I85PREP
⌘ K
CDEP study materials
CDEP study resources

CDEP practice questions

One question at a time. Build your CDEP understanding.

First 50 of 200 questions. Each one lists the choices, the correct answer, and an explanation. The checker at the top of this page saves progress in this browser when it is available.

Continue in the practice tool above to load more questions, 50 at a time.

Question 1Data Ethics Foundations

A data ethics review board is defining what 'informed consent' means for a new analytics platform. Which condition is essential for consent to be considered valid?

  1. The participant must be offered financial compensation for their data.
  2. The participant must sign a paper document rather than consent digitally.
  3. The participant must understand the purpose, risks, and scope of data use before agreeing. Correct answer.
  4. The participant must waive all future rights to withdraw their data.

Explanation

Valid informed consent requires that the person understands what they are agreeing to, including purpose, risks, and scope of use, and agrees voluntarily. Compensation, format of signature, and waiver of withdrawal rights are not defining conditions and can even undermine voluntariness.

Question 2Data Ethics Foundations

Which principle of data ethics is most directly concerned with collecting only the data that is actually needed for a stated purpose?

  1. Data portability
  2. Data redundancy
  3. Data monetization
  4. Data minimization Correct answer.

Explanation

Data minimization holds that only data adequate, relevant, and necessary for the specified purpose should be collected and processed. Portability concerns transferring data between services, while monetization and redundancy are not ethical principles in this sense.

Question 3Data Ethics Foundations

An organization wants to reuse customer data collected for billing to train a marketing recommendation model. Under purpose limitation, what is the most appropriate first step?

  1. Assess whether the new purpose is compatible and obtain fresh consent or a lawful basis if not. Correct answer.
  2. Proceed only if the model is trained on anonymized billing amounts.
  3. Delete the billing records immediately to avoid any conflict.
  4. Proceed because the data already exists in the company's systems.

Explanation

Purpose limitation requires that data be used only for purposes compatible with those originally specified. Reuse for a materially different purpose such as marketing model training needs a compatibility assessment and, where required, a new lawful basis or consent.

Question 4Data Ethics Foundations

A team publishes a dataset with names and addresses removed but leaves precise geolocation timestamps. What ethical risk remains?

  1. Re-identification of individuals through linkage and inference from location patterns. Correct answer.
  2. Loss of data utility making the dataset worthless.
  3. None, because direct identifiers have been removed.
  4. Increased storage costs for the publisher.

Explanation

Removing direct identifiers does not guarantee anonymity; precise location and time data can be linked with other sources or used to infer identity and sensitive habits. This is the classic re-identification risk that de-identification alone does not eliminate.

Question 5Data Ethics Foundations

Which statement best captures the ethical concept of 'algorithmic fairness'?

  1. An algorithm's outputs do not create unjustified disparate treatment or impact across groups. Correct answer.
  2. An algorithm produces the same output for every input.
  3. An algorithm is written by a diverse team of engineers.
  4. An algorithm runs in constant time regardless of dataset size.

Explanation

Algorithmic fairness concerns whether a system's outputs distribute benefits and harms justly and avoid unjustified disparities across groups. Identical outputs, runtime complexity, and team composition do not by themselves establish fair outcomes.

Question 6Data Ethics Foundations

A hiring model historically trained on past decisions systematically rejects qualified applicants from one demographic group. Which response best reflects an ethical remediation approach?

  1. Remove the protected attribute from the feature set and redeploy unchanged.
  2. Deploy the model and monitor rejection rates after launch.
  3. Increase the model's overall accuracy on the training set.
  4. Audit the training data and outcomes, then correct the model and validate disparities before deployment. Correct answer.

Explanation

Ethical remediation requires investigating how historical bias entered the data and outcomes, correcting the model, and validating that disparities are reduced before deployment. Simply dropping a protected attribute or improving aggregate accuracy does not address the underlying disparity.

Question 7Data Ethics Foundations

What does 'transparency' primarily require of an organization that deploys automated decision systems?

