Digital Work, People Analytics, and Artificial Intelligence

Introduction

Digital technologies have transformed the employment relationship. Work may now be performed from a home office, a client site, a platform, or several locations connected through cloud systems. Recruitment, scheduling, performance assessment, payroll, employee communication, and dismissal may be supported—or partly determined—by algorithms. These developments create opportunities for flexibility and productivity, but they also raise important questions of labour law, industrial relations, privacy, equality, evidence, and managerial accountability.

For human resource professionals, digital work is not simply a technology issue. It is a legal and institutional issue concerning who controls work, how decisions are made, what information is collected, and how employees can challenge decisions. A technically efficient system may still be unlawful if it discriminates, violates privacy, denies due process, or interferes with collective representation.

This chapter uses a general comparative perspective. Specific rights differ across jurisdictions, so practitioners must consult applicable employment, privacy, equality, occupational safety, collective bargaining, evidence, and cybersecurity legislation. International standards and national laws increasingly converge around several principles: necessity, proportionality, transparency, human oversight, security, non-discrimination, and effective remedy.

Learning Objectives

By the end of this chapter, students should be able to:

1. Distinguish remote, hybrid, platform, and digitally managed work.
2. Evaluate electronic surveillance against privacy, labour, and equality principles.
3. Explain algorithmic management and assess automated recruitment, performance, and dismissal systems.
4. Apply responsible people-analytics methods to turnover, disputes, bargaining, and compliance data.
5. Identify employee data rights and basic cybersecurity responsibilities.
6. Assess digital evidence and design a defensible governance process for AI-enabled HR.
7. Recommend fair, transparent, and accountable approaches to digital work and industrial relations.

1. The Changing Employment Relationship

1.1 Remote and hybrid work

Remote work is work performed away from the employer’s ordinary premises using digital communication and information systems. Hybrid work combines remote and on-site work. These arrangements may improve autonomy, access to talent, inclusion for some workers, and continuity during emergencies. They may also create risks: blurred working time, isolation, unequal visibility, inadequate equipment, unsafe home workspaces, weak information security, and unequal access to promotion.

The legal classification of remote work matters. A remote employee normally remains an employee, with continuing rights relating to wages, working time, occupational safety, leave, equality, consultation, and collective representation. Working from another country may additionally trigger rules concerning immigration, taxation, social security, data transfers, and applicable labour law.

A sound remote-work policy should address:

• eligibility and business justification;
• working hours, availability, rest periods, and overtime approval;
• equipment, technical support, expenses, and reimbursement;
• health and safety, including ergonomic and psychosocial risks;
• confidentiality, access control, and incident reporting;
• performance expectations based on outputs rather than constant visibility;
• reasonable accommodation and accessibility;
• consultation with employee representatives; and
• review, withdrawal, and dispute-resolution procedures.

Remote work does not eliminate managerial responsibility. Employers should assess risks without demanding intrusive access to the employee’s private home. A virtual meeting background, for example, should not become an indirect inspection of domestic circumstances.

1.2 Platform work and employment status

Platform work is organized through a digital intermediary that matches customers and workers, allocates tasks, processes payments, or rates performance. Examples include delivery, ride services, online freelancing, domestic services, and crowdwork. Platforms often describe workers as independent contractors, but legal status depends on the actual relationship, not merely contractual language.

Relevant indicators may include the platform’s control over price, work allocation, acceptance standards, monitoring, branding, discipline, customer relationship, and ability to substitute another worker. Misclassification can deny minimum wages, social insurance, paid leave, collective rights, and protection from unfair termination.

Platform workers also face distinctive algorithmic risks. A low customer rating may reduce access to tasks; an automated fraud flag may suspend an account; and opaque ranking systems may effectively discipline workers without a hearing. Regulation in many jurisdictions is moving toward greater disclosure, human review, and presumptions or tests for employment status.

2. Electronic Surveillance and Employee Privacy

Electronic surveillance includes monitoring email, network use, location, keystrokes, attendance, biometric data, cameras, call recordings, screenshots, or productivity indicators. Monitoring may serve legitimate purposes such as cybersecurity, safety, quality assurance, wage compliance, or protection of confidential information. The same tools can become excessive or coercive when used to measure every moment of work.

