Answers are grounded in this paper and cite their sources.
The spread of generative AI has made "brain outsourcing"—delegating cognitive tasks to AI—a common phenomenon in higher education, especially in low-tier private undergraduate colleges. Existing studies explain it mainly at the individual level, neglecting the institutional consequences of the technologies used to govern it. Drawing on design-based research principles, this study conducted a three-group quasi-experiment in a Data Structures course at such a college: an adversarial workflow (Group A, n = 36), a negotiative workflow (Group B, n = 40), and a traditional control group (n = 44), triangulated across five chains of evidence. Results show that Group A's direct-dumping rate fell to 11.1% (vs. 57.5% and 75.0%), yet concealed forms—cosmetic modification and division-of-labor outsourcing—rose compensatorily, leaving the composite outsourcing index unchanged across groups (p = 0.864). Process engagement rose sharply while output authenticity stayed low (comment–code matching 3.62/5; only 27.8% of difficulty tags authentic), indicating "performative engagement". Only 1 of 120 students adopted a self-regulated strategy. In Group A, perceived control rose (+2.17) and negotiation space fell (−0.91) without meaningful growth in autonomy; the 18.3-point advantage over the control group in comprehension depth shrank to 6.8 points in transfer application. The "non-outsourceable" workflow thus changed outsourcing's form, not its deep structure, alienating learning into an adversarial game of evading detection—a new hidden curriculum. Agent design should shift from adversarial surveillance to dialogic support.
No takes yet. Share an insight, caveat, or question.
Chen et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: