Beyond the h-index: Introducing the Fair Impact Score (FIS) for Network-Driven, Sentiment-Aware Bibliometrics

This repository contains the advanced, production-ready reference implementation of the Fair Impact Score (FIS) framework, a multi-dimensional alternative to the traditional h-index and raw citation counts, as introduced in the companion manuscript "Beyond the h-index: Introducing the Fair Impact Score (FIS) for Network-Driven, Sentiment-Aware Bibliometrics" (Kandi, 2026). Conventional evaluative metrics are vulnerable to linear bias, systemic citation manipulation (e.g., citation cartels), and semantic blindness toward negative or debunked research. The FIS algorithmic pipeline addresses these weaknesses through an optimized Three-Layer Correction Filtering Model, supplemented by contextual equity safeguards: Structural Layer (with Temporal Decay): Weighs the authority of citing sources using recursive graph-based centrality measures (e.g., PageRank, Eigenfactor). This version incorporates an exponential temporal decay modifier (e−λΔt) to ensure scores reflect contemporary relevance rather than legacy inertia. Behavioral Layer (Normalized Network Similarity): Detects and mitigates cartel-like citation loops and excessive self-citation using an adjusted Jaccard-style overlap coefficient (S(p,c)). A dynamic, field-normalized threshold (θf) scales according to domain-specific rolling averages, protecting large-scale “Big Science” collaborations from unfair penalization. Semantic Layer (Multi-Anchor Continuous Sentiment Analysis): Analyzes citation text blocks with natural language processing (NLP). Multiple intra-document contexts are parsed per citing paper, and sentiment is mapped continuously along a σ(c)∈[0.0,1.0] spectrum to capture nuanced academic rhetoric, hedge-phrasing, and polite disagreement. Contextual Equity Layer (Optimized Dynamic Damping): Mitigates institutional bias (the “Matthew Effect”) through a logarithmic Robin Hood multiplier (γ(c)) with a strict non-negative boundary (max⁡(0.1,… )), ensuring emerging and independent scholars are not disadvantaged by cold-start funding disparities. Repository Contents fis_framework.py: The finalized core Python module implementing normalized similarity, sentiment multipliers, exponential temporal decay, institutional damping coefficients, and aggregate researcher scoring functions. Usage and Open Science This computational template is curated in compliance with responsible metrics principles, including the San Francisco Declaration on Research Assessment (DORA). Execution instructions and metadata integration are provided in the configuration file. All source files are released under open-source terms to encourage peer review, transparent auditing, and validation across disciplines.