Park et al. 2026 argue that deskilling belongs on the top-tier AI-governance agenda alongside alignment, bias, and misuse: their model "reframes the policy question from 'should we adopt AI?' to 'how capable should AI be allowed to become in specific domains,'" warning that "incremental AI improvement can trigger discontinuous societal consequences" near a critical threshold K*≈0.85.
The cited text states verbatim that the threshold 'reframes the policy question from "should we adopt AI?" to "how capable should AI be allowed to become in specific domains,"' that 'incremental AI improvement can trigger discontinuous societal consequences,' that current frameworks focus on 'alignment, bias, and misuse' while deskilling 'warrants comparable attention,' and that K*≈0.85.
Written by Kimi K3 via Ollama Cloud · checked by GLM-5.3 via Ollama Cloud · 1 Oct, 04:46
Source chain
AEvery quote below was checked, without a model, to appear verbatim in its source.
- 01
“reframes the policy question from “should we adopt AI?” to “how capable should AI be allowed to become in specific domains before mandatory capability-preservation measures are required?””
Full-text passage · p. 17 · evidence E2
3 Discussion Our model offers four principal insights for AI governance. First, the critical thresholdK ∗ reframes the policy question from “should we adopt AI?” to “how capable should AI be allowed to become in specific domains before mandatory capability-preservation measures are required?” The existence ofK ∗ implies that incremental AI improvement can trigger discontinuous societal consequences—a concern invisible to linear risk assessments. Current AI governance frameworks focus on alignment, bias, and misuse [47, 48]; our findings suggest that deskilling risk warrants comparable attention. Notably, Scheffer and colleagues have shown that critical transitions in complex systems are often preceded by generic early-warning signals such as critical slowing down, analogous to those identified in social polarization dynamics [49], and increased variance [50, 51]. Monitoring these signals in AI-dependent skill metrics could provide advance warning of approachingK∗, particularly as frontier models approach or exceed human-level performance across broad task domains.
Passage read from arxiv.org, which may be a preprint rather than the published version.
- 02
“the model identifies a critical thresholdK ∗∼0.85 (scopedependent; broader AI scope lowersK∗) beyond which capability collapses abruptly”
Full-text passage · p. 1 · evidence E3
The enrichment paradox: critical capability thresholds and irreversible dependency in human–AI symbiosis Jeongju Park1, Musu Kim 1, Sekyung Han 1,* 1Department of Electrical Engineering, Kyungpook National University, Daegu, Republic of Korea *Corresponding author: skhan@knu.ac.kr Abstract As artificial intelligence assumes cognitive labor, no quantitative framework predicts when human capability loss becomes catastrophic. We present a two-variable dynamical systems model coupling capability (H) and delegation (D), grounded in three axioms: learning requires capability, practice, and disuse causes forgetting. Calibrated to four domains (education, medicine, navigation, aviation), the model identifies a critical thresholdK ∗∼0.85 (scopedependent; broader AI scope lowersK∗) beyond which capability collapses abruptly—the “enrichment paradox.” Validated against 15 countries’ PISA data (102 points,R 2 =0.946, 3 parameters, lowest BIC), the model predicts that periodic AI failures improve capability 2.7-fold and that 20% mandatory practice preserves 92% more capability than the simulation baseline (which includes a 5% background AI-failure rate).
Passage read from arxiv.org, which may be a preprint rather than the published version.