King's College London

The project is co-created with a PhD student with master’s students as participants and evaluators. 

The Problem – Facilitating self-directed learning around what characterises a good research question beyond the tutor saying ‘yeah – it sounds interesting’. 

On the MSc in Global Health and Master of Public Health, full time students completing their degree in a year are expected to undertake a research project as part of a 60-credit dissertation module in parallel with taught modules delivered in semester 1 and 2.   The project is therefore predominantly self-directed with guidance and support being provided by the supervisor.  However, the appointment of a supervisor is dependent on the student identifying a valid area of research.  A key first and pivotal step is to identify a relevant, researchable question which is grounded in an appraisal of the existing evidence base and reflects a gap. This is a time-consuming complex task surmised in a frequently posed question by students – is this a good research question – with a belief that their tutor can give them a definitive answer.  Although students are signposted to multiple resources to help develop their own self-evaluation about what is a good research question, these resources lack the interactivity students value through posing the question to their tutor(s).  However, the complexity of the answer is not always amenable to the tutorial time available and may inadvertently minimises the need for students undertake the background reading and thinking. The failure to understand the fundamental importance of a good research question will then often be reflected in poor study design and findings. 

The Solution 

In healthcare, Artificial Intelligence-based (AI) decision support systems are increasingly used to support systematic, methodical, and robust decision making about patient care and treatment pathways.  

The aim of this project is to deploy a large language model (LLM) such as ChatGPT to effectively create a decision support system that encompasses key prompts and responses that facilitates: 

• Students understanding about what the key characteristics of a good research question are and how those maps onto/or don’t map onto the responses being provided, 
• Formative feedback enabling critical reflection and engagement with the task.  
• Signposting to wider resources to inform their decision making. 
• A record of how their understanding develops and progression to a more clearly defined research question/and or problem. 

The developed tool will be piloted with current masters students in the Department and subsequent evaluation and feedback eliciting the degree to which it has enabled a more robust, systematic and methodical decision-making resulting in a ‘good research question’. 

We would propose that this project addresses all three categories: 

• Pedagogy focus with an objective to integrate AI into teaching and feedback mechanisms. 
• AI Tech focus with an objective to develop and evaluate an AI system to facilitate decision making. 
• Resource creation with the objective that the tool created is available to all masters students undertaking dissertation projects with prompts and response being modified to reflect the specifics of a particular programme/and or research design. 

Please note Gregory Kell is also a collaborator on this project. Gregory Kell — King's College London (kcl.ac.uk)