Unlocking the Power of Data Democratisation with ChatGPT in Higher Education: Reflections from the University of Law
In the ever-evolving landscape of higher education, the challenge of harnessing data for actionable insights remains paramount. As the Director of Learning Analytics at the University of Law, a key focus of mine has been bridging the gap between the extensive data available and academic teams' practical use of this data.
The traditional methods of delivering insights through dashboards and complex data manipulation have shown their limits in terms of scalability and adaptability for particular use cases. For instance, providing granular data on an individual student’s access pattern to a few specific items across multiple teaching modules, which change weekly to reflect progress through the teaching sessions, is difficult given the number of academics, modules and items. This reality propelled us to seek an innovative solution that aligns with the principle of data democratisation at the University of Law..
Why is data democratisation essential?
Data democratisation aims to make data accessible for all, turning it from a specialist's domain into a universally accessible tool. This approach is grounded in cultivating a data-friendly culture, providing the necessary tools and training, and empowering self-service analysis. ChatGPT, a generative AI tool, emerges as a potential game-changer, lowering the entry barriers to data analysis and enabling a more inclusive environment for data interaction. Therefore, data democratisation will result in Universities being able to improve the timeliness of identification and speed of intervention which will result in an improved learner experience.
What did we find?
Our pilot projects at the University of Law have explored the integration of ChatGPT within our data analysis processes, seeking to democratise data access and analysis across academic teams. While the journey has been challenging, ranging from initial staff apprehension to the technical nuances of prompt engineering, the results are promising. Our pilot projects have supported academic teams using Generative AI (ChatGPT) to gain actionable insights into their virtual learning environment data (Blackboard). The use cases include:
Use Case #1: Weekly checks on a small number of items
Each week, the Module Leader identifies one or two items in their Blackboard module that are important indicators of successful learning. The Module Leader wishes to determine which students have accessed these items. The Module Leader will then contact students who have not accessed the material to encourage them to engage with it to ensure good outcomes. They also wish to identify students who have accessed the items so they can email a supportive message about their good engagement behaviour.
This use case is challenging to implement through a pre-created dashboard as the items will differ across modules, the module leader must combine the data across several virtual learning environment modules, and the naming convention will not be consistent. Therefore, providing the Module Leader the opportunity to draw insights using an innovative Generative AI tool is a game-changer.
To date, the Module Leader's feedback has been very positive. Firstly, they can complete the process in ChatGPT, which takes less than 5 minutes to produce lists of students to contact. Secondly, the outcomes of the email interventions have been positive as students visit the materials they have missed.
Use Case #2: Calculating an engagement level for all students at a set point in the teaching calendar
The academic team wishes to take a snapshot of their engagement in Week 3 of teaching. The engagement calculation is derived from a student's access to 15 selected items, combined with their scores on five Blackboard multiple-choice tests. The outcomes should be as a percentage and mapped to a RAG score to indicate who to contact and the message type. For instance, a calculated engagement score of 80-100% is 'Green with a positive message of successful behaviours', 50-79% is 'Amber', and less than 49% is 'Red with a checking in are you OK message'. In addition, although the academic team sends these messages, they also need to include the student's Academic Coach (Personal Tutor) for visibility.
This use case can be delivered through a "one-off" report created through Excel. However, combining calculations and secondary data sets, including Academic Coach lists, requires significant Excel skills. Therefore, the pilot was to identify if someone could achieve this by using the lower barrier to entry Generative AI tools.
What lessons are transferable to other institutions?
Several lessons from our pilot projects emerge for institutions looking to embark on a similar path:
1. The importance of starting small and iteratively refining your approach cannot be overstated. The journey of integrating generative AI into data processes is as much about cultural change as it is about technological adoption.
2. Fostering a community of practice among staff to share insights and challenges can significantly accelerate learning and adoption.
3. Ethical considerations, particularly those related to data privacy and analysis accuracy, must guide the deployment of these technologies.
Overall, institutions should include generative AI data analysis tools within their data ecosystem.
As we continue to explore the possibilities of ChatGPT and generative AI in data democratisation, the vision is clear: "to empower academic teams to identify meaningful insights and make effective interventions low barrier Generative AI tools have a crucial role to play. While challenges remain, the potential for transformative impact on decision-making and student support is immense".
The journey at the University of Law is ongoing, and the final destination might not include ChatGPT. However, Generative AI tools should be an active component of a university's data analysis ecosystem. The path forward is promising, heralding a new era of data empowerment in higher education by including Generative AI for data analysis.