  1. Publishing all internal meeting minutes related to the project.
  2. Guaranteeing that no errors will occur in automated decisions.
  3. Providing meaningful information about how decisions are made and the role of automation. Correct answer.
  4. Disclosing the source code of every model publicly.

Explanation

Transparency means giving affected people and regulators meaningful insight into the logic, role, and consequences of automated decisions. It does not require open-sourcing code, publishing all internal records, or promising error-free operation.

Question 8Data Ethics Foundations

A hospital uses an AI triage tool whose recommendations clinicians cannot override. Which ethical concern is most acute?

  1. Excessive transparency toward patients.
  2. Loss of meaningful human control and accountability. Correct answer.
  3. Reduced electricity consumption of the system.
  4. Overly frequent model retraining.

Explanation

When humans cannot override automated recommendations, accountability becomes unclear and harmful errors cannot be corrected in context. Meaningful human oversight is a core safeguard in high-stakes automated decision-making.

Question 9Data Ethics Foundations

Which practice best embodies the ethical principle of accountability in data governance?

  1. Collecting user feedback through optional surveys.
  2. Assigning clear ownership and documented responsibility for data decisions and their impacts. Correct answer.
  3. Storing data in multiple locations to prevent loss.
  4. Encrypting all data at rest and in transit.

Explanation

Accountability requires identifiable owners and documented responsibility for how data is used and who answers for its effects. Redundancy, encryption, and surveys are useful controls but do not by themselves establish accountability.

Question 10Data Ethics Foundations

A research team plans to share a dataset containing sensitive health information with external partners. Which combination of measures most directly supports ethical data sharing?

  1. Public posting of the raw dataset to encourage transparency.
  2. Aggressive marketing of the dataset to maximize reuse.
  3. Data-use agreements, access controls, and clear limits on onward sharing. Correct answer.
  4. Removing the dataset's documentation to protect confidentiality.

Explanation

Ethical sharing of sensitive data relies on enforceable agreements, controlled access, and restrictions on further distribution. Public posting of raw sensitive data or removing documentation undermines both privacy and responsible use.

Question 11Data Ethics Foundations

What is the primary ethical purpose of conducting a data protection impact assessment before launching a high-risk processing activity?

  1. To replace the need for any ongoing monitoring.
  2. To identify and mitigate risks to individuals before harms occur. Correct answer.
  3. To calculate the financial return on the data project.
  4. To satisfy marketing requirements for the product launch.

Explanation

A data protection impact assessment is a proactive tool to identify, evaluate, and reduce risks to individuals before high-risk processing begins. It does not serve marketing or finance goals and does not remove the need for continued monitoring.

Question 12Data Ethics Foundations

An analyst notices that a model's error rates are much higher for users with disabilities. Which action is most consistent with ethical data practice?

  1. Remove disability-related records from the dataset entirely.
  2. Report the disparity only after a public complaint.
  3. Ignore the disparity because overall accuracy is acceptable.
  4. Document the disparity, investigate its causes, and mitigate before or during deployment. Correct answer.

Explanation

Disparate error rates signal potential harm to a vulnerable group and must be documented, investigated, and mitigated. Ignoring the gap, deleting records, or waiting for complaints all fail the duty to prevent foreseeable harm.

Question 13Data Ethics Foundations

Which scenario best illustrates a conflict of interest in data ethics decision-making?

  1. A reviewer asks for additional documentation before approving a project.
  2. A reviewer consults external ethics guidelines when evaluating a proposal.
  3. A reviewer who owns equity in the vendor supplying the model approves its deployment. Correct answer.
  4. A reviewer evaluates a project with no connection to its sponsors.

Explanation

A conflict of interest arises when a decision-maker has a personal financial stake in the outcome, compromising impartiality. Independent review, requests for more evidence, and use of guidelines all support objectivity rather than undermine it.

Question 14Data Ethics Foundations

A company wants to sell aggregated user data to advertisers. Which question is most central to the ethical evaluation?

