Surfacing the Invisible: AI and the IS Discipline
The current debate about Artificial Intelligence (AI) in scientific research concerns problems that appear in the published record – fabricated citations, AI-generated submissions, the strain on peer review, vacuous prose devoid of substance. These problems are real and measurable, but they represent only the tip of a far deeper iceberg. Beneath them lie effects that produce no signal the system is trained to detect: activities that did not take place, capacities that were not built, guarantees that no longer hold. This commentary surfaces these invisible costs for the information systems discipline. When doctoral students outsource the difficult work of writing, reading, and deriving to AI, the apprenticeship mechanism through which scholarly expertise has always been formed is quietly dismantled. When AI-mediated literature workflows replace deep engagement with targeted retrieval, the cross-domain encounters that have historically generated the field's most consequential research questions are lost. And when the activities that build evaluative judgment disappear, the disciplinary commons on which editors, reviewers, and tenure committees depend is eroded. Two further dynamics compound the problem: empirical evidence increasingly rests on commercial models that cannot be re-inspected or reproduced, and publication counts have been decoupled from the cognitive engagement they once represented. The harder question, and the one current evaluation mechanisms are not equipped to answer, is how a scientific community detects what is no longer being attempted. The IS discipline must move deliberately and intentionally toward revaluing what AI makes scarce rather than what it makes abundant. We offer concrete suggestions for a path forward here.