Algorithmic Decision Systems — Detecting and Challenging Automated Decisions

Automated decision systems — algorithms, risk scores, predictive models — now determine whether you are approved for a loan, offered an apartment, called for a job interview, or classified as a flight risk in a criminal proceeding. These systems process data about you and output a decision or recommendation, often without any human reviewing your individual situation. In many cases, the institution making the decision does not fully understand the algorithm it is using.

The practical consequences are significant. Errors in your underlying data — an incorrect address, a misattributed debt, a criminal record belonging to someone with a similar name — can propagate through multiple systems and produce cascading adverse decisions across credit, housing, and employment simultaneously. Discriminatory patterns in training data can produce outputs that disadvantage protected classes in ways the operator cannot explain. And because the decision appears to come from an objective system, it is often harder to challenge than a decision made by an identifiable individual.

This tutorial covers how to identify whether an automated system was involved in a decision, how to obtain the underlying data, and the legal tools available to dispute errors and request human review.

Steps

1. Identify whether the decision was automated. Signs that an algorithm was involved: decisions are made unusually quickly, within minutes or hours of your application; the denial reason is expressed as a category code or score rather than a specific explanation; the language of the denial is identical to many other reported denials; or you are told by a representative that “the system” made the decision and they cannot override it. In credit decisions, the adverse action notice itself is required to state whether a credit score was used. In employment, many applicant tracking systems screen resumes algorithmically before a human sees them.

2. Request the adverse action notice if you didn’t receive one. When a creditor, insurer, or employer takes adverse action based in whole or in part on a consumer report or credit score, the Fair Credit Reporting Act (FCRA) requires them to provide a written notice stating: the name and contact information of the consumer reporting agency used; the score or report that contributed to the decision; and the specific reason codes for the adverse action. If you did not receive this notice, request it in writing. Send your request certified mail with return receipt. Document the date sent and any response received.

3. Obtain your underlying data file from every relevant source. The decision may have been based on data from a credit bureau (Equifax, Experian, TransUnion), a specialty consumer reporting agency (LexisNexis Risk Solutions for insurance; Checkr or Accurate for background checks; CoreLogic for tenant screening; Certegy for banking), or an internal scoring model. You are entitled to a free copy of your file from any consumer reporting agency that provided data used in an adverse decision — the adverse action notice must identify which agency was used. Request your free annual credit reports at annualcreditreport.com. For specialty agencies, request your file directly from each agency by name.

4. Identify errors and dispute them with the data furnisher. Review every entry in your file for accuracy: name, address history, employer, account status, payment history, public records, and any items that do not belong to you — a common problem when someone with a similar name has a criminal record or outstanding debt. The FCRA requires consumer reporting agencies to investigate disputes within 30 days (or 45 days if you provide additional information). File disputes directly with the reporting agency, and separately with the company that furnished the data — the original creditor, the court, the landlord. Both are required to correct verified errors. Document every dispute in writing and keep copies of everything you send and receive.

5. Request human review of the decision. Under the FCRA, Equal Credit Opportunity Act (ECOA), and Fair Housing Act, you have the right to receive specific reasons for an adverse decision and, in some contexts, to request human review of a decision made by an automated system. This right is not always clearly stated by institutions — you may need to assert it explicitly. Write to the institution: “I am requesting human review of this decision pursuant to [applicable law]. Please identify the individual responsible for reviewing my file and provide their contact information.” Send certified mail. The institution’s response — or failure to respond — is itself part of your record.

6. Escalate to regulatory bodies. If errors are not corrected within the required timeframe, or if you believe a decision reflects illegal discrimination, file a complaint with the relevant regulator: the Consumer Financial Protection Bureau (CFPB) for credit and financial decisions; the Department of Housing and Urban Development (HUD) for housing discrimination; the Equal Employment Opportunity Commission (EEOC) for employment discrimination; or your state attorney general for state consumer protection violations. Regulatory complaints trigger mandatory institutional responses and create a documented record. Many institutions respond to regulatory complaints significantly faster than to direct consumer disputes.

7. Document the appeal process as an evidence trail. Every letter you send, every response you receive, every phone call you make — date, time, representative name, and what was said — is part of the record. If you later pursue legal action (a private right of action is available under the FCRA for willful noncompliance), your documentation is the evidentiary foundation. Apply the same discipline as you would to any institutional dispute: contemporaneous notes, preserved originals, dated copies stored in a location the institution cannot access.

Common Mistakes

• Assuming a human reviewed your file. In most automated decision contexts — credit, tenant screening, resume screening — no human saw your individual file before the decision was made. Don’t frame your dispute around persuading a reviewer who may not exist; focus on correcting the underlying data that fed the system.
• Disputing the decision before correcting the underlying data. If the data used to make the decision is wrong, fix the data first. Disputing the decision without correcting the source data often results in a reinvestigation that returns the same result, using the same incorrect inputs.
• Missing dispute windows. The FCRA requires consumer reporting agencies to investigate disputes within 30 days. Some state laws have different timelines. Courts and administrative bodies have filing deadlines for discrimination claims — often 180 or 300 days from the adverse action. Know your deadlines before you begin.
• Treating the process as adversarial from the start. Many adverse decisions result from data errors, not intentional discrimination. A clear, documented dispute focused on specific errors often resolves faster than a broad complaint. Escalate to regulatory bodies when direct dispute fails — not as a first step.
• Accepting “that’s our policy” as a final answer. Institutional representatives often don’t know the legal rights of the people they are speaking with. “Our system made the decision” is not a legal justification for an adverse action. Your right to human review, adverse action notice, and data access exists regardless of the institution’s stated internal policy.

Resources

• Documenting Interactions with Authorities — how to maintain a usable record through a dispute process
• AI Surveillance and Reducing Your Exposure — how data about you is collected and sold
• Survival Resources: Legal — legal aid and advocacy organizations
• Glossary — definitions for technical terms used in this tutorial