Description

Category

Computer Science
 

Scope

Master Thesis 30 hp

2 students completing 30 credits each
 

Background

The rapidly increasing use of generative AI introduces challenges. We need to be careful to not lose elements of foundational verification principles enabling trust in software. As programming moves up in abstraction we need to adapt and develop our current solutions for quality assurance.

For example, when decisions are delegated to agents in software development we need to consider how we do quality assurance of agents and how we inspect agent decisions. How do we trace decision-making and determine responsibility? How do we determine how much we can trust automation via GenAI? We need to shape the design of GenAI-based software development assistance to what we need. Delegation to agents can provide high productivity gains but must be built so that engineers can trust tha...