Affective traces of AI labor: workers' emotions, job-market narratives, and emerging vulnerabilities in Madagascar

This paper looks at how people in Madagascar talk about, promote, and debate AI-related work through social media. Conversations about artificial intelligence usually focus on algorithms, innovation, and technical infrastructure, but they rarely explore what it's actually like for workers in less visible digital labor markets to encounter AI in their everyday job searches and work experiences. In Madagascar, where outsourcing, BPO services, freelancing, multilingual jobs, and online employment are becoming more common, AI shows up as both a potential opportunity and a new source of uncertainty and risk.

The study examines posts, job advertisements, comments, and discussions about AI, ChatGPT, digital work, freelancing, BPO, data annotation, transcription, translation, content creation, remote work, and online employment. It uses a mixed-methods approach that combines qualitative content analysis, affective discourse analysis, descriptive quantitative analysis, and AI-assisted text classification. The quantitative side tracks how AI-related job postings have changed over time, what kinds of work are being advertised, and what proportion of comments express emotions like fear, distrust, enthusiasm, hope, or suspicion. The qualitative side digs into how workers make sense of AI-related labor, whether they see it as a way to earn income, gain global mobility, build skills, or whether they frame it in terms of replacement, exploitation, uncertainty, or exclusion.

AI tools help analyze larger amounts of text by classifying posts, detecting emotional language, and identifying patterns in how workers describe their experiences. But these outputs are not treated as the final answer. They are verified, interpreted, and grounded in context through manual coding and close reading, especially when dealing with irony, multilingual expressions, local meanings, or ambiguous emotional signals.

The paper also takes research ethics seriously. Because it relies on social media data, strict anonymization procedures are applied, personally identifiable information is removed, and no quotations or details are included that could expose individuals or online communities. Extra care is taken with semi-public employment groups, where discussions might be accessible but are still socially sensitive.

The paper argues that workers' emotions aren't just personal reactions to technological change. They're affective traces of how AI labor markets are actually experienced, negotiated, and contested from Madagascar. By analyzing social media discussions around AI-related work, the study reveals emerging vulnerabilities, precarious employment, pressure to constantly learn new skills, dependence on platforms, scams, linguistic inequalities, and fears of being replaced, while also showing how AI is imagined as a promise of income, mobility, and a way into global digital labor markets.