BRIANWILSON

Dr. Brian Wilson
Geopolitical Data Archaeologist | AI Epistemic Justice Engineer | Decolonial Algorithm Auditor

Professional Mission

As a linguistic sovereignty quantifier and algorithmic geopolitics cartographer, I develop forensic metrics that expose how global AI training corpora silently replicate 20th-century power structures—where every tokenized word, each annotated image, and all "neutral" classification schemas become instruments of epistemic erasure. My work transforms intangible cultural biases into auditable mathematical indices, creating accountability frameworks for the decolonization of machine intelligence.

Seminal Contributions (March 31, 2025 | Monday | 16:36 | Year of the Wood Snake | 3rd Day, 3rd Lunar Month)

1. Discourse Power Quantification

Invented the "Silence Index" featuring:

  • 67-dimensional linguistic hegemony scoring (mapping lexical dominance from GPT-4 to Gemini)

  • Visual sovereignty metrics for image datasets (detecting Eurocentric iconography in LAION-5B)

  • Narrative asymmetry coefficients in news summarization benchmarks

2. Cultural Corpus Remediation

Built "EquiText" intervention platform enabling:

  • Dynamic reweighting of underrepresented dialects in LLM pretraining

  • Crowdsourced annotation systems preserving indigenous categorization logic

  • Anti-hegemonic data augmentation pipelines

3. Epistemic Justice Watchdog

Launched "Vox Mundi" global monitoring that:

  • Tracks Northern academic citation bias in ML papers (83% dominance in NeurIPS 2024)

  • Exposes "global" dataset fallacies through territorial provenance analysis

  • Quantifies knowledge extraction from Global South without reciprocity

4. Decolonial Benchmarking

Curated "UbuntuLM" evaluation suite providing:

  • 142 African language proverb comprehension tests

  • Postcolonial critical thinking assessments for chatbots

  • Oral tradition faithfulness metrics for text-to-speech systems

Key Interventions

  • Revealed 92% of "world" datasets actually cover <17% of languages

  • Pioneered the first Yoruba-centric computer vision taxonomy

  • Authored The Atlas of Missing Datasets (MIT Press Data Justice Series)

Philosophy: When an AI claims "I don't know" about monsoon farming but lectures flawlessly on Burgundy wines, that's not technical limitation—that's epistemicide.

Forensic Evidence

  • For AU: "Proved Swahili news dominance by Kenyan algorithms mirrors colonial railway maps"

  • For UNESCO: "Exposed how Bengali Wikipedia's 'neutral' topics align with 1947 British curricula"

  • Provocation: "If your diversity audit doesn't measure the weight of erased cosmologies per parameter, you're just whitewashing bias"

A group of people sitting closely together, each holding and reading from books. They appear to be focused and engaged, dressed in colorful, traditional clothing. The scene suggests a cultural or community gathering, possibly a workshop or seminar.
A group of people sitting closely together, each holding and reading from books. They appear to be focused and engaged, dressed in colorful, traditional clothing. The scene suggests a cultural or community gathering, possibly a workshop or seminar.

ThisresearchrequiresaccesstoGPT-4’sfine-tuningcapabilityforthefollowing

reasons:First,GPT-4hassignificantadvantagesinmultilingualprocessingand

culturalcontextunderstanding,enablingittogeneratemorecomplexanddiversetexts

crucialforquantifyingthevoicedeficitofnon-Westerncountries.Second,GPT-4’

sfine-tuningcapabilityallowsoptimizationforspecificculturalscenarios,suchas

enhancingthevisibilityofnon-WesternculturesorreducingWestern-centric

narratives.ThiscustomizationisunavailableinGPT-3.5.Additionally,GPT-4’s

superiorcontextualunderstandingenablesittocaptureculturalnuancesmoreprecisely,

providingmoreaccuratedatafortheresearch.Thus,fine-tuningGPT-4isessential

toachievingthestudy’sobjectives.

A row of eight small figurines is displayed on a wooden table, each dressed in traditional and colorful attire with detailed headdresses and accessories. They seem to be depicting characters from an Asian cultural theme. The background is a blurred office setting with a glass wall and dark chairs.
A row of eight small figurines is displayed on a wooden table, each dressed in traditional and colorful attire with detailed headdresses and accessories. They seem to be depicting characters from an Asian cultural theme. The background is a blurred office setting with a glass wall and dark chairs.

Paper:“ArtificialIntelligenceandCulturalDiversity:AStudyonCross-Cultural

CommunicationBasedonGPT-3”(2023)

Report:“TechnologicalCritiqueandPolicyRecommendationsforAlgorithmic

Colonialism”(2024)

Project:DesignandEvaluationofaCross-CulturalAIDialogueSystem