Intelligent Decision Assistance Versus Automated Decision-Making: Enhancing Knowledge Work Through Explainable Artificial Intelligence

Automation bias is a real workplace risk: The paper identifies that over-reliance on automated AI decisions can cause workers to uncritically accept AI recommendations, potentially leading to poor decision-making when they defer too heavily to the system.

Deskilling of knowledge workers threatens AI development: When AI systems make decisions for workers rather than with them, those workers lose the expertise and critical thinking skills needed to train, validate, and improve AI systems in the future—creating a problematic cycle.

Explainable AI without recommendations may be the solution: Rather than presenting AI-generated answers, the paper proposes showing workers how the AI reasons through a problem while letting workers make their own final decisions, preserving both human judgment and expertise.

"Intelligent Decision Assistance" offers a middle path: This approach balances AI's computational benefits with human oversight by supporting knowledge workers' decision-making rather than replacing it, addressing concerns about both automation bias and skill erosion.

This approach is empirically supported: The paper cites evidence from multiple studies showing that providing AI explanations without concrete recommendations helps maintain worker engagement and decision quality compared to fully automated systems.