iTutorGroup Tutor Application Screen

What this models

This example runs on the Intake-rule-class screen written above the hiring funnel network: 7 components and 17 pathways between them. Every context in the Lab is a stylized model, never a reconstruction of any actual deployment, and each assumption behind it carries a provenance label.

Evidence base: 3 assumed · 7 published baseline. In the Lab, the shaded evidence band behind each headline readout draws its width from the least-established class below.

• baseline

  This models the authored-intake-rule pattern documented in the case file, not a reconstruction of the actual software, and the register is the point. The EEOC's own releases describe tutor application software that was programmed to automatically reject; there is no model, no score, no training data and no learned proxy anywhere in this record. The widely repeated first-workplace-artificial-intelligence-settlement label is legal-press framing, and the EEOC's releases never use the term. The case's doctrinal value is that the regulator treated an automated screen as ordinary actionable age discrimination regardless of mechanism, which is why it resolved as intentional disparate treatment.

• baseline

  The conduct on this diagram is attributed, not adjudicated. The rejection rule, the count of more than 200 qualified United-States applicants rejected on age, and the resubmission narrative are the EEOC's allegations, resolved by a consent decree that carries no admission of liability; iTutorGroup denied the allegations and disputed that the tutors were employees at all, characterizing them as independent contractors. The $365,000 payment, the decree's obligations and the monitoring term are the decree's own terms and are stated as facts. The underlying conduct was never adjudicated.

• baseline

  The applicant-supplied form is drawn as its own component because the remedy acts on it rather than on any decision logic. The dossier enumerates exactly one external feed for this deployment: the application form fields, decisively the birthdate field and the applicant's sex. The consent decree enjoins requesting applicants' birth dates before an offer, which deletes the field the rule computed on from the intake form. Governance acting on the data-collection edge rather than the decision node is this deployment's distinguishing structural fact, and the form is where it lands.

• assumed

  The strongest pathway on the board is one authored configuration reaching every application, and it is drawn as a self-coupling on the intake screen. A deterministic threshold is perfectly correlated across cases by construction: where a learned screen's mistakes merely correlate, this one's are identical, which is why more than 200 rejections in a two-month window share a single cause rather than many. The reading is structural. The EEOC's complaint puts the rule's terms on the record; no error rate for this deployment has ever been published, and none is asserted here.

• baseline

  The read a person would take of the screen's decision before a rejection notice goes out is drawn empty, because that is what the sources say. The dossier's finding is that the record documents no human review point between the rule and the rejection notice: rejection was automatic at application intake and the employer-side reviewers saw only applicants who came through the screen. Drawing the pathway rather than omitting it keeps the missing check visible and priceable. What the public record does not identify is who authored the rule, why the thresholds differed by sex, or whether any internal review approved or missed it, so the pre-detection internal governance is modeled as undocumented rather than as affirmatively absent beyond what the no-review structure supports.

• baseline

  The pathway that carried the screen's decisions to the regulator is drawn faint because the record shows how it opened. Detection was external and applicant-powered: the charging party applied with her real birthdate and was rejected immediately, re-applied about a day later with an identical application except a more recent birthdate, and was offered an interview. One changed field, about a day, and no inside access — the lowest-capacity external audit channel there is. Roughly three and a half years then ran from the documented rejection window to an enforceable remedy.

• assumed

  Standing demand is read at the middle rung and the manual floor at the top, and both readings come from the same fact. The record documents more than 200 qualified applicants rejected on the rule in a two-month window, and it documents no queue, no backlog and no overwhelmed reviewer; total application volume is not public, so a strained reading would import a narrative the sources do not carry. The manual floor is read high because the automated screen sat on top of an intact human hiring process: the staff exercised real judgment over everyone who reached them, and the regulator's own theory is that the applicants the screen removed were qualified. The decree makes the same bet, aiming its training obligations at those involved in hiring tutors.

