An Explainable Auditing Framework for Assessing Financial Statement Risk Based on Explainable Artificial Intelligence: A Mixed-Methods Approach

Abstract

This study aimed to design and validate an explainable auditing framework for assessing financial statement risk based on explainable artificial intelligence. Grounded in the pragmatist paradigm, which enables the integration of inductive and deductive approaches, this study employed a mixed-methods methodology with a sequential exploratory design and was developmental–applied in terms of its objective. In the qualitative phase, 39 selected scientific sources were analyzed from an initial pool of 188 sources using Sandelowski and Barroso’s seven-stage model. The coding process yielded 102 codes, which were subsequently organized into 42 initial codes, 18 organizing codes, and four overarching codes: “explainable auditing based on explainable artificial intelligence,” “antecedents,” “outcomes,” and “mediating factors.” The reliability of the qualitative analysis was confirmed using Cohen’s kappa coefficient (0.742). In the quantitative phase, data were collected through a questionnaire administered to 98 auditors, accountants, and information technology specialists. The instrument’s validity was confirmed using the content validity ratio (CVR) and content validity index (CVI), while its reliability was established using Cronbach’s alpha. Partial least squares structural equation modeling (PLS-SEM) was employed to test the conceptual model. The findings indicate that organizational and infrastructural requirements, as well as professional trust-building, play significant roles in producing operational and economic outcomes and improving audit quality. Moreover, auditors’ trust in explainable artificial intelligence models functions as a mediating factor affecting certain outcomes.