2.1 A proportionality framework

A useful evaluation asks five questions:

1. Purpose: What legitimate and specific objective is being pursued?
2. Necessity: Is monitoring genuinely needed, or can the objective be achieved through a less intrusive method?
3. Proportionality: Do the benefits justify the intrusion and risks?
4. Transparency: Were employees informed clearly before monitoring began?
5. Safeguards: Who can access the data, how long is it retained, and how can an employee challenge misuse?

Monitoring should be limited to relevant data. Continuous webcam recording, for example, is generally more intrusive than recording system access logs. Productivity scores based on keystrokes may measure activity rather than meaningful work and may disadvantage employees whose jobs require reflection, care, collaboration, or screen-free tasks.

Privacy is not necessarily absolute inside the workplace. However, employees do not surrender all privacy when they accept employment. Notice should explain the categories of data collected, purposes, legal basis where applicable, recipients, retention period, automated decision-making, and complaint routes. Consent may be weak in employment because of unequal bargaining power; a lawful basis should not be assumed merely because a worker clicked “accept.”

2.2 Special categories of data

Biometric identifiers, health information, precise location, union membership, political opinions, and disability-related information are often especially protected. Processing such information requires strong justification, access restrictions, security controls, and sometimes a formal impact assessment. An employer should not collect sensitive information simply because software makes collection easy.

3. Algorithmic Management and Artificial Intelligence

Algorithmic management occurs when computational systems assist or determine decisions about work allocation, scheduling, evaluation, pay, promotion, discipline, or termination. Artificial intelligence generally refers to systems that perform tasks associated with human cognitive processes, such as prediction, classification, language processing, or pattern recognition. The distinction is practical rather than absolute: a simple rule-based scheduling system can be legally significant even if it is not advanced AI.

A typical HR system may contain several stages:

1. Input: applications, attendance records, ratings, survey responses, or work outputs.
2. Processing: rules, statistical models, machine learning, or generative AI.
3. Output: ranking, risk score, recommendation, alert, or decision.
4. Action: interview invitation, schedule, coaching, discipline, or termination.
5. Feedback: later outcomes are fed back into the system, potentially reproducing past patterns.

This creates a crucial principle: a human signature does not automatically make a decision human. If a manager routinely accepts an algorithmic recommendation without examining its basis, the organization has retained nominal rather than meaningful human oversight.

3.1 Automated hiring

Recruitment systems may parse résumés, rank applicants, generate interview questions, analyze language or video, and predict job performance. Risks include:

• historical bias embedded in training data;
• proxy discrimination, such as using postcode as a proxy for race or socioeconomic status;
• inaccessible assessments for disabled applicants;
• cultural or linguistic bias in language analysis;
• unlawful use of personal or publicly available information;
• unclear explanations and inability to correct inaccurate data; and
• automation of exclusion without a meaningful appeal.

A responsible process begins with job-related criteria validated against the actual requirements of the role. It should test whether each feature is necessary, monitor outcomes across relevant groups where lawful, provide reasonable accommodations, and ensure that a qualified human reviews adverse decisions.

3.2 Automated performance management and dismissal

Systems may assign productivity scores, predict absence, identify “flight risk,” or flag alleged misconduct. Such outputs are indicators, not facts. A prediction of resignation is not proof of disloyalty; a low activity score is not proof of poor performance; and a fraud alert is not proof of wrongdoing.

Dismissal is particularly sensitive. Before adverse action, the employee should ordinarily receive notice of the concern, relevant evidence, an opportunity to respond, consideration by an authorized decision-maker, and access to applicable grievance or appeal processes. Automated termination without explanation or review conflicts with basic procedural fairness and may breach employment law or collective agreements.

4. Testing AI Systems for Legality, Bias, and Accountability

AI governance should operate across the system life cycle rather than begin after harm occurs.

4.1 Pre-deployment assessment

Before procurement or deployment, HR should document:

• the purpose and affected decisions;
• the vendor, model type, training-data provenance, and update process;
• data categories and legal basis for processing;
• groups likely to experience disparate impact;
• accessibility and accommodation requirements;
• security, retention, and cross-border transfer arrangements;
• human decision rights and escalation routes; and
• measurable criteria for suspension or withdrawal.

A data protection impact assessment may be required by law and is valuable even when not mandatory. An equality or algorithmic impact assessment should examine whether apparently neutral variables reproduce protected-group disadvantage.

4.2 Bias and performance testing

A confusion matrix can organize the results of a binary classification system. It distinguishes true positives, false positives, true negatives, and false negatives. For example, in recruitment, a “positive” result might mean selecting an applicant for interview. A false positive is an applicant selected by the model who does not meet the defined criterion; a false negative is an applicant rejected despite meeting it.