  1. Whether users were informed and whether the sale is consistent with the original consent. Correct answer.
  2. Whether the aggregation was performed with a specific software library.
  3. Whether the company's servers are located in a particular country.
  4. Whether the dataset is compressed to save storage.

Explanation

The ethical core of secondary commercial use is whether it falls within what users were told and agreed to, and whether they were adequately informed. Technical details like libraries, server location, and compression do not determine whether the use respects users' expectations.

Question 15Data Ethics Foundations

What does the principle of 'data quality' contribute to ethical data practice?

  1. It guarantees that data will never contain errors.
  2. It requires that all data be stored indefinitely.
  3. It mandates that data be collected from a single source.
  4. It ensures data is accurate, relevant, and fit for the purpose it is used for. Correct answer.

Explanation

Data quality in an ethics framework means data is accurate, relevant, and appropriate for its intended use, reducing harm from bad decisions. It does not promise error-free data, indefinite retention, or single-source collection.

Question 16Data Ethics Foundations

A public-sector body plans to publish statistics derived from citizen data. Which approach best balances openness with privacy?

  1. Withhold all statistics to avoid any privacy risk.
  2. Publish individual-level records to maximize transparency.
  3. Apply statistical disclosure control and publish aggregate results. Correct answer.
  4. Publish records with names removed but leave all other fields intact.

Explanation

Statistical disclosure control reduces the risk of re-identification while still enabling public benefit from aggregate results. Publishing individual records or merely stripping names leaves serious re-identification risk, while withholding everything sacrifices legitimate openness.

Question 17Data Ethics Foundations

Which of the following best describes the ethical duty of 'data stewardship'?

  1. Maximizing the volume of data collected for future opportunities.
  2. Restricting data access to a single individual permanently.
  3. Managing data responsibly on behalf of others with care for their interests. Correct answer.
  4. Transferring all data responsibilities to a third-party vendor.

Explanation

Data stewardship is the responsible management of data on behalf of others, prioritizing their interests and the common good. It is not about hoarding data, outsourcing responsibility entirely, or concentrating access in one person.

Question 18Data Ethics Foundations

A team discovers that a vendor's model was trained on data scraped without permission. What is the most defensible ethical response?

  1. Rebrand the model to avoid association with the vendor.
  2. Deploy the model while negotiating a retroactive license.
  3. Continue using the model but avoid mentioning the data source.
  4. Suspend use, assess the impact, and seek a lawful and ethical alternative. Correct answer.

Explanation

Continuing to benefit from improperly obtained data compounds the original harm. The defensible path is to pause use, evaluate consequences, and obtain a lawful, ethically sourced alternative. Concealment, retroactive licensing during use, and rebranding do not remedy the harm.

Question 19Data Ethics Foundations

Which statement about anonymization versus pseudonymization is accurate?

  1. Anonymized data always retains a re-identification key held by the controller.
  2. Pseudonymization and anonymization are identical processes with different names.
  3. Anonymization requires encryption, while pseudonymization forbids it.
  4. Pseudonymized data can often be re-linked to individuals with additional information, while properly anonymized data cannot. Correct answer.

Explanation

Pseudonymization replaces identifiers but keeps a means of re-linking, whereas anonymization aims to make re-identification impossible. The two are distinct, and anonymized data does not retain a controller-held re-identification key.

Question 20Data Ethics Foundations

A model is used to allocate social benefits. Which feature of the system most demands a documented appeals process?

  1. The model runs on cloud infrastructure.
  2. The model was developed by a small team.
  3. The model uses a modern programming language.
  4. The decisions materially affect people's access to essential resources. Correct answer.

Explanation

When automated decisions affect access to essential benefits, affected people need a way to contest errors and seek review. Programming language, hosting, and team size do not determine the need for appeals.

Question 21Data Ethics Foundations

Which behavior best reflects 'privacy by design'?

  1. Adding privacy features after a product has launched.
  2. Building privacy protections into the system architecture from the outset. Correct answer.
  3. Relying on users to configure all privacy settings themselves.
  4. Collecting maximum data to allow future privacy features.

