Time: 09:30 AM
This is a PhD Stage 1 seminar to present and discuss the proposed gap in the literature review, identified research questions, early results and plans for further investigations. All are welcome to attend, and to provide constructive feedback that will assist the student in preparing the future work for thesis submission.
Thesis Title: Participatory Causal Modelling for Learning Analytics
PhD Candidate: Ben Hicks
Supervisors: Kirsty Kitto and Simon Buckingham Shum
Delivery Mode/ Location: Hybrid. Join us in the CIC Ideation Studio (CB22.01.IdeationStudio) or click here to join via Zoom
Learning Analytics is not just about using data to understand learning and the environments in which it occurs. It also aims to improve the learning process into the future. This necessitates a causal understanding of learning processes; if I do this, how will it change the outcome?
To develop a causal understanding from learning data we have to understand the context in which the learning occurs, and apply this to a rigorous statistical framework. One way in which to do this involves the use of causal inference frameworks. However, making use of this relatively new modelling technique requires that data experts and context experts need to form a shared understanding about the system they are analysing. This is a difficult epistemic gap to bridge, as developing a shared understanding between technical and non-technical stakeholders is challenging, as is translating contextual knowledge into an appropriate statistical apparatus. So despite the importance of making causal claims in order to improve learning a traditional statistical approach is generally adopted in Learning Analytics. This explicitly rules out causal claims unless they are accompanied by a randomised controlled trial (RCT), but RCTs are particularly difficult to run in education due to both practical and ethical concerns.
Graphical Causal Models offer something new; a collaborative way to construct theories about the causal mechanisms of the world, that carries a direct interpretation for making causal claims from statistical associations. The visual formalism of the model requires little technical knowledge to engage with, whilst informing the statistical modeller about how to make stronger causal claims.
In my thesis I will apply these models to two cases; a student retention model and a conceptualisation of student belonging. The models will be co-constructed between stakeholders with a wide range of expertise in the modelling process, using the visual formalism to help foster a shared understanding of the system. This process of participatory modelling has the potential to allow a wide range of experts to be actively involved in the co-creation of models that support making causal claims. This stage 1 document presents my progress to date, and my plans for the remainder of my candidature.
Ben Hicks is a PhD student at the Connected Intelligence Centre, UTS and a Data Analyst with Charles Sturt University building Learning Analytics systems. Ben has worked with a wide variety of mathematical and statistical models, from water catchments and policy to student engagement and retention. He has taught mathematics for over a decade across three continents, occasionally well but always playfully. Ben’s research focuses on collaborative methods of modelling causality to help bring the abstract world of modellers and data closer to the world of teaching and learning practitioners.