Paper Number
2954
Paper Type
Complete
Description
This paper reports on the 4th phase of a multi-phase ‘human-in-the-loop’ approach to conducting systematic literature reviews (SLRs). This paper explores the integration of Generative Artificial Intelligence (GenAI) into the Grounded Theory (GT) methodology to advance qualitative research techniques. Utilizing a 'human-in-the-loop' approach, we use a Large Language Model (LLM) (specifically ChatGPT4) to perform Open, Axial, and Selective (OAS) coding on a set of DevOps research abstracts. We visualize the ChatGPT4-generated coding output (10 categories and 7 relationships) as a DataOps conceptual model, organized around the core category (‘Adapting to Agile and DevOps Practices’). We conclude the paper with an evaluation of our coding output and a reflection on our ‘human-in-the-loop’ approach. Our work highlights two considerations for human-AI collaboration: (i) enhancing the efficiency and creativity of qualitative data analysis and (ii) prompting a re-evaluation of the researcher’s role/responsibility in enhancing methodological transparency.
Recommended Citation
Sammon, David; McCarthy, Stephen; Thummadi, Babu Veeresh; Wibisono, Arif; and Fitzgerald, Brian, "Exploring the Potential of Large Language Models (LLMs) for Grounded Theorizing: A Human-in-the-Loop Configuration" (2024). ICIS 2024 Proceedings. 3.
https://aisel.aisnet.org/icis2024/adv_theory/adv_theory/3
Exploring the Potential of Large Language Models (LLMs) for Grounded Theorizing: A Human-in-the-Loop Configuration
This paper reports on the 4th phase of a multi-phase ‘human-in-the-loop’ approach to conducting systematic literature reviews (SLRs). This paper explores the integration of Generative Artificial Intelligence (GenAI) into the Grounded Theory (GT) methodology to advance qualitative research techniques. Utilizing a 'human-in-the-loop' approach, we use a Large Language Model (LLM) (specifically ChatGPT4) to perform Open, Axial, and Selective (OAS) coding on a set of DevOps research abstracts. We visualize the ChatGPT4-generated coding output (10 categories and 7 relationships) as a DataOps conceptual model, organized around the core category (‘Adapting to Agile and DevOps Practices’). We conclude the paper with an evaluation of our coding output and a reflection on our ‘human-in-the-loop’ approach. Our work highlights two considerations for human-AI collaboration: (i) enhancing the efficiency and creativity of qualitative data analysis and (ii) prompting a re-evaluation of the researcher’s role/responsibility in enhancing methodological transparency.
Comments
20-Theory