• baseline

  The decree era is drawn as structure because the decree built structure where the record documents none. A second record holds the charge, the pleadings, the decree's obligations and the written notices of discrimination complaints the decree routes to the EEOC, and a compliance function works from it: policy, internal memo, repeated trainings, and that reporting duty. The regulator's monitoring of that function is drawn as the board's firmest check, on the published term of at least the next five years or longer if iTutorGroup resumes hiring tutors in the United States. Part of the machinery has never had occasion to operate: because the company had ceased hiring United States tutors, the obligation to notify and interview the rejected applicants waits on a resumption the public record does not show.

• baseline

  The pressure applied to this board is about the regulator's channel and not about this case, and the distinction is carried on every surface. The decree instantiated a reporting-and-monitoring edge at the height of a proactive federal posture toward automated hiring; that posture then changed measurably, through the January 2025 removal of the EEOC's artificial-intelligence and algorithmic-fairness guidance, a loss of quorum from January 2025 until 7 October 2025, and a published 2026 agenda in which artificial intelligence in hiring does not appear. Executive Order 14281 of 23 April 2025 directs agencies to deprioritize enforcement involving disparate-impact liability; this decree rests on intentional disparate treatment, so that directive does not reach its theory, and nothing here suggests the decree is threatened. Removing guidance does not repeal the Age Discrimination in Employment Act, and no post-decree activity in this case was identified.

• assumed

  No applicant outcome is computed here. This Lab reads institutional propagation only, and the people this screen rejected are boundary-only: the count of more than 200, the $365,000 distributed among them, the per-claimant amounts that were never made public, and every question about who should have been hired live in the case file and are never derived from anything on this diagram. The count itself is a floor stated as more than 200, attributed to the EEOC's complaint, with no public denominator, so no rate can be computed from it either.

What this example does not show

• No applicant outcome is modeled here. This Lab reads institutional propagation only, and the people the screen rejected are boundary-only: the count of more than 200, the $365,000 distributed among them through a claims process, and the per-claimant amounts that were never made public live in the case file and are never computed on this diagram. The count is a floor attributed to the EEOC's complaint, and no public denominator exists, so no rate can be derived from it.
• The conduct is attributed, not adjudicated. The rejection rule, the count of rejected applicants and the resubmission narrative are the EEOC's allegations, resolved by a consent decree that carries no admission of liability. iTutorGroup denied the allegations and disputed that the tutors were employees at all, characterizing them as independent contractors. What is stated flatly here are the decree's own terms: the $365,000 payment, the obligations, and the monitoring period.
• This was a rule, not machine learning, and the popular label for the case is wrong about that. The EEOC's own releases describe application software programmed to automatically reject; there is no model, no score and no training data in the record. The first-workplace-artificial-intelligence-settlement framing belongs to the legal press. The case matters because a regulator treated an automated screen as ordinary actionable age discrimination regardless of mechanism, which is also why the sex-differentiated thresholds appear in the decree's injunctions while the suit itself was brought only under the Age Discrimination in Employment Act.
• The decree's term is conditional and part of its machinery has never operated. The published wording is monitoring for at least the next five years or longer if iTutorGroup resumes hiring tutors in the United States, which is not a flat five-year term. Because the company had already ceased hiring United States tutors when the decree was entered, the obligation to notify and interview the rejected applicants waits on a resumption the public record does not show happening.
• The pressure on this board is about the regulator and not about this case. The removal of the EEOC's artificial-intelligence guidance in January 2025, the loss of quorum until 7 October 2025, and Executive Order 14281's direction to deprioritize disparate-impact enforcement are documented changes in the agency's posture. This decree rests on intentional disparate treatment, so that directive does not reach its theory, no post-decree activity in this case was identified, and nothing here implies the decree is threatened. Removing guidance does not repeal the statute.
• The court filings themselves were not read. The joint notice of settlement with the proposed decree and the full docket returned HTTP 403 to every client attempted from the research environment, so every decree term asserted here is corroborated across the two EEOC releases and three law-firm analyses rather than read from the filed document. The exact date the decree was entered and the mechanics of the claims process are not independently verified and are not asserted.

