Patsy Nwogu

Testing policy through simulation and statistical modelling.

AI Governance Lead, City Business AI (Traavu Ltd)

Birmingham, United Kingdom/Speaks English, Igbo and Yoruba

Patsy Nwogu, researcher in computational policy modelling
1.6mbeneficiaries across 345 primary health centres, through health data infrastructure at Orodata Science
152local authorities modelled in IPS-England, validated against DfE records
2public simulations anyone can run and interrogate, IPS-England and GRID

About

I am a researcher in computational policy modelling. I build instruments that show how a policy distributes access, delay and protection across a real population, including its most vulnerable members.

My research stays with populations that policy research often leaves thin: people with little capacity to self-report or contest, who carry the cost when a design fails. My public work includes IPS-England, a model of how children with special educational needs are identified in England and routed to support; GRID, on underreporting as a safeguarding consequence of mismatched policy; and IPS-Nigeria, being built with researchers, parents and interest groups for creating an educational access pathway for children with special needs in Nigeria. IPS-Nigeria is the first test of computational policy modelling as a pre-enactment methodology, not just a post-policy one. The same logic also sits on oversight conditions in AI operations, and the policies meant to hold them.

I have led and collaborated on projects across journalism, health data infrastructure, data and governance management and AI governance. This unconventional path gives me something unique: I do not study people as an abstract population, I work one to one, every day, with the people most affected by the laws I study and that experience keeps my work solutions-focused rather than decorative. I am building a body of work that people can actually use, not just read, research built to be interacted with by the people it is meant to serve.

Recognition

Tools

Methods

  • Agent-based simulation
  • Monte Carlo simulation
  • Statistical modelling

Languages and Libraries

  • Python
  • Mesa
  • JavaScript
  • React
  • Recharts

Publishing

  • GitHub
  • Zenodo
  • Harvard Dataverse
  • Netlify
  • ORCID

Scroll

Policy is written for a population.

It lands on people, and people are not the same.

When a policy isn't calibrated to the environment it lands in, the most vulnerable carry the cost first.

Simulation tests the policy before it tests the people.

Gilbert, Ahrweiler, Barbrook-Johnson, Narasimhan and Wilkinson (2018), Computational Modelling of Public Policy: Reflections on Practice. Journal of Artificial Societies and Social Simulation 21(1) 14.

Portfolio

IPS-England · agent-based model of England's SEND identification pathway

More money buys more plans, not fewer children turned away.

A model of the statutory route a child travels from an unmet special educational need to an Education, Health and Care plan, run across all 152 local authorities. Pick a change to the system and see what happens to the children who leave it without a plan.

50,081

children leave the pathway without a plan each year

Modelled baseline

Change the system

Resourcing and timing Appeal access
Ten-seed means from IPS-England, children leaving without a plan per year. Bars start at zero.

20.5 wksmodelled statutory processdeadline: 20 weeks
45.8%plans issued within 20 weeksEngland: 46.1%
107,952plans issued per yearEngland: 110,708
8,492reach a plan only by appealingmedian 51.4 weeks

GRID Observatory · safeguarding underreporting in specialist SEND schools

Safeguarding depends on what gets recorded.

Non-verbal learners cannot disclose harm, so protection rests on what practitioners capture. A concern is only recorded when three conditions hold at once: capacity to document, trained classification and protected time. Move them.

U = 1 − (C × A × T)

Underreporting 96.9%
Capture · C × A × T 3.1%
Largest marginal return Lift C or T Each point on C or T is worth 0.125 points of capture.

Papers

Preprint

The Mathematics of Safeguarding

Safeguarding policy · Agent-based simulation · SEND education

Preprint

The Environmental Gap in Agentic AI Governance

AI policy · Infrastructure assessment · Pre-deployment governance

In preparation

Simulating African Fintech Futures

Financial policy · Simulation · Built environment

Journal

Contact

I'm looking to collaborate with researchers in law, education, social science and policy.