Skip to content

GRA · AI and Financial Model Risk

Spatial financeWhere capital meets hazard, exposure and transition

Uncertainty quantification and model validation at AI and Financial Model Risk

Uncertainty quantification estimates how far a model’s output could be wrong, from data, parameters, structure and use. On this platform it supports model risk management by showing ranges, sensitivities and the conditions under which outputs should not be trusted.

01Register

Uncertainty quantification and model validation at AI and Financial Model Risk on the record.

Method recordSTEM method
  1. MethodUncertainty quantification and model validation
  2. Applied inAI and Financial Model Risk
  3. Canonical pageSTEM Commons
02Application

What uncertainty quantification and model validation does here.

Uncertainty quantification and model validation characterise how uncertainty in inputs, parameters and model structure affects predictions. They establish how much confidence a model’s outputs deserve. In artificial intelligence and financial stability, uncertainty quantification shows where risk estimates are robust and where they are fragile. It guides where better data or models would change decisions. The canonical description is on the STEM Commons.

Prediction intervals for loss models

Loss forecasts are reported with intervals from bootstrap or Bayesian methods, so decisions see how wide plausible outcomes are.

Model ensembles

Several model specifications are run in parallel, and their disagreement is used as a measure of structural uncertainty for decision makers.

Model risk capital or buffers

Uncertainty estimates inform conservative margins or buffers that firms hold for known model weaknesses, with their basis documented.

03Monitor

Uncertainty analysis.

The screen a team would use to follow this method: its measures, states and decisions. Names are examples and values are illustrative.

Uncertainty analysis

AI and model risk · Uncertainty quantification · Run R-7264

Sensitivity rankedGRA · AI and Financial Model Risk

Total-order sensitivity

Output distribution

Inputs

Models in inventoryBetamodels, statedexpert elicitation
Overdue validationsBetamodels, statedexpert elicitation
High-tier models with open findingsBetamodels, statedexpert elicitation
Models with drift alertsNormalmodels, statedexpert elicitation
Uncertainty analysis: the screen’s structure, terms and states for AI and model risk.Example names and illustrative values; a design, not a record or a live system.
04Data

What the method consumes.

Datasets and records used by uncertainty quantification and model validation on this platform, with the role each plays and the label its values carry. Sources are listed in the source register.

DataRolLabel
Model outputs and residualsError estimationMODELLED
Alternative specificationsStructural uncertaintyMODELLED
Back-testing dataRealised errorsOBSERVED
05Exposure, loss and capital

What it means for exposure, loss and capital.

The exposure, loss, scenario and capital views a risk officer, supervisor or ministry of finance would ask for, with the frameworks they report against. Structures follow supervisory and actuarial practice. Values are illustrative: never prices, ratings or advice.

Transmission channels

How a shock in Uncertainty quantification and model validation at AI and Financial Model Risk reaches credit, market, underwriting, liquidity and operational risk, and feeds back.

Channel structure after the Basel Committee’s work on climate-related risk drivers and transmission channels; links shown are typical paths, not estimated exposures.
NGFS climate scenarios · Network for Greening the Financial SystemIFRS S2 Climate-related Disclosures · ISSBTCFD recommendations · Financial Stability Board TCFDPrinciples for the effective management of climate-related financial risks · Basel Committee on Banking Supervision
06Validation

How results are checked.

Predictive distributions are tested for calibration and sharpness against held-out observations. Verification confirms that code solves the intended equations.

CheckHow it is done
Interval coverageStated intervals are checked against realised outcomes
Sensitivity reportingKey assumptions are varied and effects reported
CommunicationUncertainty is shown with results, not in appendices

Structural uncertainty and unknown unknowns are difficult to represent. Computational cost limits exploration of high-dimensional models.

07Proses

How the process runs.

Process explorer · Process

Uncertainty quantification and model validation at AI and Financial Model Risk: Hazard, exposure, vulnerability

10 steps4 stages

Proses

Uncertainty quantification and model validation at AI and Financial Model Risk: Hazard, exposure, vulnerability

In IPCC AR6 Working Group II, climate risk results from interactions between hazards and a system’s exposure and vulnerability, and responses can add risk.

  1. 01 Drivers

    1. 01HazardDataset · Independent validationThe potential occurrence of a natural or human-induced physical event or trend that may harm people, property, services or ecosystems.
    2. 02ExposureDataset · Independent validationThe presence of people, livelihoods, species, ecosystems, infrastructure or assets in places and settings that could be adversely affected.
    3. 03SensitivityDataset · Independent validationThe degree to which a system or species is affected, either adversely or beneficially, by climate variability or change.
    4. 04Capacity to cope and adaptDataset · Independent validationThe ability of systems, institutions, humans and other organisms to adjust to damage, take opportunities or respond to consequences.
  2. 02 Assessment

    1. 05Risk from responsesComponent · Independent validationRisks from climate responses arise when responses miss their objectives, or from trade-offs or negative side effects.
    2. 06VulnerabilityComponent · Independent validationThe propensity or predisposition to be adversely affected. It includes sensitivity to harm and lack of capacity to cope and adapt.
    3. 07RiskModel · AI and Financial Model RiskRisk results from dynamic interactions between hazards and the system’s exposure and vulnerability. Each may be uncertain and change.
  3. 03 Response

    1. 08Risk management and responseStep · Independent validationRisk management is plans, actions, strategies or policies that reduce the likelihood or size of adverse consequences.
    2. 09Residual riskRecord · Independent validationThe climate impact risk that remains after adaptation and mitigation efforts. Adaptation can redistribute risk between areas and groups.
  4. 04 Decision

    1. 10Decision by the responsible partyDecision · Independent validationThe framing names no single decision maker. Decisions on responses rest with the responsible party under its own rules.

Outline of an external standard or method, from its public sources. It describes what the framework says, not a record of any run. It does not show that Nexus or any partner performs, certifies or approves a step. A gate is a step of the method, never a Nexus approval.

09Stewardship

Who maintains the method.

Systems Science and Modelling stewards uncertainty quantification and model validation for the Consortium and reviews how it is applied across platforms.

11Vrae

Questions and answers.

What are the limits of uncertainty quantification and model validation in this field?
Structural uncertainty and unknown unknowns are difficult to represent. Computational cost limits exploration of high-dimensional models.
How are results checked?
Predictive distributions are tested for calibration and sharpness against held-out observations. Verification confirms that code solves the intended equations.
Are the values on this page real?
The data names, roles and checks are real descriptions of practice. The canvas shows a design of the method’s use, not a run; no result is reported here.
Where else is this method applied?
The canonical method page lists every platform that applies it.
VerslaeDra by

Dra bewyse vanaf hierdie bladsy by.

Reports is the Consortium’s quarterly publication. What you know about this subject can become evidence, a case or a correction in the next edition, credited to you and published under CC BY 4.0 where it is your own work.

Global Risks AllianceAI and Financial Model RiskMethods and TechnicalExponential Technology
KI- en finansiële modelrisiko

Bring an exposure or capital question to the Alliance.

Information and enquiries are open today. An enquiry receives a reference and names the next step; applications, payments, compute allocations and registrations open only when their team, terms and review capacity exist.