Ikibazo gihuza ibibazo · Hagati Urwego

Rotterdam Welfare Fraud Risk Algorithm Quiz

Checks the Rotterdam fraud-risk model’s human-review role, official audit limits, proxy risks, model-test scope and current-status distinctions.

Ikibazo 1 Bya 8

Ni uruhe ruhare Analytics ya Rotterdam Uitkeringsfraude yerekana ibyago byagize?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Ni uruhe ruhare Analytics ya Rotterdam Uitkeringsfraude yerekana ibyago byagize?
    • It automatically convicted recipients of fraud
    • It helped select recipients for a follow-up review
    • It set benefit amounts nationally
    • It replaced all municipal caseworkers
  2. Ni iki Urukiko rw'Ubugenzuzi bwa Rotterdam rwanzuye ku byerekeye ivangura rishingiye ku ivangura mu isuzuma ryarwo 2021?
    • It proved every flagged person was treated unlawfully
    • It certified the model as bias-free
    • It warned about proxy-bias risk but did not determine that model outcomes were discriminatory
    • It found no transparency problems
  3. Kuki inyandiko ziperereza zamateka zishobora kuba intego yo guhugura?
    • They may reflect past selection practices as well as underlying noncompliance
    • They cannot contain administrative data
    • They are always a random sample of all recipients
    • They eliminate human judgment
  4. Ni izihe mpungenge za prokisi 2021 ubugenzuzi nigisubizo cya leta byaganiriye?
    • Proxy risk applies only to face recognition
    • Language or literacy information could correlate with origin or another protected characteristic
    • Nationality was the only input and no proxy concern existed
    • Postal codes legally prove fraud
  5. Niki gishobora kugenzurwa ibizamini by'icyitegererezo cyagaruwe bishobora gushiraho?
    • The exact real-world error rate for all Rotterdam recipients
    • A judicial conclusion about discrimination
    • Whether every investigated person committed fraud
    • How selected input changes affect scores for that artifact and test profiles
  6. Who made the final assessment of benefit legality under the municipality’s 2021 description?
    • A municipal employee after reviewing the case
    • A national court for every invitation
    • The risk score alone
    • A private credit bureau
  7. What does the public algorithm-register entry document, and what remains unproven?
    • A successor model completed in 2023 and still operating
    • The recovered code as a certified copy of the register record
    • A selection algorithm dated January 2017–February 2022 and stopped in early 2022; its exact link to the recovered artifact is unproven
    • A current national system that decides benefit eligibility
  8. Which conclusion best matches the official evidence about the case?
    • The later register proves the older system remains active
    • The Rotterdam model was invalidated in the separate SyRI judgment
    • Removing protected fields guarantees fair referrals
    • Opacity and proxy risks warranted scrutiny, but the 2021 audit did not measure whether outputs were discriminatory