Exploring moral paradoxes of fast fashion consumption via active learning

Australians now surpass the USA and UK as the world's leading consumers of fast fashion, despite nearly two-thirds expressing concerns about its environmental impact. This level of consumption carries serious social and environmental costs; from textile waste equivalent to filling four Sydney Harbour Bridges each year to unsafe working conditions across global supply chains.

At the same time, Australian sustainable fashion brands are growing in popularity. Together, these trends reveal a critical consumption paradox where consumers are aware of the harms of fast fashion yet struggle to fully transition to sustainable fashion alternatives. This paradox raises an important question: how do consumers move from fast fashion to sustainable fashion consumption despite knowing its harmful effects?

Traditional qualitative research methods, such as interviews, provide rich insights into consumer lived experiences. However, these methods are often limited by social desirability bias and the influence of social media discourse. This project overcomes these challenges using Active Learning, a semi-supervised machine learning method that combines computational analysis with human judgement.

In this method, the algorithm scans a large, unlabelled dataset and identifies the most uncertain instances – the "grey codes" – for manual review. Through this human-in-the-loop process, qualitative researchers iteratively identify and code complex moral paradoxes that that are often missed through passive coding. Using interviews with fast fashion consumers and social media data, the project will generate a gold-standard dataset and develop a novel theoretical framework explaining how consumers morally disengage from the harms of fast fashion.

By integrating Active Learning with qualitative analysis, the project advances theory by uncovering the moral and cognitive mechanisms sustaining hyperconsumption. It will also demonstrate a rigorous model for qualitative researchers using machine learning, offering practical insights for policymakers promoting sustainable consumption.

MDAP's expertise is essential to this project's core methodology. The team will lead data science component, sourcing and cleaning social media data from platforms including YouTube, TikTok, Instagram and Reddit. MDAP will implement and fine-tune the Active Learning framework, using uncertainty sampling to iteratively select the least certain segments for human annotation.

This complex technical task requires expertise in machine learning, natural language processing, and large-scale data architecture. The research team will provide domain expertise, develop the initial codebook and perform human annotation to ensure theoretical fidelity. Together, we will document the process to create a consultancy package for future qualitative scholars.

Expected outcomes include peer-reviewed publications, actionable insights for policymakers, and a rigorous model for qualitative researchers using machine learning. Findings will inform policy reforms strengthening the Australian Modern Slavery Act 2018 and promote consumer-centred sustainability measures.