Explanation

Privacy by design embeds privacy protections into systems from the beginning rather than retrofitting them. It does not offload responsibility entirely to users or justify excessive collection.

Question 22Data Ethics Foundations

An ethics committee must decide whether to approve a study using data from a vulnerable population. Which consideration is most important?

  1. Whether the study will produce statistically significant results.
  2. Whether the data can be sold to a commercial partner.
  3. Whether additional safeguards protect participants and whether the burden is justified. Correct answer.
  4. Whether the study can be completed within the budget.

Explanation

Research involving vulnerable populations requires extra safeguards and a justified balance of burden and benefit. Statistical significance, budget, and commercial resale are not the primary ethical criteria.

Question 23Data Ethics Foundations

What distinguishes 'explainability' from mere 'transparency' in automated decision-making?

  1. Explainability is only relevant to systems with no human involvement.
  2. Explainability focuses on making the reasons for specific outputs understandable to affected people. Correct answer.
  3. Explainability means the system never makes mistakes.
  4. Explainability requires publishing all training data publicly.

Explanation

Explainability concerns whether the reasons behind particular decisions can be communicated understandably to those affected. It does not require publishing training data, guarantee correctness, or apply only to fully automated systems.

Question 24Data Ethics Foundations

Which is the best example of a 'dark pattern' relevant to data ethics?

  1. A settings page that explains each data use.
  2. A privacy notice written in plain language.
  3. A design that makes rejecting data collection far harder than accepting it. Correct answer.
  4. A clear consent screen with equally prominent accept and reject buttons.

Explanation

Dark patterns manipulate interface design to steer users toward choices that may not reflect their true preferences, such as making refusal difficult. Clear, balanced, and plain-language designs support genuine choice rather than subvert it.

Question 25Data Ethics Foundations

A data ethics policy states that decisions should be 'contestable'. What does this most directly require?

  1. That every automated decision be reversed automatically.
  2. That decisions never be revisited once issued.
  3. That affected individuals have a practical way to challenge decisions and obtain review. Correct answer.
  4. That all decisions be made by committees rather than software.

Explanation

Contestability means people affected by a decision have a practical route to challenge it and have it reviewed. It does not require automatic reversal, a ban on software decisions, or finality without review.

Question 26Data Ethics Foundations

Which ethical principle is most directly violated when an organization collects email addresses for a newsletter and then sells them to third-party brokers without informing subscribers?

  1. Data minimization and storage limitation
  2. Purpose limitation and informed consent Correct answer.
  3. Accountability and audit logging
  4. Accuracy and completeness

Explanation

Consent-based collection ties data to a stated purpose, and repurposing it for sale exceeds that purpose. Purpose limitation requires that data be used only for the purposes disclosed at collection, and informed consent requires that the new use be communicated and agreed to.

Question 27Data Ethics Foundations

In data ethics, what does the principle of 'fairness' most commonly require of a data-driven system?

  1. That comparable people are treated comparably and unjustified disparate impacts are identified and addressed Correct answer.
  2. That the system maximizes overall accuracy on the training set
  3. That all users receive identical outputs regardless of context
  4. That the system runs on hardware available to all users

Explanation

Fairness concerns the distribution of benefits and harms across groups, not uniformity of output. It asks whether differences in treatment or performance are justified and whether unjustified disparate impact has been detected and remedied.

Question 28Data Ethics Foundations

A retail analytics team wants to infer customers' pregnancy status from purchase histories to target promotions. Which ethical consideration is most directly raised by this practice?

  1. Whether the promotion copy uses accessible fonts
  2. Whether the inference is stored in a relational database
  3. Whether inferring sensitive attributes without awareness or consent is a justified intrusion Correct answer.
  4. Whether the model was written in an open-source language

Explanation

Inferences about sensitive personal attributes such as health or pregnancy status raise privacy and autonomy concerns even when the underlying data was lawfully collected. The ethical question is whether such inference is proportionate, disclosed, and consented to.