Useful measures include:

Here, TP means true positives, FP false positives, FN false negatives, and TN true negatives. Precision asks how many predicted positives were correct; recall asks how many actual positives were identified. These measures apply to a defined outcome and require reliable labels. They do not, by themselves, establish fairness or job validity.

For group comparisons, an analyst may compare selection rates:

The subscript g identifies a demographic or other legally relevant group. Differences may signal disparate impact, but they do not prove unlawful discrimination. Context, job-relatedness, sample size, measurement validity, legitimate accommodations, and applicable legal tests must be examined. Fairness metrics can also conflict: improving one measure may worsen another, especially when groups have different base rates. Therefore, statistical testing must be combined with legal analysis and qualitative investigation.

Testing should occur before launch and periodically afterward. Monitoring should examine data drift, changes in workforce composition, vendor updates, and feedback from affected workers. Results should be documented, and a model should be paused when serious unexplained harm appears.

4.3 Explainability and contestability

Explainability need not reveal protected trade secrets or source code in every case. It does require a comprehensible account of the decision’s relevant factors, the system’s role, the evidence used, and the route for correction. Contestability means that an affected person can question the result, supply additional information, and obtain review by a competent human who has authority to change the outcome.

5. People Analytics and Industrial Relations

People analytics uses workforce data to understand patterns and support decisions. Appropriate uses include identifying workload risks, evaluating training, monitoring turnover, planning staffing, and detecting recurring safety or grievance problems. Analytics becomes problematic when correlation is treated as causation, when data are collected for one purpose and reused for another, or when aggregate analysis becomes individualized surveillance.

5.1 Responsible metrics

A basic turnover rate can be expressed as:

The period must be specified, and the numerator should distinguish voluntary resignations, involuntary dismissals, retirements, transfers, and redundancy where relevant. Average headcount may be calculated as the mean of beginning and ending headcount when that approximation is reasonable. A high turnover rate is a signal for investigation, not a diagnosis.

For example, if an organization records 48 separations during a year and average headcount is 600:

The calculation is descriptive. It does not show whether turnover reflects poor management, seasonal work, restructuring, labour-market competition, or healthy mobility.

A dispute dashboard might include grievance volume, time to acknowledgment, time to resolution, settlement rate, repeat issues, representation status, and outcomes by issue type. A bargaining dashboard might track proposals, response times, unresolved clauses, costing assumptions, strike or lockout notices where lawful, and ratification results. Data should be aggregated and access-controlled so that dashboards do not expose union membership or individual complainants unnecessarily.

5.2 Data interpretation and industrial relations

Analytics can strengthen collective bargaining by identifying workload, scheduling, safety, pay-equity, or turnover patterns. It can also damage trust if management presents selective data as objective truth. Employee representatives should understand definitions, data quality, sampling, exclusions, and limitations. Where collective consultation rights exist, the introduction of monitoring or automated management may itself require information and consultation.

A responsible dashboard therefore includes a data dictionary, source and refresh date, coverage limitations, confidence or uncertainty information where relevant, and an owner accountable for interpretation. It should distinguish facts, estimates, and recommendations.

6. Digital Evidence and Due Process

Digital evidence includes emails, messaging records, access logs, GPS records, audit trails, system-generated scores, video, metadata, and vendor reports. Its usefulness depends on authenticity, relevance, completeness, reliability, and lawful acquisition.

An evidence-handling process should:

1. define the issue under investigation;
2. preserve relevant data through a documented legal hold;
3. record who collected it, when, and by what method;
4. protect original files and calculate integrity checks where appropriate;
5. separate personal, privileged, and irrelevant information;
6. give the employee a fair opportunity to respond; and
7. retain and delete material according to law and policy.

A dashboard screenshot without underlying records may be difficult to interpret. A model-generated explanation may not accurately represent the factors actually used. Vendor secrecy does not remove the employer’s responsibility to provide a fair process. Employers should contract for audit access, incident notification, data portability, retention controls, and cooperation with lawful investigations.

7. Cybersecurity and Resilient Digital Work

Cybersecurity is part of the employment relationship because employers hold payroll, identity, health, performance, and communications data. Core controls include multi-factor authentication, least-privilege access, encryption, secure backups, patching, device management, phishing awareness, vendor due diligence, and incident-response plans.

