AI could reshape political decision-making, but at what cost?

Bureaucratic theory, shaped by Max Weber, stresses order, hierarchy, and administrative expertise. The entry of AI into governance raises pressing questions about whether algorithmic decision-making would strengthen bureaucratic efficiency or disrupt it by shifting authority away from human experts and into the hands of opaque machine systems. The study highlights the danger of bureaucracy being replaced or overshadowed by AI-driven structures that lack accountability or transparency.

Artificial intelligence is moving rapidly into the political domain, raising new questions about ethics, fairness, and governance. A recent study published in Philosophies examines how large language models (LLMs) and other AI systems could influence the future of politics.

The authors of the study "The Use of Artificial Intelligence in Political Decision-Making" assess the promises and perils of AI through the lens of political philosophy, bureaucratic theory, and conflict analysis, while also testing how contemporary AI tools respond to political decision-making tasks.

How AI challenges traditional theories of politics

The study frames the debate around three major schools of political thought, realist politics, bureaucratic theory, and conflict theory, and explores how AI aligns with or disrupts each. Realist politics, rooted in the Machiavellian tradition, emphasizes the preservation of power at all costs. According to the authors, this framework resonates strongly with AI’s logic, which often prioritizes efficiency and calculative reasoning over moral considerations. In this sense, AI risks reinforcing a political culture where outcomes are judged by pragmatism rather than principles.

Bureaucratic theory, shaped by Max Weber, stresses order, hierarchy, and administrative expertise. The entry of AI into governance raises pressing questions about whether algorithmic decision-making would strengthen bureaucratic efficiency or disrupt it by shifting authority away from human experts and into the hands of opaque machine systems. The study highlights the danger of bureaucracy being replaced or overshadowed by AI-driven structures that lack accountability or transparency.

Conflict theory, closely linked to Marxist analysis, examines power imbalances and systemic inequalities. Here, the risk lies in AI amplifying existing disparities. Because large language models are trained on vast datasets that may include cultural, social, or political biases, the technology could perpetuate discrimination based on race, gender, class, or geography. The authors stress that without deliberate safeguards, AI in politics could entrench inequality under the guise of neutrality.

Where AI is already reshaping political processes

While the concept of AI in politics often evokes futuristic scenarios, the study points to current real-world applications already shaping governance. Governments and public institutions worldwide are adopting AI in areas ranging from fraud detection in welfare systems to predicting criminal activity, monitoring disease outbreaks, and enhancing emergency response. In citizen engagement, AI-driven chatbots and sentiment analysis tools are increasingly being used to gauge public opinion and deliver services.

These applications demonstrate efficiency and cost-effectiveness, but they also reveal deep ethical trade-offs. Fraud detection algorithms, for example, have helped reduce misuse of public funds but have also wrongly flagged vulnerable individuals, creating hardships. Predictive policing tools, designed to optimize security, have been criticized for embedding racial and social biases. Even in healthcare, disease surveillance through AI can improve response times but may raise privacy concerns.

The authors argue that these examples serve as a warning for political decision-making at large. The efficiency of AI cannot be divorced from the ethical risks it introduces. When applied to governance, the same biases and limitations could skew political decisions, marginalize communities, and weaken trust in institutions.

What AI models reveal about political values

To explore how AI systems might handle political decision-making, the authors conducted direct tests with ChatGPT and Claude, two leading large language models. They asked these tools to design political decision-making frameworks. Both models leaned heavily toward pragmatic, cost-benefit approaches, prioritizing efficiency and measurable outcomes. Ethical concerns such as inclusion, equality, and fairness only emerged when explicitly prompted.

This experiment illustrates a critical flaw in the assumption that AI is neutral. The models’ responses reflected not only their training data but also embedded cultural assumptions. The study notes, for instance, that Claude demonstrated a distinctly U.S.-centric worldview, framing political scenarios through American cultural and institutional lenses. Such biases raise serious concerns about the global application of AI in politics, where context-sensitive decision-making is crucial.

To sum up, AI, while powerful in speed and computational ability, risks narrowing political decision-making to purely quantitative terms. This dynamic could sideline ethical reflection, reduce space for public deliberation, and reinforce existing power structures rather than challenge them.