Home // FUTURE COMPUTING 2025, The Seventeenth International Conference on Future Computational Technologies and Applications // View article

Interstitial b-SHAP-centric Amalgam for the Enhancement of an AI-centric Construct Validity Approach

Authors:
Steve Chan

Keywords: artificial intelligence systems; machine learning; construct validity; explainability; interpretability.

Abstract:
This paper describes an Artificial Intelligence (AI)-based Construct Validity Verification Methodology (CVVM) being advanced. The proposed methodology includes an amalgam utilization of temporal-centric Finite-Change Shapley-Owen values along with, among others, Generic Shapley-Owen values and Variance-Based Shapley-Owen values (i.e., a bespoke SHAP amalgam or b-SHAP implementation), CRiteria Importance through Intercriteria Correlation (CRITIC), and Preference Ranking Organization Method for Enrichment Evaluation (PROMETHEE) for enhancing the interpretability of not only the machine learning constituent components of an AI system, but also the interstices (e.g., between/among individual components as well as amalgams/clusters of components locally/globally). This approach extrapolates upon and furthers current proposals for the utilization of SHAP in local, global, and glocal (a hybridized intermediary of local and global) contexts. It turns out that this Interstitial SHAP-centric Amalgam (ISA), by better correlating features with each interim intended construct, potentially segues to better interpretability and construct validity at the component, interstitial, and overall system level, particularly when ISA is conjoined with a well-counterpoised Multi-Attribute Decision-Making (MADM)/Multi-Objective Decision-Making (MODM) Subjective Measures (SM)/Objective Measures (OM) paradigm and a modified Constriction Factor (CF)-Particle Swarm Optimization (PSO)-Robust Convex Relaxation (RCR)-Long Short-Term Memory (LSTM)-Deep Convolutional Neural Network (DCNN) (CPRLD) metaheuristic architectural construct.

Pages: 19 to 26

Copyright: Copyright (c) IARIA, 2025

Publication date: April 6, 2025

Published in: conference

ISSN: 2308-3735

ISBN: 978-1-68558-259-3

Location: Valencia, Spain

Dates: from April 6, 2025 to April 10, 2025