Sources and evidence

What this example rests on, claim by claim. Every entry resolves to the same ledger the Evidence Registry publishes.

• The U.S. Equal Employment Opportunity Commission alleged that iTutorGroup, Inc., Shanghai Ping'An Intelligent Education Technology Co., Ltd. and Tutor Group Limited — three integrated companies providing English-language tutoring to students in China through United-States-based tutors working fully remotely from their homes — had programmed their tutor application software to automatically reject female applicants aged 55 or older and male applicants aged 60 or older, rejecting more than 200 qualified United-States applicants because of their age. After conciliation failed the Commission sued under the Age Discrimination in Employment Act on 5 May 2022, No. 1:22-cv-02565 (E.D.N.Y.); then-Chair Charlotte Burrows framed the filing for the agency's algorithmic-enforcement agenda with the words 'Even when technology automates the discrimination, the employer is still responsible.' The parties filed a joint notice of settlement with a proposed consent decree on 9 August 2023 and the court approved the decree in September 2023, announced by the Commission on 11 September 2023. iTutorGroup pays $365,000 to be distributed among the more than 200 rejected applicants through a claims process, without admitting liability; per-claimant amounts were not made public. iTutorGroup denied the allegations and disputed that the tutors were employees at all, characterizing them as independent contractors, a question the settlement resolved without adjudication.

  empirical
  • Government U.S. Equal Employment Opportunity Commission (2023, September 11). iTutorGroup to Pay $365,000 to Settle EEOC Discriminatory Hiring Suit (EEOC v. iTutorGroup, No. 1:22-cv-02565, E.D.N.Y.). https://www.eeoc.gov/newsroom/itutorgroup-pay-365000-settle-eeoc-discriminatory-hiring-suit
  • Government U.S. Equal Employment Opportunity Commission (2022, May 5). EEOC Sues iTutorGroup for Age Discrimination https://www.eeoc.gov/newsroom/eeoc-sues-itutorgroup-age-discrimination
  • Trade press HR Dive (2023, August 10). Tutoring firm settles claim alleging its recruiting algorithm screened out applicants over 60 https://www.hrdive.com/news/AI-age-bias-recruiting-eeoc-suit/690556/

• The screen at issue was an authored decision boundary rather than anything learned, and the distinction is the case's doctrinal value. The EEOC's own releases describe 'tutor application software' that was 'programmed to automatically reject female applicants aged 55 or older and male applicants aged 60 or older' — a deterministic, sex-differentiated threshold computed from the birthdate field collected on the application form, with no score, no training data, and no published error rate anywhere in the record. It sat at the top of the hiring funnel, upstream of any human reviewer: rejection was automatic at application intake, no human review point between the rule and the rejection notice is documented, and the employer's own hiring staff saw only applicants who came through the screen. Rejected applicants received no statement of the reason. The settlement was widely described in the legal press as the EEOC's first workplace artificial-intelligence settlement, a framing that belongs to that press and not to the agency, whose releases never use the term; the case's significance is that the regulator treated an automated screen as ordinary actionable age discrimination regardless of mechanism, which is why it proceeded as intentional disparate treatment. All of the conduct described here is the Commission's allegation, resolved by a decree carrying no admission of liability.

  empirical
  • Government U.S. Equal Employment Opportunity Commission (2023, September 11). iTutorGroup to Pay $365,000 to Settle EEOC Discriminatory Hiring Suit (EEOC v. iTutorGroup, No. 1:22-cv-02565, E.D.N.Y.). https://www.eeoc.gov/newsroom/itutorgroup-pay-365000-settle-eeoc-discriminatory-hiring-suit
  • Government U.S. Equal Employment Opportunity Commission (2022, May 5). EEOC Sues iTutorGroup for Age Discrimination https://www.eeoc.gov/newsroom/eeoc-sues-itutorgroup-age-discrimination
  • Trade press Greenberg Traurig LLP (2023). EEOC Secures First Workplace Artificial Intelligence Settlement (GT Alert) https://www.gtlaw.com/en/insights/2023/8/eeoc-secures-first-workplace-artificial-intelligence-settlement

