Paper Type
Complete
Abstract
Artificial Intelligence (AI) systems are increasingly embedded in professional reasoning, yet their outputs often arrive as incomplete claims lacking transparent warrants and rebuttals. Drawing on a 16-month ethnography of AI integration in a pioneering Department of Radiology deploying 17 FDA-cleared models across three configurations (automated triage, diagnostic decision support, and second-opinion augmentation), we introduce argumentation labor to describe the cognitive and communicative meta-work required to adjudicate machine-generated reasoning. Using Toulmin's argumentation model and the metahuman systems framework, we identify four forms of this labor (Reconstructive, Comparative, Adjudicative, and Integrative) operationalized through delegation, monitoring, cultivation, and reflection. We position argumentation labor as the cognitive successor to articulation work, showing how professional expertise repairs the limits of machine reasoning while professionals absorb asymmetric liability, with direct implications for the design of sustainable human–AI collaborative systems in high-stakes domains.
Paper Number
1392
Recommended Citation
S.G., Akhil and Lyytinen, Kalle, "Invisible Labor of Argumentation: Ethnography of Radiologists Using AI" (2026). AMCIS 2026 Proceedings. 3.
https://aisel.aisnet.org/amcis2026/sigcnow/sigcnow/3
Invisible Labor of Argumentation: Ethnography of Radiologists Using AI
Artificial Intelligence (AI) systems are increasingly embedded in professional reasoning, yet their outputs often arrive as incomplete claims lacking transparent warrants and rebuttals. Drawing on a 16-month ethnography of AI integration in a pioneering Department of Radiology deploying 17 FDA-cleared models across three configurations (automated triage, diagnostic decision support, and second-opinion augmentation), we introduce argumentation labor to describe the cognitive and communicative meta-work required to adjudicate machine-generated reasoning. Using Toulmin's argumentation model and the metahuman systems framework, we identify four forms of this labor (Reconstructive, Comparative, Adjudicative, and Integrative) operationalized through delegation, monitoring, cultivation, and reflection. We position argumentation labor as the cognitive successor to articulation work, showing how professional expertise repairs the limits of machine reasoning while professionals absorb asymmetric liability, with direct implications for the design of sustainable human–AI collaborative systems in high-stakes domains.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
SIG CNOW