Security should not become a justification for unlimited surveillance. Access logs should be used for defined security purposes, and anomalous activity should be investigated proportionately. Employees need clear reporting channels for lost devices, suspected phishing, accidental disclosure, and unsafe systems. Training should be practical and role-specific rather than limited to annual acknowledgments.

A cyber incident may create labour-relations consequences: payroll interruption, unsafe work, inability to schedule staff, disclosure of union or health data, and disputes about whether employees must use compromised systems. Continuity plans should identify critical processes, alternative communication channels, pay arrangements, and consultation responsibilities.

8. Employee Data Rights and Governance

Depending on jurisdiction, employees may have rights to notice, access, correction, deletion or restriction, portability, objection, confidentiality, and safeguards against solely automated decisions. Some rights are qualified by legitimate business needs, legal retention duties, security, litigation holds, or collective agreements. HR must not promise an absolute right that local law does not provide, but it should establish a clear procedure for requests.

A practical governance model assigns responsibility across three levels:

• Board or senior leadership: approves risk appetite and receives serious-risk reports.
• HR, legal, privacy, IT, and equality specialists: design controls, assess systems, and investigate incidents.
• Managers and employees: use systems lawfully, protect information, report concerns, and participate in review.

A useful policy is to maintain an AI and monitoring register containing the system’s purpose, owner, affected workforce, data categories, supplier, risk assessment, testing results, review date, and incident history. High-impact systems should receive more intensive review than low-risk administrative tools.

9. Industrial Relations in the Digital Workplace

Digitalization changes not only individual rights but also collective power. Remote employees may be dispersed, platform workers may lack a common workplace, and algorithmic scheduling can make collective action difficult. At the same time, digital communication can support organizing, consultation, virtual bargaining, and rapid information sharing.

Collective agreements should consider:

• consultation before introducing monitoring or AI;
• disclosure of system purpose and relevant performance criteria;
• limits on intrusive surveillance;
• data access for representatives in aggregated form;
• human review of automated discipline and dismissal;
• algorithmic impact assessments;
• training and redeployment when technology changes jobs;
• working-time boundaries and right-to-disconnect arrangements;
• cybersecurity responsibilities; and
• procedures for technological disputes.

Employers should avoid using analytics to undermine lawful organizing or infer union membership from communication patterns. Employee representatives should seek technically meaningful information rather than merely receiving a general statement that “AI is being used.”

10. Illustrative Case: The Atlas Scheduling System

Background

Atlas Retail, a fictional national retailer, introduced an AI-assisted scheduling system. The system predicted customer demand, assigned shifts, and generated an “attendance reliability” score using lateness records, shift swaps, absence data, and manager ratings. Workers with low scores received fewer desirable shifts. The company also used automated alerts to recommend disciplinary action.

Problem

Within six months, part-time caregivers and workers with disabilities reported reduced hours. The system treated approved medical absences and unavoidable public-transport delays as negative indicators. Managers were told that the score was “objective” and rarely changed recommendations. The union requested information about the model and its effects.

Analysis

HR first preserved the relevant model versions, data dictionary, scheduling records, and decision logs. It discovered that approved protected absences had not been excluded consistently. The model also used shift-swap frequency, which penalized employees whose care responsibilities required flexibility. The company had given general notice of digital scheduling but had not explained scoring, review rights, or retention.

The organization compared outcomes by employment status, disability accommodation status where lawfully available, shift type, and location. It found a substantial reduction in desirable shifts for accommodated workers, but recognized that the analysis could not alone establish unlawful discrimination. Legal, privacy, equality, IT, HR, and union representatives jointly reviewed the system.

Approach and solution

Atlas suspended automated disciplinary recommendations and restored human review. It removed protected absences from the reliability score, tested alternative scheduling variables, created an appeal process, and gave workers an accessible explanation of scores. It provided back pay or shift restoration where an individual review showed measurable loss. The collective agreement was amended to require consultation before material changes to algorithmic scheduling.

The company introduced quarterly impact testing, vendor audit rights, role-based access, retention limits, and a system register. Managers were trained that a model output was a prompt for inquiry, not proof of misconduct.

Outcome

Scheduling disparities declined, appeals were resolved more quickly, and employee representatives gained access to aggregated impact reports. Some efficiency gains were lower than originally forecast, but the organization reduced legal, reputational, and industrial-relations risk.

Learning points

1. A lawful business objective does not legalize every method used to pursue it.
2. Protected information may be mishandled indirectly through variables that encode its effects.
3. Human oversight must include authority, time, information, and independence.
4. Collective consultation improves both legitimacy and system quality.
5. Impact assessment must continue after deployment.

