Context and rationale
Many Geography students focus on localised geographical problems for their independent dissertation research. While such investigations can yield novel insights, scaling up to analyse patterns across larger areas often reveals broader relationships. In geospatial analysis, where students often work with satellite-derived data, efficiently handling large datasets typically requires coding skills. Although many Geography programmes, including ours, offer introductions to basic programming, a technical skills gap remains; Jacobs et al. (2016) highlight that programming is increasingly essential for scientists, yet a lack of coding knowledge continues to limit the scope of undergraduate research in our field. Additionally, students often spend excessive time manually sourcing and compiling geospatial datasets, leaving less time for in-depth analysis that could enhance their academic outcomes.
To address these challenges, I am trialling the use of GenAI for streamlining data acquisition and processing. By providing a fast and reliable means of obtaining large-scale geospatial data, integrating GenAI in their workflow allows students to focus more on deeper geographical analysis, aligning with our assessment priorities. By enabling students to write prompts instead of complex code, GenAI lowers technical barriers, making large-scale geographical investigations more accessible. This shift particularly benefits students exploring broad spatial phenomena, such as mapping land cover change due to wildfires, or deforestation, who previously lacked the technical means to conduct such analyses effectively.
What we did
To illustrate this approach, I will describe my work with a student who chose to investigate deforestation in the Amazon, mostly driven by cattle farming. I introduced this to the student as follows:
- In an early meeting, I demonstrated how to effectively use ChatGPT for generating JavaScript code tailored for Google Earth Engine – a web browser-based platform for retrieving and analysing satellite data at scale. The code produced by ChatGPT can be entered directly into Google Earth Engine, which executes the code and generates relevant datasets, which can then be exported for further analysis. The student took notes, with a view to repeating the steps on their own. We focused on how to effectively ‘prompt engineer’ and then iterate with ChatGPT to arrive at code that a) works without error, and b) produces appropriate datasets.
- Subsequently, the student independently applied the techniques they had learned, using ChatGPT to generate code for Google Earth Engine. It is common that the first attempts don’t work and result in error messages from Google Earth Engine. Therefore, this process is inherently iterative, requiring the student to refine their prompts and engage in multiple interactions with ChatGPT to achieve the desired code output. In addition, students are expected to independently verify the accuracy of data obtained through Google Earth Engine, as they would in any project of this nature.
- This approach is more efficient than a manual data search, dramatically accelerating the data collection process – tasks that would take weeks of effort could now be completed in a fraction of the time, deepening the student’s grasp of ChatGPT’s capabilities, building their critical reflection and data literacy skills, and, ultimately, freeing up more time for in-depth data analysis and discussion.
- The student is still working on this dissertation. We have also discussed and considered how they best present their work. We decided that the methodology section of their dissertation needs to include an honest discussion of how they used GenAI tools in pursuit of their objectives, and that the annotated code generated by ChatGPT is included in the appendices. A critical reflection on the process of using GenAI in their research is also expected as part of the write-up.
I felt that this approach builds upon the student’s existing familiarity with GIS while introducing new tools that expand their analytical capabilities. The emphasis remains on geographical analysis rather than programming proficiency, though students must demonstrate understanding of how they obtained their data.
Useful tools
I have dabbled with several GenAI tools (e.g. ChatGPT, CoPilot, SciSpace). We decided to use ChatGPT as the student was a little more familiar with it, and at the time it proved to be the most capable for generating usable Google Earth Engine code. ChatGPT and CoPilot now seem to be equally capable for this task.
Benefits of new practice
- Increasing scope of possible dissertation projects – projectscan potentially be more ambitious and more impactful. Students can retrieve data over larger spatial scales, and over lengthier time periods, enabling more interesting analysis.
- Time saving – students can streamline the data collection process, potentially dedicating more time to meaningful analysis.
- Future-proofing – students develop (or build on) skills that we know will be valued by graduate employers; competency with GenAI is becoming an essential workplace skill.
- Critical evaluation – students learn to critique the relevance and quality of AI-generated output.
- More engaging and ambitious projects – this has an added benefit of fostering a culture of innovation, and collaboration between staff and students. We want students to push boundaries, and to challenge us!
- Professional Development – embedding GenAI into project-based learning and teaching helps staff to stay current with emergent technologies Lessons learned
Lessons learned
Thorough planning and road mapping are important when integrating GenAI tools into educational programmes or modules. Areas to consider and plan for include: how to introduce their use to students and colleagues; how to demonstrate and develop relevant skills; how to assess, and whether specific guidelines for staff and students are required regarding assessment and presentation of work.
Collaborating and discussing with the Programme Lead and fellow colleagues is invaluable to get buy-in and to consider everybody’s views. Such dialogue can help develop a robust implementation plan, alignment with programme aims and objectives as well as complementing learning outcomes.
It is important to introduce these tools in a way that enhances the existing curriculum. This might require staff to balance between innovation and core educational objectives and associated processes.
Future work or plans
There is potential to further expand the use of GenAI tools to support students during the analysis stage of their projects. GenAI could be employed to assist with troubleshooting and refining analytical processes, for example.
We currently have Faculty-level guidelines around the acceptable use of AI, and SoGEES staff are increasingly adding ‘AI statements’ to assignment briefs regarding the acceptable and unacceptable use of AI.
From next academic year (25/26) messaging to second- and third-year students working on developing and undertaking their dissertation research will address the use of AI head-on. The dissertation lead delivers a briefing to the second-year cohort toward the end of semester 1, and in semester 2 of their second-year students engage with a series of dissertation-related workshops – there is potential to deliver a workshop focusing specifically on leveraging GenAI. There is no expectation for students to engage with AI – it is an individual choice, and it is not appropriate or useful in many instances – but there is a need to build awareness that it can be used responsibly.
Reference:
Jacobs, C. T., Gorman, G. J., Rees, H. E., & Craig, L. E. (2016). Experiences With Efficient Methodologies for Teaching Computer Programming to Geoscientists. Journal of Geoscience Education, 64(3), 183–198. https://doi.org/10.5408/15-101.1
This case study has been captured by the Academic Development team and showcases current practice from BSc (Hons) Geography; BSc (Hons) Geography with Ocean Science; BA (Hons) Geography; BA (Hons) Geography with International Relations. We would like to thank Matt Westoby (School of Geography, Earth and Environmental Sciences) for participating in this case study.