Learning Analytics (LA) is a field within Technology-Enhanced Learning (TEL) that borrows methods from Artificial Intelligence (AI) and helps understanding and optimizing learning environments based on the analysis of educational data. However, the adoption of LA suffers from an on-going debate about the need to provide a human centered and trustworthy approach to AI. In TEL this debate focuses on the tension between the potential of LA to achieve a more effective education and its impact on human behavior and wellbeing.

Meanwhile, the COVID-19 pandemic has created a large disruption in education. The crisis has illustrated the need for a profound transformation of education and its technological support, which also implies a transition to new modes of Hybrid Learning (HL). HL goes beyond more established ways of TEL, blurring traditionally separate dichotomies in teaching and learning: physical vs. digital spaces, onsite vs. online education, formal vs. informal learning, individual vs. social active learning methodologies. For instance, having a debate among students, some of which are in the classroom while others are at home, using videoconferencing tools and shared virtual whiteboards, is not an uncommon situation in the "new normal".

The blurring of dichotomies in HL poses significant challenges:

  • how to provide teachers with actionable learning indicators about activities happening in a hybrid of learning spaces and pedagogical approaches where not all students are expected to use the same tools, be located in the same (physical and/or digital) spaces, etc.;
  • how to help students regulate their individual and collaborative learning, interacting with teachers and/or students face-to-face or remotely, using a changing set of technological tools, etc., even in situations of potential social isolation. LA has the potential to address these limitations, but it needs to do so while also considering how its solutions impact the behavior of the human actors (teachers and learners, but also education managers, researchers...) and their wellbeing.

Therefore, the goal of the H2O Learning project is to build Trustworthy and Human-Centered LA (TaHCLA) solutions to support human stakeholders when designing, orchestrating and (self-, socially- or co-) regulating learning in HL. The contributions will consider key requirements for trustworthy AI, as defined by the European Commission:

  • fostering human (i.e., teachers, learners…) agency;
  • enabling transparency of the LA systems;
  • seeking socio-emotional and inclusive wellbeing; and
  • promoting accountability by enabling the assessment of algorithms and design processes. 

To do so, the project will propose:

  1. a set of principles, indicators, visualizations and interventions for TaHCLA systems that will enhance educator and learner agency in making informed decisions that improve design, orchestration and regulation of individual and social learning in HL;
  2. frameworks with conceptual and technical elements for transparent, wellbeing-driven, accountable HL systems; and,
  3. a set of pilot experiences in educational contexts promoting STE(A)M and Sustainable Development Goals. The project will:
  • early and continuously involve stakeholders in goal definition, co-design, co-orchestration, and evaluation processes;
  • explore multimodal data gathering and analysis techniques supporting the mixture of spaces and methods in HL;
  • address interoperability in the solutions proposed.