11. Implementation Framework for HR Professionals

A practical implementation sequence is:

Map: identify systems, decisions, data flows, affected workers, vendors, and representatives.

Classify risk: distinguish administrative tools from systems affecting pay, access to work, discipline, promotion, or dismissal.

Justify: document purpose, necessity, alternatives, legal basis, and expected benefits.

Consult: involve privacy, legal, equality, cybersecurity, managers, employees, and unions as appropriate.

Test: validate job relevance, accessibility, accuracy, disparate impacts, security, and explainability.

Pilot: use a limited deployment with monitoring and a defined stop rule.

Decide: preserve meaningful human judgment and due process for high-impact decisions.

Monitor: review outcomes, complaints, incidents, model changes, and workforce effects.

Remediate: correct data, compensate harm where appropriate, suspend unsafe tools, and communicate lessons.

This framework treats compliance as a continuing management process rather than a one-time form.

12. Limitations and Critical Perspectives

Digital systems are not neutral observers. They reflect choices about what counts as performance, whose data are available, and which outcomes are valued. Historical data may encode earlier discrimination. Quantitative metrics may encourage gaming and suppress professional judgment. Privacy notices may be unreadable, and consent may not be meaningful in unequal employment relationships.

There are also technical limits. Small samples can produce unstable group comparisons. Missing data can conceal disadvantage. Correlation does not establish causation. A fair result on a test dataset may deteriorate after deployment. Vendors may update models without adequate notice. Generative AI may produce confident but inaccurate explanations or disclose confidential information.

Accordingly, organizations should retain non-digital channels: a human manager, union representative, privacy officer, ombudsperson, or appeal body. Some decisions may be unsuitable for automation regardless of accuracy because dignity, trust, or procedural fairness requires personal judgment.

Contemporary Developments

Current developments include regulation of high-risk AI, stronger scrutiny of automated employment decisions, legal recognition of platform-worker protections, right-to-disconnect initiatives, biometric restrictions, pay-transparency measures, and collective bargaining over algorithmic management. Generative AI is entering recruitment, policy drafting, employee service desks, learning, and investigation support. Its use requires special care because generated text can be inaccurate, biased, or based on confidential prompts.

The future HR professional will need combined competence in employment law, data governance, industrial relations, statistics, cybersecurity, and organizational ethics. The central professional question is not whether a system is innovative, but whether it improves work while respecting rights and preserving accountability.

Chapter Summary

Digital work includes remote, hybrid, platform, and algorithmically managed arrangements. These models can provide flexibility and efficiency but raise questions about employment status, working time, health and safety, privacy, equality, and collective representation. Electronic monitoring should have a legitimate purpose, be necessary and proportionate, be transparent, and include safeguards.

Algorithmic recruitment, scheduling, performance management, and dismissal require life-cycle governance. Human oversight must be meaningful, and adverse decisions should be explainable, reviewable, and supported by due process. Bias testing can use measures such as precision, recall, and group selection rates, but statistical results must be interpreted with legal, contextual, and qualitative evidence.

People analytics can support responsible decisions about turnover, disputes, bargaining, safety, and staffing when definitions, data quality, access, and limitations are clear. Digital evidence must be lawfully collected, preserved, authenticated, and disclosed fairly. Cybersecurity protects both organizational continuity and employee rights. Strong governance combines leadership accountability, specialist review, employee participation, collective bargaining, impact assessment, and effective remedies.

Key Terms

Algorithmic management: Use of computational systems to allocate, monitor, evaluate, or control work.

Artificial intelligence: Computational systems performing tasks associated with human cognitive processes.

Automated decision-making: A decision made partly or wholly through a system without ordinary human judgment.

Bias: Systematic distortion that can produce unfair or inaccurate outcomes.

Collective bargaining: Negotiation between an employer and employee representatives concerning employment conditions.

Contestability: The ability of an affected person to challenge and seek review of a decision.

Data minimization: Collection and use of no more personal data than necessary for a defined purpose.

Digital evidence: Electronic information relevant to an investigation, dispute, or legal proceeding.

Electronic surveillance: Monitoring of employee communications, activity, location, devices, or performance through technology.

Explainability: A comprehensible account of how relevant data and system processes contributed to an outcome.

False negative: A negative classification when the relevant positive condition is actually present.

Human oversight: Competent human supervision with authority to review and change an automated outcome.

