SHAPR: A Solo Human-Centred and AI-Assisted Practice Framework for Research Software Development

TL;DR

This paper addresses the challenge of solo HDR research software development in the age of AI, where rapid AI-assisted prototyping can undermine methodological rigour and accountability. It introduces SHAPR, a Solo, Human-centred, AI-assisted PRactice framework that operationalises Action Design Research for solo contexts by specifying roles, artefacts, reflective practices, and lightweight governance. SHAPR is treated as the primary design artefact and unit of analysis, evaluated formatively for coherence and alignment with ADR principles, rather than empirically instantiated. The framework aims to support knowledge production and HDR researcher training by linking Human-AI collaboration, artefact management, and reflective learning, laying a foundation for broader empirical applications and cumulative research on AI-assisted research software development.

Abstract

Research software has become a central vehicle for inquiry and learning in many Higher Degree Research (HDR) contexts, where solo researchers increasingly develop software-based artefacts as part of their research methodology. At the same time, generative artificial intelligence is reshaping development practice, offering powerful forms of assistance while introducing new challenges for accountability, reflection, and methodological rigour. Although Action Design Research (ADR) provides a well-established foundation for studying and constructing socio-technical artefacts, it offers limited guidance on how its principles can be operationalised in the day-to-day practice of solo, AI-assisted research software development. This paper proposes the SHAPR framework (Solo, Human-centred, AI-assisted PRactice) as a practice-level operational framework that complements ADR by translating its high-level principles into actionable guidance for contemporary research contexts. SHAPR supports the enactment of ADR Building-Intervention-Evaluation cycles by making explicit the roles, artefacts, reflective practices, and lightweight governance mechanisms required to sustain human accountability and learning in AI-assisted development. The contribution of the paper is conceptual: SHAPR itself is treated as the primary design artefact and unit of analysis and is evaluated formatively through reflective analysis of its internal coherence, alignment with ADR principles, and applicability to solo research practice. By explicitly linking research software development, Human-AI collaboration, and reflective learning, this study contributes to broader discussions on how SHAPR can support both knowledge production and HDR researcher training.

Figures (3)

• Figure 1: Overview of the SHAPR framework. The figure illustrates the conceptual structure of SHAPR, positioning Action Design Research as the guiding methodology for solo research software development. The framework integrates iterative Building--Intervention--Evaluation cycles, human--AI collaboration, and artefact-centred evidence generation to support reflective and learning-oriented research practice.
• Figure 2: Human--AI role multiplexing in solo research software development under SHAPR. The figure depicts conceptual human roles enacted by a solo researcher alongside AI-assisted support functions, emphasising human-centred governance and the non-delegation of decision authority to AI systems.
• Figure 3: Artefact and evidence flow across Action Design Research Building--Intervention--Evaluation cycles. The figure shows how research software artefacts, documentation, prompts, and reflections evolve iteratively and are treated as research evidence supporting reflection, abstraction, and learning across cycles.