From Framework to Program: Sequencing AI Autonomy in an Applied AI Associate Degree for Agentic Work
Year: 2026
Type: preprint
Abstract: Agentic AI systems now perform multi-step work that a person reviews only before or after execution. Graduates entering technical work therefore need to decide how much authority to give these systems, bound what they may do, and verify what they did. We describe a proposed 31-credit core for an Applied AI Associate of Applied Science (A.A.S.) degree at a U.S. community college, designed to build that capacity. The program instantiates the Calibrated Delegation Framework, which keys AI autonomy to the consequence of error and to the developmental value of a task to the learner. Its central design move is to plan, across four terms, the migration of tasks between allocation regimes: early courses run a "struggle first" regime in which students attempt work before consulting AI, and later courses add bounded, audited delegation of consequential work. Every learning outcome carries one of five competency dimensions and, where applicable, one of the framework's four operational dimensions (Bounding, Allocating, Auditing, Withholding). Ten courses include six new courses, two redesigned courses, and two capstones in an on-campus center that serves small-business clients. Assessment combines course-embedded reconciliation tasks with four program-level measurement points kept separate from grades. We report the design, the alignment evidence, and the assessment plan; no student outcome data are yet available. We close with practical implications for programs adapting the design and with its constraints.
Source: EdArXiv (OSF Preprints)
Author: Agelo Thalassinidis
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W7234065761