Hybrid work: Work combining remote and on-site working arrangements.

People analytics: Systematic analysis of workforce data to inform organizational decisions.

Platform work: Work organized or mediated through a digital platform.

Precision: The proportion of predicted positive cases that are true positives.

Recall: The proportion of actual positive cases identified by a system.

Remote work: Work performed away from the employer’s ordinary premises using digital systems.

Transparency: Clear information about a system’s purpose, operation, data, and consequences.

Turnover rate: Separations during a period divided by average headcount, expressed as a percentage.

Review Questions

Short Answer Questions

1. What distinguishes remote work from hybrid work?
2. State three questions in a proportionality assessment of employee surveillance.
3. Why is an algorithmic recommendation not necessarily proof of misconduct?
4. Define precision and recall in a classification system.
5. Identify four cybersecurity controls relevant to HR data.

Long Answer Questions

1. Explain how platform work challenges traditional tests of employment status and collective representation.
2. Discuss the legal and ethical risks of automated recruitment and propose safeguards.
3. Explain how people analytics can support, and potentially undermine, industrial relations.
4. Describe the requirements for fair use of digital evidence in an employment investigation.
5. Evaluate the role of collective bargaining in governing algorithmic management.

Analytical / Application Questions

1. An employer uses keystroke counts to rank remote employees. Analyze the privacy, equality, validity, and industrial-relations issues and recommend an alternative evaluation approach.
2. A hiring model selects 30% of Group A applicants and 18% of Group B applicants for interview. What additional information would be required before concluding that the model is unlawful or biased? Explain why selection rates alone are insufficient.
3. Design a governance plan for an AI system that predicts absence and recommends disciplinary action. Include data controls, impact testing, human review, employee rights, cybersecurity, and monitoring after deployment.

Multiple Choice Questions with Answers

Question 1

Which principle requires an employer to consider whether a less intrusive monitoring method could achieve the same legitimate objective?

A. Necessity

B. Automation

C. Portability

D. Seniority

Correct Answer: A. Necessity

Question 2

A platform worker is most likely to be legally classified by examining:

A. The label used in the contract alone

B. The actual degree of control and dependence in the relationship

C. The worker’s preferred job title

D. The platform’s advertising slogan

Correct Answer: B. The actual degree of control and dependence in the relationship

Question 3

In a classification system, precision is calculated as:

A. TP ÷ (TP + FN)

B. TN ÷ (TN + FP)

C. TP ÷ (TP + FP)

D. FP ÷ (TP + FP)

Correct Answer: C. TP ÷ (TP + FP)

Question 4

Why should a low algorithmic performance score not automatically establish misconduct?

A. Algorithms never use data

B. A score is an indicator requiring contextual and human review

C. Managers are prohibited from reviewing scores

D. Performance cannot ever be measured

Correct Answer: B. A score is an indicator requiring contextual and human review

Question 5

Which is an example of meaningful human oversight?

A. A manager automatically accepts every model recommendation

B. A vendor alone decides all appeals

C. A trained decision-maker reviews evidence and has authority to change the outcome

D. An employee receives no explanation but may resign

Correct Answer: C. A trained decision-maker reviews evidence and has authority to change the outcome

Question 6

If an organization records 20 separations and has an average headcount of 400, its turnover rate is:

A. 2%

B. 5%

C. 20%

D. 80%

Correct Answer: B. 5%

Question 7

Which practice best supports defensible digital evidence?

A. Deleting inconvenient records immediately

B. Preserving relevant data and documenting collection and integrity

C. Using only screenshots without context

D. Allowing unrestricted access to all HR files

Correct Answer: B. Preserving relevant data and documenting collection and integrity

Question 8

Which control limits employees’ access to only the information required for their roles?

A. Least privilege

B. Open access

C. Data duplication

D. Continuous recording

Correct Answer: A. Least privilege

Question 9

Why can postcode be problematic as a hiring-model variable?

A. It is never a data field

B. It may act as a proxy for protected or socioeconomic characteristics

C. It guarantees job performance

D. It prevents all statistical analysis

Correct Answer: B. It may act as a proxy for protected or socioeconomic characteristics

Question 10

A responsible people-analytics dashboard should include:

A. Only a final recommendation

B. A data dictionary, source information, limitations, and accountable ownership

C. Names of all complainants by default

D. Unverified predictions presented as facts

Correct Answer: B. A data dictionary, source information, limitations, and accountable ownership