• The practice surfaced through a single applicant's experiment on the input rather than through any internal control. The charging party applied with her real birthdate and was rejected immediately, then re-applied about a day later with an application identical in every respect except a more recent birthdate, and was offered an interview. That comparison required no inside access, one changed field, and roughly a day, and it is the only comparison of a rejection against a counterfactual anywhere in the public record. The decree's reapplication invitations cover applicants rejected in March and April 2020, fixing the documented rejection window at two months, and roughly three and a half years then ran from that window to an enforceable remedy: rejections March to April 2020, a charge and failed conciliation, suit on 5 May 2022, joint notice of settlement 9 August 2023, decree approved September 2023. No public record identifies who authored the rule, why the thresholds differed by sex, or whether any internal review approved or missed it.

  empirical
  • Trade press HR Dive (2023, August 10). Tutoring firm settles claim alleging its recruiting algorithm screened out applicants over 60 https://www.hrdive.com/news/AI-age-bias-recruiting-eeoc-suit/690556/
  • Trade press Greenberg Traurig LLP (2023). EEOC Secures First Workplace Artificial Intelligence Settlement (GT Alert) https://www.gtlaw.com/en/insights/2023/8/eeoc-secures-first-workplace-artificial-intelligence-settlement
  • Government U.S. Equal Employment Opportunity Commission (2023, September 11). iTutorGroup to Pay $365,000 to Settle EEOC Discriminatory Hiring Suit (EEOC v. iTutorGroup, No. 1:22-cv-02565, E.D.N.Y.). https://www.eeoc.gov/newsroom/itutorgroup-pay-365000-settle-eeoc-discriminatory-hiring-suit

• The consent decree acted on the rule's input and on the oversight structure rather than on any decision logic. Its terms: injunctions against hiring discrimination based on age or sex; a prohibition on requesting applicants' birth dates before an offer, which removes from the intake form the field the rule computed on; a new anti-discrimination policy and an internal memo; multiple anti-discrimination trainings for those involved in hiring tutors; written notice to the EEOC of discrimination complaints, converting a formerly internal channel into a regulator-visible one; invitations to the applicants rejected in March and April 2020 to reapply; and, because iTutorGroup had already ceased hiring tutors in the United States, an obligation to notify and interview those applicants if it resumes United States operations, so part of the decree's machinery has never had occasion to operate. The Commission's own release and Seyfarth's 2024 recap give the duration as monitoring compliance 'for at least the next five years or longer if iTutorGroup resumes hiring tutors in the United States' — a floor with a conditional tail, not a flat five-year term. No post-decree enforcement activity in this case was identified.

  empirical
  • Government U.S. Equal Employment Opportunity Commission (2023, September 11). iTutorGroup to Pay $365,000 to Settle EEOC Discriminatory Hiring Suit (EEOC v. iTutorGroup, No. 1:22-cv-02565, E.D.N.Y.). https://www.eeoc.gov/newsroom/itutorgroup-pay-365000-settle-eeoc-discriminatory-hiring-suit
  • Trade press Greenberg Traurig LLP (2023). EEOC Secures First Workplace Artificial Intelligence Settlement (GT Alert) https://www.gtlaw.com/en/insights/2023/8/eeoc-secures-first-workplace-artificial-intelligence-settlement
  • Trade press Seyfarth Shaw LLP (2024). EEOC-Initiated Litigation, 2024 Edition (EEOC v. iTutorGroup entry) https://www.content.seyfarth.com/publications/EEOC-Initiated-Litigation-2024/36/

