Announcing the Winners of the CDCS Digital Research Prizes 2025
We are thrilled to announce the winners of this year’s CDCD Digital Research Prizes. Congratulations to all our winners and to all those whose fabulous work was nominated!
Event photography by Gintare Kulyte.
Best Small Data-Driven Project: Measuring the energy poverty premium in Great Britain and identifying its main drivers based on longitudinal household survey data
Fiona Rasanga
Collaborators: Tina Harrison and Raffaella Calabrese.
The study addresses a pressing socio-economic issue: the “energy poverty premium,” where poor households in Great Britain pay more for energy than their wealthier counterparts. Given the concerns on fuel poverty, rising cost of living and increasing energy price volatility, this research is not only timely but also provides policy relevant insights for energy policy and social welfare. The study’s objectives are twofold; first, to measure how much more poor households pay for energy, and second to identify they key determinants of these costs. The research questions are not only well-articulated but also grounded in existing literature on energy affordability and economic inequality. Additionally, by focusing on these research questions, we make significant contributions to existing literature on poverty premium.
Find out more about the study here: Measuring the energy poverty premium in Great Britain and identifying its main drivers based on longitudinal household survey data
Best Dataset: National Security and Defence Documents Dataset (1987-2024)
Andrew Neal
Collaborator: Roy Gardner
This dataset, launched on Edinburgh DataShare in August 2024, represents a transformative methodological innovation for critical security studies. Gathering over 3,600 views since its launch, this comprehensive resource provides unprecedented access to 573 national security and defence documents from 113 countries spanning 1987-2024, along with extensive metadata and computational tools designed for systematic analysis of security discourse.
This dataset introduces a methodological shift in a field that has been constrained by its reliance on case studies and theoretical development. While analysis of security threat construction in discourse has been central to this subdiscipline since its inception, it has lacked the tools and data for systematic comparative analysis.
Rather than simply adding to existing knowledge, this methodological innovation fundamentally changes what kinds of questions can be asked and answered within security studies. By enabling systematic comparison of how governments construct and communicate security threats, the dataset creates new possibilities for addressing core questions in the discipline, with implications that extend to policy development and international relations.
Find the dataset here: SUPERSEDED: National Security and Defence Documents Dataset (1987-2024)
Best Impact from a Data-Led Project: Equally Safe Online (ESO)
Poppy Gerrard-Abbott, Fiona McNeil and Ioannis Konstas
Collaborators: Christine Mcmello, Claire Houghto, Gavin Abercrombi, Tanvi Dinka, Simone Frend, Aiqi Jian and Nancie Gunson
Taking forward the Scottish Government’s Equally Safe strategy, and with support from Ofcom, Equally Safe Online (ESO) is an interdisciplinary, collaborative project where social science meets Machine Learning/NLP. Experts in safeguarding, violence against women and girls (VAWG), and digital education from SPS and the School of Education are working with computer scientists in the School of Informatics and Heriot-Watt (HW) to build a user bot for prevention, intervention, and support in online gender-based violence (OGBV), with particular attention on social media platforms.
Read more here: Equally Safe Online
Best Novel use of a Digital Method:
Do or Do Not: LLM Understandings of Collective vs Individual Action
Ari Stillman
Large Language Models (LLMs), such as GPT-4 and LLaMA, have revolutionized the field of NLP, particularly due to their generative capabilities. These models excel in tasks requiring flexible, creative, and interactive text generation. Unlike models such as BERT and RoBERTa, which are more suited to structured tasks like text classification and sentiment analysis, generative models are better equipped to handle more open-ended tasks and to produce coherent, contextually aware outputs. While LLMs like GPT-4 and LLaMA may not be fine-tuned for specific tasks like sentiment analysis, they are still capable of performing these tasks reasonably well through careful prompting.
One of the most significant advantages of LLMs is their ability to understand and generate more nuanced forms of communication, such as humor, irony, and sarcasm. This makes them potentially more effective than traditional NLP methods in analyzing online communities where such forms of expression are common. However, LLMs may not always surpass traditional models when it comes to specific tasks – especially those requiring precise labeling or domain-specific knowledge.
Best Data Visulation: Paths in Education
Paula Lago
Collaborator: Stuart King
In Uruguay, only half of young people complete mandatory secondary education. A country designs its education system to equip its population with the skills they will need to access decent employment and to exert their rights as citizens. But what happens when just half of a country’s young population manages to complete mandatory education —the very foundation designed to be accessible to all?
This has been a persistent challenge in Uruguay for at least 30 years. Despite being a beacon of democracy, human rights, and social progress in Latin America, this remains a structural issue that has yet to be addressed effectively.
While key statistics are frequently reported in mass media, a crucial gap remains. Who are these people? Why don’t they manage to finish? What barriers do they face in secondary education? What kinds of futures are possible in a country where half the population is uneducated?
This project aims to address this gap by using data visualisation to decode both quantitative and qualitative data, and elicit empathy. It combines statistical analysis with personal stories to expose the realities behind the numbers, making the issue clearer and more tangible.
Find out more here: Paths in Education
See the full list of winning and commended projects in the CDCS Digital Research Prizes 2025.