About the Project

Supervisory team: Dr Bahar Rastegari and Dr Panayiotis Danassis

Agentic AI is rapidly affecting diverse aspects of our life, including social, technological and financial interactions. They may seem like human agents, yet their "thinking" and "reasoning" is not the same. The aim of this project is to study, understand and design the strategic interaction between LLM agents.

As Agentic AI is rapidly moving from research labs into the real world, we expect it to affect many aspects of our social, technological, and financial world. A large number of interactions will be mediated via AI systems giving rise to complex multi-human-multi-AI systems.

Traditionally, multi-agent interactions are studied through the lenses of Game Theory -- which aims to study the effects of self-interested behaviour, -- and mechanism design -- which aims to align the personal with the collective incentives. Classical game theory offers powerful tools for analysing strategic behaviour, yet assumes fully rational, utility-maximising agents operating in well-defined games. In contrast, large language model (LLM) agents reason via learned heuristics, finite context windows, prompts, and training data rather than explicit utility optimisation. As a result, classical notions of negotiation, belief formation, and best-response strategies break down when applied to LLM agents.

The aim of this project is to study the emergent strategies and stable points as we shift from rational agents to foundation-model-based agents, opening the path to more robust, strategic, and socially aligned Agentic AI.

The scope of this project is wide by design, with details to be refined collaboratively with the prospective PhD applicant. Potential applications include market negotiations, collaborative resource allocation, and AI governance simulations. For example, consider the case of Autonomous Supply Chain Negotiation. Multiple LLM agents, each representing different companies or suppliers, could negotiate delivery schedules, prices, and production priorities. These agents can learn and update their beliefs, strategically adjust offers, form coalitions, and negotiate.

Entry requirements:

You must have a UK 2:1 honours degree, or its international equivalent in one of the following or related disciplines:

• computer science
• Artificial Intelligence
• mathematics

We are looking for a highly motivated candidate with a strong academic background and a passion for conducting high quality research.

Experience or a strong interest in areas such as game theory, mechanism design, and LLMs be advantageous.

Fees and funding:

Full scholarships include tuition fees, a tax-free stipend at the UKRI rate for up to 3.5 years (totalling £20,780 for 2025/26, rising annually). UK, EU and Horizon Europe students are eligible for scholarships. Chinese Scholarship Council funded students are eligible for fee waivers. Funding for other international applicants is very limited and highly competitive. Overseas students who have secured or are seeking external funding are welcome to apply. For more information, please visit our postgraduate research funding pages.

How to apply:

Apply now

You need to:

• choose programme type (Research), 2026/27, Faculty of Engineering and Physical Sciences
• select Full time or Part time
• search for programme PhD Computer Science (7089)
• add name of the supervisor in section 2 of the application

Applications should include:

• research proposal
• your CV (resumé)
• 2 academic references
• degree transcripts and certificates to date
• English language qualification (if applicable)