HAIC-SE: A Framework for Human-AI Collaboration in Software Engineering

Name of participant: Ludwig Felder

Project’s name: HAIC-SE: A Framework for Human-AI Collaboration in Software Engineering

Project description:

Generative AI, particularly large language models and agentic systems, is reshaping software development. Yet sustained productivity gains in professional settings often fail to materialize. The reason is not insufficient model capability but poorly designed human-AI interaction: unclear goals are not reliably translated into actionable steps, small variations in phrasing or context produce divergent results, and integration into team conventions, domain knowledge, and quality requirements remains fragile.
HAIC-SE (Human-AI Collaboration in Software Engineering) addresses this gap. It treats software development as a collaborative process in which goals are specified, context is built, results are verified, and approaches are continuously adapted. The project structures the development cycle as a sequence of reciprocal human and AI contributions, focusing on the DevOps phases Plan (intent capture), Code (context-aware implementation), and Test (verification), plus cross-cutting feedback loops.
The work combines three strands: systematic identification and validation of collaboration patterns, development of adaptive interfaces that align with individual work styles and team standards, and an evaluation framework integrating classical software engineering metrics with human-computer interaction indicators such as cognitive load.
Implementation is iterative and industry-driven in cooperation with DATEV eG, using real tasks, repositories, and workflows. The resulting building blocks integrate into established toolchains, reducing transfer risk and yielding reusable artefacts.

Software Campus Partner: TU München & DATEV eG

Implementation period: 01.01.2026 – 31.12.2027