IBI-002163_significativepublie
Differential Calibration Bias
Biais de calibration différentielle
IBI Score
6.5
out of 10
Severity3_significative
Frequencyfrequent
Scope groups4
Detection difficulty3
🇫🇷 Définition (FR)
Probabilités prédites ne correspondant pas aux fréquences observées de façon différenciée.
🇬🇧 Definition (EN)
Predicted probabilities not matching observed frequencies differentially across groups.
TriAxis Taxonomy
Structural Classification
O
Origin
O2
Computational & Statistical
L
Lifecycle
L6
Evaluation
S
Specific
S18
Group attribution bias
💡
Concrete Example
COMPAS (ProPublica 2016) : 2× plus susceptible de classer faussement une personne noire comme à haut risque.
📊 Detection Metrics
Demographic Parity DifferenceDisparate Impact RatioEqualized Odds Difference
⚙ Alert Thresholds
Disparate Impact Ratio0.8
Equalized Odds Difference0.1
Demographic Parity Difference0.1
📚 Sources & Scientific References
- •Dissecting racial bias in an algorithm used to manage the health of populationsZiad Obermeyer, Brian Powers, Christine Vogeli, Sendhil Mullainathan (2019)
- •Fair prediction with disparate impact: A study of bias in recidivism prediction instrumentsAlexandra Chouldechova (2017)
🛡 Impact & Mitigation
⚠ Documented evidence (Studies & Incidents)
- Outil de recrutement expérimental d'Amazon abandonné pour biais envers les femmes : ProPublica analyse 7 000 dossiers COMPAS : l'algorithme classe 2× plus souvent faussement les accusés noirs.
Impact: Faux positifs : 44,9% (noirs) vs 23,5% (blancs). - ProPublica report: recidivism risk models are racially biased (billet d'ORCAA commentant l'enqu�te � Machine Bias � de ProPublica sur COMPAS) : Source : https://orcaarisk.com/articles/2016/10/12/propublica-report-recidivism-risk-models-are-racially-biased
Impact:
🎯 Vulnerable groups impacted
Black / Afro-descendantWorking classLGBTQ+ personsWomenMinority language speakersArab / MENA personsAsian personsMotor disabilitySeniors (60+)Youth (18-25)Non-binary personsRural areas
🛠 Mitigation methodologies
- AUDIT_TIERS : Third-party Audit
🏢 Sectors concerned
Finance & Banking (Risk: /5)Housing & Real Estate (Risk: /5)HR & Recruitment (Risk: /5)Justice & Security (Risk: /5)Insurance (Risk: /5)
⚖ Applicable regulations
🇪🇺 Europe
- •AI_ACT_10 : AI Act (UE) 2024/1689
- •AI_ACT_9 : AI Act (UE) 2024/1689
- •RGPD_22 : RGPD (UE) 2016/679
🇺🇸 United States
- •US_TX_TRAIGA : Texas Responsible AI Governance Act (HB 149)
🌍 International
- •ISO_42001 : ISO/IEC 42001:2023
- •UNESCO_2021 : UNESCO - Recommandation sur l'éthique de l'IA
- •OECD_AI_PRINCIPLES : OCDE - Principes IA
- •COE_AI_CONVENTION_INT : Conseil de l'Europe - Convention IA
🇫🇷 France
- •FR_CNIL_AI : CNIL - Recommandations IA
🌐 Canada
- •CA_AIDA : Canada - Loi sur l'IA et les données (AIDA)
📋 Metadata
Contributor
Mohamadou SIDIBE
Institution
AI BIAIS INITIATIVE
Version
1
Created on
October 14, 2026
Updated on
November 4, 2026
UUID
04b57d38-46c4-4e59-91c1-b30d10c8b847
Explore other biases
250 documented biases, classified by the TriAxis taxonomy.