|
|
 |
| |
|
|
|
Ethical agent framework for Socio-Technical Alignment of Human-AI Interaction in AI-driven Software Engineering |
|
|
|
PP: 1669-1705 |
|
|
doi:10.18576/amis/200614
|
|
|
|
Author(s) |
|
|
|
Amany A. Slamaa,
Ahmed I. Sharaf,
|
|
|
|
Abstract |
|
|
| The integration of generative AI particularly LLMs into software engineering is fundamentally reshaping software development practices from technical activity into a complex socio-technical endeavor. This shift introduces critical challenges in trust calibration, cognitive load, collaboration, provenance and ethical governance. Through a systematic state-of-the-art analysis of literatures from 2018-2026, this study synthesizes the evolution of human-computer interaction in human-AI collaboration within AI-driven software engineering. The study identifies a paradigm shift from deterministic and tool-centric automation to probabilistic, agentic collaboration, Human-centric process. It highlights transformation from capability to governance and control over AI. Further, it addresses persistent gaps in ethical and socio-technical aspects through Human-AI interaction in automated software engineering. In response, the study proposes a novel ethical agent framework operationalizes socio-technical and ethical-socio-technical alignments across four key dimensions; cognitive load, collaboration, trust and ethical-social impact. The Delphi study is used to assess the feasibility and sustainability of a suggested ethical AI agent model through six main criteria, which were feasibility, usefulness, trust calibration, reduction of cognitive load, ethics adequacy and integration readiness. Final conclusions show overall consensus is 92.3% with highest agreement for Ethical Adequacy by 97% and the lowest agreement for Integration Readiness by 89%. These results validate the strength of the framework and its suitability to the prototype implementation in SDLC processes. The contributions provide structured analysis of Human-AI interaction, expose underexplored research avenues, and offer an integrative model to foster responsible, effective and human-centric AI adoption in modern automated software engineering workflows. |
|
|
|
|
 |
|
|