Paper Type

Complete

Abstract

This paper advances Augmented Cognitive Extension (ACE), a process theory that explains how delegation to agentic AI can either scaffold or erode human cognition over time. We argue that delegation to agentic systems creates a paradox: it expands cognitive bandwidth while attenuating the regulatory capacities that sustain disciplined reasoning. ACE resolves this paradox by theorizing three developmental states—entanglement, cognitive extension, and cognitive atrophy—and two competing pathways through which delegation stabilizes into competence-reinforcing or competence-depleting equilibria. Movement among these states is governed by calibration and negotiation, enacted through five regulatory capacities: attentional synchronization, epistemic calibration, affective attunement, risk construal, and mandate control. We derive these capacities from the execution gaps and reconfigured epistemic boundaries that distinguish agentic delegation. The resulting model contributes to IS theory by explaining why structurally similar agentic systems yield divergent cognitive outcomes across users.

Paper Number

1559

Comments

SIG SVS

Download

Share

COinS
 
Aug 15th, 12:00 AM

Agentic AI and Delegation Paradox: An Augmented Cognitive Extension Perspective

This paper advances Augmented Cognitive Extension (ACE), a process theory that explains how delegation to agentic AI can either scaffold or erode human cognition over time. We argue that delegation to agentic systems creates a paradox: it expands cognitive bandwidth while attenuating the regulatory capacities that sustain disciplined reasoning. ACE resolves this paradox by theorizing three developmental states—entanglement, cognitive extension, and cognitive atrophy—and two competing pathways through which delegation stabilizes into competence-reinforcing or competence-depleting equilibria. Movement among these states is governed by calibration and negotiation, enacted through five regulatory capacities: attentional synchronization, epistemic calibration, affective attunement, risk construal, and mandate control. We derive these capacities from the execution gaps and reconfigured epistemic boundaries that distinguish agentic delegation. The resulting model contributes to IS theory by explaining why structurally similar agentic systems yield divergent cognitive outcomes across users.