Human learning and learning analytics in the age of artificial intelligence

In the last few decades, we have faced rapid technological development, unlike anything ever seen before. It seems that with this exponential growth of possibilities, we are facing more questions than answers about civilisation's future. As education and knowledge are cornerstones of this process, this special section highlights some aspects of this topic and puts them into perspective. In pedagogical practice, AI opens many important issues about teachers' roles and competences and students' roles as autonomous, self-directed learners. Of special interest is how teaching and learning interact to provide individualised, personalised learning experiences and connections to knowledge networks. AI and analytics support teachers' informed decisions to facilitate student-centred learning (Starčič & Vukan, 2019). Learning takes place across three domains: the cognitive, social-affective and psychomotor (Bruner, 1966). Traditional teaching follows learning taxonomies, which are more domain-specific and outcome-related. Current AI adaptive learning environments support process-oriented teaching. AI can capture the learning process and the learner's psychological states at every step of the process, providing relevant feedback concerning the cognitive and psychomotor process, social interaction and affectional states (Sullivan & Keith, 2019). A multimodal analytics approach has the potential to provide understanding based on physiological, representational and behavioural data (Blikstein & Worsley, 2016) using gesture and physical actions (Cukurova, Luckin, Millán, & Mavrikis, 2018). Designing and developing AI requires a deep understanding of human learning and designing. For example, social robots help to learn about human learning (Mubin, Stevens, Shahid, Mahmud, & Dong, 2013). AI is a powerful tool for exploring the human learning process and making it more visible (Luckin, Holmes, Griffiths, & Forcier, 2016). This help comes in two ways: in the design phase when learning science informs AI design, and during the learning process where AI is a powerful tool for gathering and analysing data about learner (Du Boulay, 2019; Luckin & Cukurova, 2019; Mavrikis, Geraniou, Gutiérrez-Santos, & Poulovassilis, 2019). In considering what we might learn from developing AI by exploring the human learning process, among the potential future issues might be the transformation in the human learning process. AI learns and improves from gathered data, which may eventually effect a change in the concept of human learning. It is anticipated that the domains of learning will be intertwined and the formal curriculum will not provide the basic framework. Student-driven learning in diverse settings will be supported by complex data about learner and learning context. In the future, how will the teacher's role be defined and which teachers' competences will be needed? The special section brings together five articles discussing current applications of AI as it affects pedagogical practices. The papers present empirical studies, one in elementary school, one in elementary and high school and three in higher education. They contribute to the current understanding of AI in educational settings by considering teachers' support to students as autonomous and self-directed learners. The teaching strategies under consideration are feedback and assessment, tutors' decision making and teaching strategies that are informed by learning analytics, eg, predictive, representational by discourse analysis and multimodal. The article by Maria Cutumisu, Doris B. Chin and Daniel L. Schwartz presents the digital game-based assessment of middle-school and college students. In focus of attention is the link between critical feedback and learning outcomes. Explored are students' strategies when they are learning on their own. The students' proactive behaviour in picking critical feedback and revising their work leads them to higher achievement comparable to students who conduct their assignments without seeking critical feedback. Authors findings indicate students' search and engagement in critical feedback as a predictor for learning and a valid outcome measure. The article by Kshitij Sharma, Zacharoula Papamitsiou and Michail Giannakos presents a potential use of AI and multimodal analytics to explain learning in adaptive self-asssessment. Authors discuss the role of effective feedback and learning analytics on students' engagement in adaptive learning activities and their achievement. They examine multimodal data such as that fusion of eye-tracking, facial expressions and arousal data as a prediction of effort and performance. The article by Mutlu Cukurova, Carmel Kent and Rosemary Luckin presents AI and multimodal data in the service of human decision-making on a case study in debate tutoring. Authors discuss the role of AI in human intelligence augmentation with AI supporting decision-making processes. Examined are prediction and classification models to increase the transparency of the intuitive decision-making processes of expert tutors. Findings indicate that tutors' decisions for reflections and feedback, which influenced by the social and emotional aspects, could be supported by explicit and traceable models. The article by Florence R. Sullivan and P. Kevin Keit presents the potential of natural language processing to support microgenetic analysis of collaborative learning discussions. This learning analytics discussion is in the area of machine learning techniques and neural network models by computational linguistics. Authors develop and critically discuss a theoretical model of a computational method for assisting in the microgenetic analysis of speech data in the co-present, collaborative problem-solving group talk. The research was conducted in educational robotics analysing video and audio-taped classroom and informal discussions. The findings could inform the curriculum and learning environment designers. Lastly, the article by Christothea Herodotou, Martin, Hlosta, Avinash, Boroowa, Bart Rienties, Zdenek, Zdrahal and Chrysoula Mangafa presents online teaching using predictive learning analytics. This large scale study at the Open University found that teacher's use of predictive learning analytics was related to learning outcomes. Teachers were using learning analytics for identifying students at risk and preventing them from failing. Having had the opportunity in the past to serve as BJET editor for Europe and currently being a member of the international advisory board, I am honoured to take part in this anniversary edition and hope that we contribute some insights about this dynamic development, AI in education. The special section discuss and calls for further research of efficient learner's support in self-assessment and autonomous decision making, including (a) Measuring the learning process and providing advanced feedback and scaffolds capturing everyday learning behaviour in diverse settings, and (b) gathering multimodal data in authentic learning contexts using ubiquitous low-cost technology.