可信LLM生态系统博士职位

How can we design LLM-enabled systems that people and organizations can trust? Study how trustworthiness concerns can be translated into requirements and architectural decisions, with attention to openness, automation, control and accountability.

Information Large Language Models are increasingly becoming part of systems in which people, organizations, software systems and AI-enabled services work together. As these systems become more open and AI is given more responsibility, new questions arise. Who needs to trust whom, or what? What are they trusting them to do? Under what conditions are they willing to do so? And what does this mean for the requirements and architecture of these systems?

Many existing frameworks describe what trustworthy AI should provide, for example transparency, accountability, privacy, reliability, security and human oversight. The challenge is to translate these concerns into concrete requirements and architectural decisions. These concerns can also conflict. More transparency may affect privacy, for example, while more automation may reduce human control.

In this PhD, you will work at the intersection of requirements engineering, software architecture and trustworthy AI. You will study how different stakeholders understand trust in LLM-enabled systems and how their expectations, together with organizational and regulatory concerns, can be translated into requirements and architectural drivers. You will also study how different stakeholders can reason about choices concerning openness, information sharing, automation, human oversight and accountability.

You will develop and evaluate methods and models that help architects and other stakeholders identify trustworthiness requirements and make informed architectural decisions. Possible outcomes include approaches for eliciting and analyzing trust requirements, architecture principles or patterns, and contributions to a reference architecture. The exact direction and artefacts will develop during the PhD, giving you room to shape your own research questions and contributions.

This PhD is part of LLM4LM (Large Language Models for Logistic Management), a collaborative research project involving TU/e, TNO and several industry partners. The project investigates how LLMs and related AI technologies can support logistics and regulatory compliance in reliable, transparent and trustworthy ways. You will have the opportunity to study real AI adoption and architectural decision-making as it unfolds, working with academic and industry partners and across different application cases.

The research will use an empirical and design-oriented approach. Depending on the research questions, this can include interviews, workshops, document analysis, case studies, and the design and evaluation of methods or models. You will build on and contribute to ongoing research within LLM4LM while developing your own PhD research trajectory.