Paper Number

1360

Paper Type

CRP

Abstract

As a holistic conceptualization of AI-advised decision-making currently does not exist, we propose such conceptualization by utilizing a proven framework: Toulmin’s Model of Argumentation. To achieve this, we break down AI advice into its core elements; namely the AI prediction, the AI explanation, and AI confidence level. We argue that each of these elements can be mapped to the argumentative elements proposed by Toulmin’s Model: The prediction constitutes grounds and claim, the explanation warrant and backing, and the confidence level the qualifier. Through this new perspective, this conceptual paper offers three main contributions: 1) We present the first holistic conceptualization for AI-advised decision-making, 2) Building on the proven explanatory powers of TMA, our novel conceptualization deepens our understand of contemporary issues in humans interacting with AI advice, and 3) The conceptualization can be used by practitioners to build more persuasive AI systems for real-world applications.

Share

COinS
 
Jun 14th, 12:00 AM

UNPACKING AI ADVICE FOR DECISION-MAKING: A NOVEL TOULMIN-BASED CONCEPTUALIZATION

As a holistic conceptualization of AI-advised decision-making currently does not exist, we propose such conceptualization by utilizing a proven framework: Toulmin’s Model of Argumentation. To achieve this, we break down AI advice into its core elements; namely the AI prediction, the AI explanation, and AI confidence level. We argue that each of these elements can be mapped to the argumentative elements proposed by Toulmin’s Model: The prediction constitutes grounds and claim, the explanation warrant and backing, and the confidence level the qualifier. Through this new perspective, this conceptual paper offers three main contributions: 1) We present the first holistic conceptualization for AI-advised decision-making, 2) Building on the proven explanatory powers of TMA, our novel conceptualization deepens our understand of contemporary issues in humans interacting with AI advice, and 3) The conceptualization can be used by practitioners to build more persuasive AI systems for real-world applications.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.