate and timely support to students in these fundamental aspects.

Interdisciplinary research from the learning sciences has helped us understand a great deal about the way that humans learn, and as a result we now have an improved understanding about how best to teach and train people. This same body of research must now be used to better inform the development of Artificial Intelligence (AI) technologies for use in education and training. In this paper, we use three case studies to illustrate how learning sciences research can inform the judicious analysis, of rich, varied and multimodal data, so that it can be used to help us scaffold students and support teachers. Based on this increased understanding of how best to inform the analysis of data through the application of learning sciences research, we are better placed to design AI algorithms that can analyse rich educational data at speed. Such AI algorithms and technology can then help us to leverage faster, more nuanced and individualised scaffolding for learners. However, most commercial AI developers know little about learning sciences research, indeed they often know little about learning or teaching. We therefore argue that in order to ensure that AI technologies for use in education and training embody such judicious analysis and learn in a learning sciences informed manner, we must develop inter‐stakeholder partnerships between AI developers, educators and researchers. Here, we exemplify our approach to such partnerships through the EDUCATE Educational Technology (EdTech) programme. What is already known about this topic? The progress of AI Technology and learning analytics lags behind the adoption of these approaches and technologies in other fields such as medicine or finance. Data are central to the empirical work conducted in the learning sciences and to the development of machine learning Artificial Intelligence (AI). Education is full of doubts about the value that any technology can bring to the teaching and learning process. What this paper adds? We argue that the learning sciences have an important role to play in the design of educational AI, through their provision of theories that can be operationalised and advanced. Through case studies, we illustrate that the analysis of data appropriately informed by interdisciplinary learning sciences research can be used to power AI educational technology. We provide a framework for inter‐stakeholder, interdisciplinary partnerships that can help educators better understand AI, and AI developers better understand education. Implications for practice and/or policy? AI is here to stay and that it will have an increasing impact on the design of technology for use in education and training. Data, which is the power behind machine learning AI, can enable analysis that can vastly increase our understanding of when and how the teaching and learning process is progressing positively. Inter‐stakeholder, interdisciplinary partnerships must be used to make sure that AI provides some of the educational benefits its application in other areas promise us.

While Artificial Intelligence in Education (AIED) research has at its core the desire to support student learning, experience from other AI domains suggest that such ethical intentions are not by themselves sufficient. There is also the need to consider explicitly issues such as fairness, accountability, transparency, bias, autonomy, agency, and inclusion. At a more general level, there is also a need to differentiate between doing ethical things and doing things ethically, to understand and to make pedagogical choices that are ethical, and to account for the ever-present possibility of unintended consequences. However, addressing these and related questions is far from trivial. As a first step towards addressing this critical gap, we invited 60 of the AIED community’s leading researchers to respond to a survey of questions about ethics and the application of AI in educational contexts. In this paper, we first introduce issues around the ethics of AI in education. Next, we summarise the contributions of the 17 respondents, and discuss the complex issues that they raised. Specific outcomes include the recognition that most AIED researchers are not trained to tackle the emerging ethical questions. A well-designed framework for engaging with ethics of AIED that combined a multidisciplinary approach and a set of robust guidelines seems vital in this context.

Collaborative problem-solving (CPS) is a fundamental skill for success in modern societies, and part of many common constructivist teaching approaches. However, its effective implementation and evaluation in both digital and physical learning environments are challenging for educators. This paper presents an original method for identifying differences in CPS behaviours when groups of students are taking part in face-to-face practice-based learning (PBL). The dataset is based on high school and university students' hand position and head direction data, which can potentially be automated deploying existing multimodal learning analytics systems. The framework uses Nonverbal Indexes of Students' Physical Interactivity (NISPI) to interpret the key parameters of students' CPS competence. The results show that the NISPI framework can be used to judge students' CPS competence levels accurately based on their non-verbal behaviour data. The findings have significant implications for design, research and development of educational technology.

The question: “What is an appropriate role for AI?” is the subject of much discussion and interest. Arguments about whether AI should be a human replacing technology or a human assisting technology frequently take centre stage. Education is no exception when it comes to questions about the role that AI should play, and as with many other professional areas, the exact role of AI in education is not easy to predict. Here, we argue that one potential role for AI in education is to provide opportunities for human intelligence augmentation, with AI supporting us in decision‐making processes, rather than replacing us through automation. To provide empirical evidence to support our argument, we present a case study in the context of debate tutoring, in which we use prediction and classification models to increase the transparency of the intuitive decision‐making processes of expert tutors for advanced reflections and feedback. Furthermore, we compare the accuracy of unimodal and multimodal classification models of expert human tutors' decisions about the social and emotional aspects of tutoring while evaluating trainees. Our results show that multimodal data leads to more accurate classification models in the context we studied.