Frictional AI. Designing Desirable Inefficiencies in Decision Support Systems for Knowledge Work

Abstract

My research involves the conceptualization of ‘Frictional AI’ as a novel approach for enhancing cognitive engagement in Decision Support Systems (DSS) through intentional design inefficiencies. Through empirical studies and theoretical analysis, I explore the balance between human intuition and automation-induced enhancements to decision-making in medical diagnostics. With the introduction and assessment of four distinct frictional protocols (cautious, comparative, judicial, and adjunct), this design framework prioritizes the efficacy and integrity of human knowledge work, ensuring that professionals are engaged, critical thinkers, capable of counteracting automation bias and deskilling–even at a slight cost in efficiency and comfort.