PhD Studentship: The Financial Judgement of AI: New Insights into Political Ideology and Role-Conditioned Bias

Project summary:  Project examines whether Large Language Models (LLMs) embed systematic political ideologies that influence their financial judgements. Extends a novel auditing framework to measure ideological orientation. Investigates how these ideological tendencies change under financial analysts or financial advisor personas, and how they affect stock forecasts, investment recommendations, and ESG portfolio advice.

Start date: 1st October 2026

Deadline: 30th April

Duration: 36 months

Funding: Funded

Funding towards:

Home fee

Stipend -UKRI stipend rate for UK students.

Funding details: Fully-funded scholarship for 3 years covers all university tuition fees (at UK level) and an annual tax-free stipend. International students are also eligible to apply, but they will need to find other funding sources to cover the difference between the home and international tuition fees. Exceptional international candidates may be provided funding for this difference. 

Number of places: 1

Number of places extra: There will be a shortlisting and interview process.

RCUK eligibility: No

Eligibility: 

Applicants should hold, or expect to obtain, a postgraduate degree in Computer Science, Artificial Intelligence, Data Science, Machine Learning, Finance, FinTech, Economics, or a related field. Candidates should demonstrate knowledge of Large Language Models, generative AI, and machine learning, with interest in financial applications. Experience in natural language processing, model evaluation, or experimental design is advantageous. Proficiency in Python and/or R and familiarity with AI/ML libraries or generative AI platforms is desirable.

Further eligibility criteria can be found: https://www.strath.ac.uk/studywithus/postgraduateresearchphdopportunities/business/accountingfinance/thefinancialjudgementofainewinsightsintopoliticalideologyandrole-conditionedbias/  

Study modes eligibility: Full-time

Fee Status: What fee status applies to applicants? Please omit any of the below that do not apply:

Project Details:  Large Language Models (LLMs) are increasingly used to support financial analysis and investment advice. Emerging research suggests that LLMs may embed hidden political ideologies that influence how they analyse information and make recommendations.

This project investigates whether LLMs exhibit systematic political ideologies and whether these orientations affect the financial judgements they provide. Because LLMs are typically designed to avoid revealing political views through direct questioning, the project extends an existing auditing framework to measure political ideology as revealed through financially grounded analytical tasks. A key innovation is the development of a new survey instrument designed to uncover political differences by examining how AI systems approach real-world financial decision problems. This approach establishes a benchmark measure of political ideology across different LLMs.

The project further explores how these ideological tendencies change when LLMs operate under professional personas commonly used in financial services. By prompting models to adopt the role of a financial analyst or financial advisor, the research examines whether persona-specific ideological orientations influence stock forecasts, investment recommendations, and ESG investment advice, including portfolio construction choices and their underlying justifications. This allows for systematic comparison between baseline ideological tendencies and those expressed within professional financial decision-making contexts.

Primary Supervisor: Professor Mark Cummins

Additional Supervisor/s: Dr. Yashar Moshfeghi; Dr. James Bowden