Question 29Data Ethics Foundations

Which action best demonstrates 'data minimization' at the point of collection?

  1. Collecting only the fields needed for the stated purpose and justifying each one Correct answer.
  2. Encrypting all collected fields at rest
  3. Collecting every field the form allows so future needs are covered
  4. Retaining all collected data indefinitely to preserve optionality

Explanation

Data minimization limits collection to what is adequate, relevant, and necessary for the specified purpose. Justifying each field forces the team to connect collection to a legitimate need rather than speculative future use.

Question 30Data Ethics Foundations

An organization publishes a model card describing a deployed model's intended use, training data sources, and measured performance across subgroups. Which ethical function does this primarily serve?

  1. It transfers legal liability for model errors to the reader
  2. It replaces the need for ongoing monitoring
  3. It provides documented transparency that supports scrutiny by users and auditors Correct answer.
  4. It guarantees the model is free of bias

Explanation

Documentation such as model cards makes the system's scope and limitations visible so that others can evaluate and challenge it. It supports transparency and accountability but cannot by itself eliminate bias or substitute for monitoring.

Question 31Data Ethics Foundations

Under the principle of storage limitation, what should guide how long a dataset is retained?

  1. The maximum capacity of the storage system
  2. The preference of the individual engineer who built the pipeline
  3. A defined retention period tied to the purpose, after which data is deleted or de-identified Correct answer.
  4. The default setting of the database vendor

Explanation

Storage limitation requires that personal data be kept no longer than necessary for the purpose for which it was collected. Retention periods should be defined, documented, and enforced with deletion or de-identification at the end.

Question 32Data Ethics Foundations

A team wants to reuse a clinical dataset for a new research question unrelated to the original consent. What is the most defensible first step?

  1. Check whether the original consent or an ethics review permits the new use, and seek re-consent or approval if not Correct answer.
  2. Delete the original consent records to avoid confusion
  3. Aggregate the data first so individual records cannot be seen
  4. Proceed because the data is already in the organization's possession

Explanation

Secondary use of data is governed by the scope of the original consent and any applicable ethics oversight. Where the new purpose falls outside that scope, re-consent or formal ethics approval is required before reuse.

Question 33Data Ethics Foundations

Which statement best captures the ethical difference between 'privacy' and 'confidentiality' in data handling?

  1. Privacy applies only to governments, while confidentiality applies only to companies
  2. Confidentiality concerns consent, while privacy concerns encryption
  3. Privacy concerns an individual's control over personal information, while confidentiality concerns limiting access to information entrusted to others Correct answer.
  4. Privacy and confidentiality are synonyms with no meaningful distinction

Explanation

Privacy is about a person's ability to control collection, use, and disclosure of their information. Confidentiality is a duty of those who hold information to prevent unauthorized access, which can apply to data that is not strictly personal.

Question 34Data Ethics Foundations

A dataset used to train a credit model contains only records of applicants who were previously approved. Which ethical problem does this sampling create?

  1. The data reflects historical selection decisions, so the model may learn and reproduce past exclusion Correct answer.
  2. The dataset is too small to train any model
  3. The dataset contains too many protected attributes
  4. The dataset violates storage limitation

Explanation

Selecting only approved applicants creates selection bias: the model learns from a population shaped by prior decisions rather than from all applicants. It can therefore reproduce and amplify historical exclusion.

Question 35Data Ethics Foundations

Which practice most directly supports the ethical principle of accountability for a data pipeline?

  1. Increasing the pipeline's throughput
  2. Encrypting the pipeline's network traffic
  3. Renaming the pipeline to reflect its purpose
  4. Assigning a named owner, documented decisions, and periodic review of the pipeline Correct answer.

Explanation

Accountability requires that someone is answerable for the system and that decisions are recorded and reviewed. Named ownership plus documentation and periodic review creates a traceable chain of responsibility.

Question 36Data Ethics Foundations

An analyst is asked to remove records showing that a model performs poorly for older users before a report is shared with leadership. What is the most ethically defensible response?

