Efficiency focused analytic review of NLP and Machine Learning Applications in Legal Reasoning and Ethical Business Decision-Making
Abstract
While business leaders have increasingly sought to manage ethical compliance in their decision-making frameworks, their understanding of what responsible AI is and what should be prioritized remains fragmented. Within the broader objective of AI-driven models supporting legal professionals in their response to complex regulatory scenarios, we conducted a multi-method analysis for benchmarking best practices. We argue a key challenge with organizations’ efforts to manage machine-assisted decisions is that it is a contextually ambiguous concept for them. To bridge this gap, the present study evaluates the hierarchical criteria and performance attributes related to an efficiency-oriented framework, whereby both normative reasoning and computational accuracy are included. We interviewed legal analysts at multinational corporate governance units in Germany, Singapore, and Canada and identified systematic barriers to managing ethical dilemmas at different stages of automated decision processes. We were able to establish a balance between the need for transparency and the predictive performance by combining the output from three different models, each aiming at contextual relevance, in a hybrid AHP-TOPSIS structure. We also identified that inconsistencies between stakeholders’ intentions to align with ethical codes and what they identify in their machine learning outcomes, reveal important evaluative discrepancies that help to improve organizational understanding of AI-related obligations. The statistical analysis of the proposed decision support system provides a prioritization model of criteria from structured data that lies in cross-sectoral standards with an improved clarification of legal accountability. The procedure shows the great potential of natural language processing in combination with multi-criteria evaluation for informed reasoning even in an ambiguous legal setting.