• The federal enforcement channel this decree instantiated changed measurably after the decree was entered, and the change is about the agency rather than about this case. In January 2025 the EEOC removed its Artificial Intelligence and Algorithmic Fairness Initiative content along with its May 2023 Title VII technical assistance on artificial intelligence and its May 2022 guidance on the Americans with Disabilities Act, following Executive Order 14179; the Department of Labor and the Office of Federal Contract Compliance Programs made parallel removals. Removing guidance repeals no law: Title VII, the Age Discrimination in Employment Act, and the Uniform Guidelines on Employee Selection Procedures are unchanged, and no federal safe harbor was created. Executive Order 14281 of 23 April 2025 directs that 'All agencies shall deprioritize enforcement of all statutes and regulations to the extent they include disparate-impact liability' and orders a ninety-day evaluation of existing disparate-impact consent judgments; this decree rests on intentional disparate treatment, so that directive does not reach its theory, and nothing in the public record suggests the decree is threatened. The Commission lacked a quorum from January 2025 until 7 October 2025, when Commissioner Brittany Panuccio was confirmed; Andrea Lucas was designated Chair on 5 November 2025, and the restored commission's published 2026 priorities centre on investigations of diversity programmes, religious accommodation, and national-origin cases, with artificial intelligence in hiring absent from the stated agenda.

  empirical
  • Trade press Cooley LLP (2025, February 21). Gone but Not Forgotten: Federal Laws Still Apply Despite AI Guidance Disappearance Act; and National Law Review (2026, March 31). The Federal Government Quietly Removed Its AI Hiring Guidance. Four States Are Writing Their Own https://www.cooley.com/news/insight/2025/2025-02-21-gone-but-not-forgotten-federal-laws-still-apply-despite-guidance-disappearance-act
  • Government Executive Order 14281, Restoring Equality of Opportunity and Meritocracy (23 April 2025), Daily Compilation of Presidential Documents DCPD-202500515 https://www.govinfo.gov/content/pkg/DCPD-202500515/html/DCPD-202500515.htm
  • Trade press Holland & Knight LLP (2025, December 30). Back in Business: EEOC's Restored Quorum Explained and a Look Forward to 2026 https://www.hklaw.com/en/insights/publications/2025/12/back-in-business-eeocs-restored-quorum-explained

Where this connects

Institutional pressures in this domain

• Workload surge — Demand outruns staffing; per-case attention shrinks and review becomes triage.
• Vendor opacity — The deploying institution cannot inspect the model, data, or update pipeline it is accountable for.
• Compliance over substance — Paper controls (sign-offs, checklists) satisfy audits while the behavior they describe erodes.
• Data & policy drift — The world, the intake process, and the rules change under a system trained on how things used to be — two mechanisms with different remedies: the statistical properties of what the system processes move (concept drift), or the mixture of inputs arriving in deployment differs from the mixture it was trained on (covariate shift).
• Reviewer bottleneck — One fixed-capacity checking stage sits between AI output and consequence; everything queues behind it.

All of them in context on the Hiring & employment screening AI domain page.

Levers available here and the patterns behind them

• Verify output — Put a verifier on the agent
• Gate record entries — Human-in-the-loop write gating
• Mark AI-written records — Provenance labeling
• Understand the system — Understand the system
• Store less data — Data minimization
• Review on schedule — Oversight cadence & retrospectives
• Assign a challenger — Structured dissent

Documented case histories

• iTutorGroup Tutor Application Screen
• A resume screener that learned the past's bias
• Vendor screening across thousands of employers (litigation live)
• Graduate-hiring AI with its audits on the record
• HireVue video assessment (vendor layer)
• The 1959 statute and the integrity video screen (Baker v. CVS Health)
• An internal promotion, a recorded screen, and a captioning request (D.K. charges against Intuit and HireVue)
• Aon pre-hire assessment suite (vendor's own tables)
• The cooperative audit: a paid source-code examination, and what happened to its verdict
• McHire and the 64-million-record custody exposure
• SiriusXM's iCIMS applicant screening
• Checkr gig-economy background screening
• The rule with no number to disclose
• The account goes dark at nine; the reason arrives on day twenty-six
• Meta Job-Ad Delivery: the guardrail and the layer below