  1. Publish the full dataset publicly to expose the problem
  2. Refuse to suppress the findings and escalate the request through the appropriate governance channel Correct answer.
  3. Comply, because leadership asked and the analyst is subordinate
  4. Remove the records but mention the issue verbally to a colleague

Explanation

Suppressing adverse performance evidence undermines transparency and accountability, and it harms the affected group. The defensible path is to decline and escalate through governance so the issue is addressed rather than hidden.

Question 37Data Ethics Foundations

Which situation best illustrates an ethical tension between individual privacy and a legitimate public interest?

  1. A hospital encrypts its patient database
  2. A public health agency wants population-level mobility data to track an outbreak while individuals could be re-identified Correct answer.
  3. A team documents its data sources
  4. A company updates its privacy notice

Explanation

Public health surveillance can serve a strong collective interest, yet granular mobility data carries re-identification risk for individuals. The tension is between the benefit of population-level insight and the privacy of the people whose data produces it.

Question 38Data Ethics Foundations

What does 'informed consent' require beyond obtaining a signature on a form?

  1. That the person understands the purpose, risks, and scope of data use and can freely decline Correct answer.
  2. That consent is collected once and never revisited
  3. That the organization's legal team approves the wording
  4. That the form is stored in a secure location

Explanation

Informed consent is substantive: the person must understand what will happen to their data, the associated risks, and the scope of use, and must be able to refuse without detriment. A signature alone does not establish that understanding or voluntariness.

Question 39Data Ethics Foundations

A team plans to deploy a facial recognition system in public spaces. Which combination of ethical safeguards is most appropriate before deployment?

  1. A pilot with no disclosure to reduce public concern
  2. A documented necessity and proportionality assessment, an impact assessment covering privacy and bias, and a mechanism for public input Correct answer.
  3. A public notice on a website and a press release
  4. A vendor contract clause and a staff training session

Explanation

High-risk surveillance requires justification that the intrusion is necessary and proportionate to a legitimate aim. It also requires assessment of privacy and bias impacts and a channel for affected communities to be heard before deployment.

Question 40Data Ethics Foundations

Which of the following best describes the ethical role of a data protection impact assessment (DPIA) in a project's lifecycle?

  1. A financial audit of the project's budget
  2. A one-time marketing approval gate
  3. A technical performance benchmark for the model
  4. An ongoing process that identifies and mitigates risks to individuals before and during processing Correct answer.

Explanation

A DPIA is a structured risk process focused on the rights and freedoms of individuals affected by processing. It should begin before processing and be revisited as the project changes, because risks evolve with design and deployment.

Question 41Data Ethics Foundations

A data broker claims its dataset is 'anonymous' because it removed names and addresses, yet the dataset includes precise timestamps and geolocation. What is the most accurate ethical assessment?

  1. The dataset is safe because it contains no names
  2. The dataset carries meaningful re-identification risk and should be treated as personal data unless proven otherwise Correct answer.
  3. The dataset is fully anonymous because direct identifiers were removed
  4. The dataset is safe as long as it is stored in the cloud

Explanation

Removing direct identifiers does not eliminate re-identification when quasi-identifiers such as precise time and location remain. Such data should be treated as personal until a rigorous assessment shows the risk is acceptably low.

Question 42Data Ethics Foundations

Which action best reflects the ethical duty to respect 'data subject rights' in an operational system?

  1. Applying deletion requests only to records created after the request
  2. Requiring individuals to visit a single office in person to make any request
  3. Building a verified process to access, correct, and delete an individual's data within defined timeframes Correct answer.
  4. Storing all requests in a shared spreadsheet without tracking

Explanation

Respecting rights means the organization can reliably find, correct, and erase personal data on request within stated deadlines. A verified, tracked process makes those rights effective rather than nominal.

Question 43Data Ethics Foundations

A model trained on historical hiring data performs well overall but has a much lower recall for one applicant group. Which metric-focused response is most ethically appropriate?

