Qualitative Coding in the Computational Era: A Hybrid Approach to Improve Reliability and Reduce Effort for Coding...

Area of Science:

• Social Sciences
• Computational Social Science
• Qualitative Research Methods

Background:

• Sociologists advocate for computational tools in qualitative research.
• Traditional computational methods often conflict with qualitative research practices.
• Existing methods lack contextual interpretation capabilities for qualitative data.

Purpose of the Study:

• To introduce a hybrid human-machine learning approach (HHMLA) for qualitative data coding.
• To demonstrate HHMLA's efficiency and reliability compared to human-only coding.
• To provide a scalable solution for qualitative data analysis.

Main Methods:

• Developed a hybrid human-machine learning approach (HHMLA).
• Integrated contemporary iterative qualitative coding with advanced word embedding models.
• Applied HHMLA to analyze 87 human-coded ethnographic interview transcripts.

Main Results:

• HHMLA codes data sets with a fraction of the effort of human-only strategies.
• Significant labor savings (hundreds of hours) demonstrated in qualitative studies.
• Improved coding reliability achieved through the hybrid approach.

Conclusions:

• HHMLA offers a promising model for coding large qualitative data sets.
• Addresses limitations of purely human or conventional computational coding approaches.
• Facilitates qualitative data analysis where human coding is logistically prohibitive.