Quantifying a firm's AI engagement: Constructing objective, data-driven, AI stock indices using 10-K filings

2025-01-03

Technological forecasting & social change

AI Summary

Current AI-themed investment strategies lack objective, data-driven metrics to quantify firms’ AI engagement, rendering them susceptible to subjective judgment. This paper proposes a novel AI exposure quantification framework grounded in 10-K annual reports of publicly listed firms. It integrates domain-adapted BERT fine-tuning, named entity recognition, an AI technology lexicon, and weighted TF-IDF to automatically extract verifiable evidence of AI adoption from unstructured regulatory filings—mapping such evidence to financially interpretable, dynamic index constituents for the first time. The resulting market-wide AI exposure index demonstrates statistically significant outperformance in backtests versus conventional benchmarks. Moreover, index values exhibit strong positive correlations with subsequent AI patent filings, AI-related hiring activity, and stock price reactions to AI-relevant events—validating both methodological innovation and practical investment utility.

Technology Category

Humans and AI: AI for AccessibilityPhilosophy and Ethics of AI: Philosophical Foundations of AINatural Language Processing: Information Extraction

Application Category

Search and Retrieval-Augmented AI: Web evaluation methodologies and metricsWeb Mining and Content Analysis: Bridging structured and unstructured dataGraph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Problem

• Artificial Intelligence
• Stock Investment
• Objective Evaluation

Innovation

• AI Investment Index
• Text Analysis
• Machine Learning in Finance