Formalizing Trust in Artificial Intelligence: Prerequisites, Causes and Goals of Human Trust in AI

Abstract

A theoretical framework that formalizes the concept of trust in AI by analyzing its prerequisites, causes, and goals. The paper distinguishes between contractual trust (based on explicit agreements) and intrinsic trust (based on perceived benevolence), arguing that most discussions of AI trust conflate these distinct concepts.

Key Findings

• Trust in AI can be decomposed into contractual and intrinsic components
• Explainability serves trust by providing evidence for trustworthiness assessment
• The framework reveals that many "trust-building" techniques actually target different trust types

Source