Skip to content

GRA · AI and Financial Model Risk

Spatial financeWhere capital meets hazard, exposure and transition

Analysis and uncertainty

Model risk is managed through inventory, documentation, conceptual soundness review, outcome analysis, benchmarking and ongoing monitoring. This page explains each step, how machine learning changes it, and why generative and agentic AI sit partly outside current guidance.

03Explainability and fairness methods

Explainability and fairness methods, with limits.

পদ্ধতিGivesসীমা
SHAP and other attribution methodsContribution of each input to a predictionCorrelated inputs blur attribution
Counterfactual explanationsWhat change would alter the decisionMay suggest infeasible changes
Group fairness metricsOutcome differences across groupsMetrics can conflict
Challenger modelsIndependent benchmarkShares data weaknesses
04প্রক্রিয়া

How the process runs.

Process explorer · Process

Analysis and uncertainty: Source to evidence record

10 steps4 stages2 gates

প্রক্রিয়া

Analysis and uncertainty: Source to evidence record

A source is registered and assessed, its record labelled and classified, and a Digital Evidence Passport carries provenance; a correction reaches every dependent record.

  1. 01 Intake

    1. 01Identify the sourceSource · The source stewardDiscovery is not ingestion. Source identification records the source class, steward, institutional role and jurisdiction.
    2. 03Register the sourceStepRegistration makes a discovered source a governed participant in the evidence lifecycle. It implies no endorsement or certification.
  2. 02 Assessment

    1. 04Jurisdiction and rights classificationGateRights classification identifies licence, consent, confidentiality and reuse limits. The result determines access and routing.
    2. 05Evidence quality assessmentStepQuality assessment decides whether a record is fit for a defined use. Quality is multidimensional and never a single score.
    3. 06Evidence labelStepEvery finding carries one of nine evidence labels. Missing data is shown as missing, never as zero.
  3. 03 Record

    1. 07Evidence classificationRecordClassification gives records lifecycle states, such as source-verified, public-safe or restricted. States can coexist and set evidence use.
    2. 08Source release and identifierRecord · NXS-GRX (canonical data, provenance and risk)A source release records rights, temporal coverage and source lineage. A UNI is allocated through a catalogue and never typed by hand.
    3. 09Digital Evidence PassportRecordThe passport records provenance, quality, rights, jurisdiction, permitted use and history, so evidence keeps its context when it travels.
  4. 04 Use and correction

    1. 02Correction reaches dependent recordsStep · NXS-GRX (canonical data, provenance and risk)A source correction propagates to dependent cuts and products without rewriting old results.
    2. 10Public-safe reviewGatePublic-safe review decides whether a record can be shared publicly. It is a distinct stage and is not validation.

Data protocols: evidence lifecycle · Data protocols: core data objects and evidence records · The Global Risks Index (NXS-GRX)Designed programme process, not a record of operation. No cycle has run and every programme is at service level R0. A gate is a programme decision, never a public-authority approval, certification or deployment authorisation. The data-protocol pages describe a design, not a running system.

05অনিশ্চয়তা

Quantifying what the model does not know.

Prediction intervals and calibration checks show when a model is outside its experience. Methods at GCRI uncertainty quantification; topic depth at explainability and fairness.

06How uncertainty is reported

How uncertainty is reported.

  1. 01

    State assumptions

    Each analysis lists its assumptions and the data it depends on.

  2. 02

    Quantify where possible

    Ranges and confidence are reported alongside central estimates.

  3. 03

    Validate and compare

    Models are checked against observations and, where possible, against other models.

  4. 04

    Publish limits

    What an analysis cannot show is stated beside what it can.

07উৎস

Where this page’s references lead.

The bodies and documents this page is read against. Status and editions change; each link goes to the issuing body.

08প্রশ্ন

Questions and answers.

Are the methods open?
Method descriptions are public on the STEM Commons. Code and data follow their own licences and access terms, listed in the source register.
প্রতিবেদনঅবদান রাখুন

এই পাতা থেকে প্রমাণ জমা দিন।

Reports হলো কনসোর্টিয়ামের ত্রৈমাসিক প্রকাশনা। এই বিষয়ে আপনি যা জানেন তা পরবর্তী সংস্করণে প্রমাণ, একটি দৃষ্টান্ত বা একটি সংশোধন হতে পারে, আপনার নামে স্বীকৃত এবং যেখানে তা আপনার নিজের কাজ সেখানে CC BY 4.0-এর অধীনে প্রকাশিত।

Global Risks AllianceAI and Financial Model RiskMethods and TechnicalExponential Technology

কখনো ব্যক্তিগত তথ্য, সংবেদনশীল অবকাঠামোর বিবরণ বা ভাগ করার অধিকার নেই এমন কিছু পাঠাবেন না। কোনো অবদান ব্যবহারের আগে পর্যালোচনা করা হয়; এটি পাঠালে লেখকত্ব বা সম্পাদকীয় অধিকার জন্মায় না, এবং পৃষ্ঠপোষকতা কখনো কোনো সিদ্ধান্ত কেনে না।

এআই ও আর্থিক মডেল ঝুঁকি

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.