Trustworthy Agent-based Decision Support in Software Development
| Reference number | |
| Coordinator | Lunds universitet - Lunds Tekniska Högskola Inst f datavetenskap |
| Funding from Vinnova | SEK 6 074 628 |
| Project duration | April 2026 - April 2028 |
| Status | Ongoing |
| Venture | Advanced digitalization - Industrial needs-driven innovation |
| Call | Industrial applied AI by advanced digitalization 2026 |
Purpose and goal
Generative AI is transforming software development - a development that offers great opportunities but also challenges. We are seeing a wave of new AI-based development tools. How do we gain insight into the decisions these tools make and to what extent can we trust them? In this project, we explore how AI-based decision support for software development should be designed to provide reliable, high-quality assistance.
Expected effects and result
We expect results in the form of new prototypes that realize agent designs, as well as evaluation with developers and benchmarks, insights into interaction with agents from a trust perspective, insights into the management of agent systems in software development, and insights into the efficiency and precision of agent design focused on automatic code updating. We expect a direct impact on our partners, and further out into the industry via workshops and sharing via open platforms.
Planned approach and implementation
The project is divided into four work packages (WP); Bosch is running WP1 focused on agent system management, Axis is running WP2 focused on agents and trust, LU is running WP3 focused on agent context and code manipulation, and LU is running WP4 focused on dissemination of results via workshops, open platforms and reports. Methodologically, we plan to work with literature and tool reviews, empirical methods, interaction and technical design, and benchmarking.