Coalition Governance Bargaining
In coalition governance, understanding how political parties negotiate, bargain, and deliberate is central to explaining policy outcomes and evaluating democratic representation. Yet, this often remains difficult to observe. The research developed as part of my PhD seeks to shed light on these hidden dynamics by developing computational tools that can infer patterns of intra-coalitional behaviour.
Specifically, I investigated how coalition parties signal distinct rhetoric in public parliamentary speech, a phenomenon known as coalition differentiation. Using a supervised text classification approach I captured distinct linguistic patterns by coalition parties in speeches. The paper presenting this measurement approach is currently under review, and in the meantime can be found here.
A second paper applies this measure to show that parties predominantly differentiate when they possess sufficient bargaining leverage. Indicating the connection between this behaviour and incumbency negotiations. This work is also under review, the working paper can be read here.
Simulating Deliberations
My research at CAIG expands on my PhD work by investigating how opaque deliberative processes can be directly simulated using computational methods. Many political negotiations and decision-making processes are difficult or impossible to observe directly, and I aim to reconstruct these interactions using large language model (LLM) powered multi-agent systems (MAS). This methodology allows the creation of “digital twins” of deliberative bodies to run counterfactual scenario simulations.
Considering the novel nature of this method, my research does not only aim to develop the pipeline required to simulate political deliberation with LLM-driven agents, but also to create robust validation frameworks for every stage, whether that is assessing the utility of synthetically generated texts, or the effect of agentic design, such as prompt engineering or agent orchestration, on simulated negotiation dynamics.
AI Mediated Political Communication
As citizens increasingly contact elected representatives online, officials face growing volumes of digital communication. This raising important questions about how AI and digital platforms are transforming the role of elected representatives.
I am investigating this specifically on the German transparency platform abgeordnetenwatch.de, which allows citizens to publicly submit questions to any elected representative. By scraping the platform, my co-authors and I compiled a dataset of more than 250,000 citizen questions and nearly 200,000 corresponding answers. A current working paper uses this data to analyse the role of polarising and emotional language. Using a causal identification strategy, we find that MPs on ideological fringes strategically introduce polarising language in their replies, reducing the deliberative quality of online discussion.
Public Parliamentary Engagement
As a research assistant, I contributed to the Inter-Parliamentary Union’s Global Parliamentary Report on public parliamentary engagement, published in 2022. My work focused on case studies of youth engagement and citizen assemblies, examining how parliaments interact with citizens and how these processes influence accountability, representation, and democratic legitimacy.
Building on this project, I authored an academic article on feedback processes in parliamentary public engagement, drawing on data from focus groups, parliamentary surveys, and interviews conducted for the report. The article, published in The Journal of Legislative Studies , examines how parliaments can use structured feedback loops to improve the quality of engagement, increase citizen participation, and fine-tune institutional practices.
Christine Sheldon 