  1. Report only overall accuracy because it is highest
  2. Remove the group attribute so the disparity cannot be measured
  3. Report subgroup-level error rates and investigate the cause before deployment Correct answer.
  4. Increase the model's overall accuracy slightly and ship

Explanation

Aggregate accuracy can conceal harms concentrated in a subgroup, so subgroup-level metrics are needed to surface them. Investigation should follow to determine whether the disparity stems from data, labels, or model design.

Question 44Data Ethics Foundations

Which statement best explains why 'explainability' is an ethical requirement and not merely a technical convenience?

  1. It guarantees the model will be accurate
  2. It removes the need for human oversight
  3. It enables affected people to understand and challenge decisions that materially affect them Correct answer.
  4. It reduces cloud computing costs

Explanation

Explainability matters ethically because people subject to consequential decisions need enough understanding to contest them. It supports due process and human oversight rather than replacing them.

Question 45Data Ethics Foundations

A team wants to use synthetic data to reduce privacy risk in a shared dataset. Which condition is most important for that claim to be ethically sound?

  1. The original data was deleted from backups
  2. The generative model has been evaluated for leakage of real records and for fitness for the intended use Correct answer.
  3. The synthetic data was generated quickly
  4. The synthetic data looks realistic to a casual viewer

Explanation

Synthetic data only reduces privacy risk if the generation process does not memorize or leak real records. It must also preserve the statistical properties needed for the intended analysis, or it will mislead downstream users.

Question 46Data Ethics Foundations

What is the most appropriate ethical response when a model's training data is later found to contain illegally obtained personal information?

  1. Sell the model before the issue becomes public
  2. Assess and mitigate the impact, including retraining or withdrawal where warranted, and disclose to affected parties as required Correct answer.
  3. Delete only the original files while keeping the trained model
  4. Continue using the model and note the issue in an internal wiki

Explanation

When training data provenance is compromised, the organization must evaluate the harm and take corrective action such as retraining or withdrawing the model. Transparency toward affected parties and regulators is part of the remedy.

Question 47Data Ethics Foundations

Which practice best reflects 'privacy by design' in a new analytics product?

  1. Adding a privacy notice after launch
  2. Encrypting data only after a breach occurs
  3. Embedding data minimization, access controls, and retention limits into the product architecture from the start Correct answer.
  4. Collecting all possible data and deciding later what to keep

Explanation

Privacy by design builds protections into the system architecture rather than bolting them on afterward. Minimization, access control, and retention limits are decided during design, when they can actually shape the product.

Question 48Data Ethics Foundations

An ethics review board is asked to approve a study using data from people in a detention facility. Which consideration is most central to its judgment?

  1. Whether the data can be processed quickly
  2. Whether the dataset is stored in a standardized format
  3. Whether the researchers hold academic titles
  4. Whether the conditions of detention compromise voluntary consent and whether additional safeguards are needed Correct answer.

Explanation

People in detention may face coercion or fear of consequences that undermine genuinely voluntary consent. Review should therefore focus on whether consent is meaningful and what extra protections are needed.

Question 49Data Ethics Foundations

Which scenario best illustrates a 'dark pattern' in a consent interface?

  1. A clear settings page where privacy options are equally prominent
  2. A pop-up where the 'accept all' button is large and prominent while 'reject' is a small grey link buried in a submenu Correct answer.
  3. A plain-language explanation of how data is used
  4. A quarterly privacy newsletter

Explanation

Dark patterns manipulate interface design to steer people toward choices that serve the organization rather than the user. Making rejection hard to find while acceptance is prominent undermines meaningful consent.

Question 50Data Ethics Foundations

What does the principle of 'contestability' most directly require of an automated decision system?

  1. That the system never makes errors
  2. That decisions are final once issued
  3. That the system's source code is published in full
  4. That affected individuals have a practical route to question the decision and obtain review by a capable human Correct answer.

Explanation

Contestability means people can challenge a decision and have it reviewed by someone able to change the outcome. It requires an accessible process and human reviewers with real authority, not merely an appeal address.