Research, Evidence and Information Quality

Introduction

Human resource professionals increasingly make decisions from mixed information: employee surveys, interviews, administrative records, labour-market reports, academic studies, dashboards, consultant briefings, and material generated or summarized by artificial intelligence. The central communication problem is not simply finding information. It is determining what can reasonably be believed, what can be claimed, and how confidently a claim should be communicated.

In graduate-level HR management, evidence literacy is therefore both a research capability and a managerial responsibility. A report recommending a new selection procedure, hybrid-work policy, pay-equity intervention, or wellbeing programme may influence careers, expenditure, legal exposure, and organizational trust. Weak evidence can produce attractive but indefensible recommendations; strong evidence can be misused through selective reporting, overstated causality, or misleading visual presentation.

This chapter presents an integrated approach to gathering, evaluating, synthesizing, and communicating evidence. It distinguishes primary from secondary evidence, explains reliability, validity, credibility, and bias, introduces practical appraisal frameworks, clarifies citation and plagiarism, and shows how to manage information overload. The emphasis is on defensible professional judgement rather than mechanical adherence to checklists.

Learning Objectives

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

1. Distinguish primary, secondary, and tertiary information and select sources appropriate to a decision.
2. Formulate answerable research questions and construct effective search strategies.
3. Assess the reliability, validity, credibility, relevance, and limitations of evidence.
4. Identify sampling, measurement, researcher, publication, confirmation, and algorithmic bias.
5. Synthesize diverse findings without confusing association with causation.
6. Apply ethical citation practices and prevent plagiarism, fabrication, falsification, and misrepresentation.
7. Communicate uncertainty and reduce information overload in HR reports.
8. Use digital and AI tools with appropriate human verification and governance.

1. Evidence as a Foundation of Professional Communication

Evidence is information used to justify a description, explanation, evaluation, prediction, or recommendation. Data become evidence only when they are relevant to a question, collected or selected through a defensible process, interpreted appropriately, and communicated with sufficient context.

A useful chain is:

Question → search and design → source → data → analysis → inference → claim → decision.

Errors can enter at every stage. A poorly framed question may retrieve irrelevant material. A representative sample may be measured with an invalid instrument. Correct statistical analysis may still support an unjustified causal claim. A sound study may be selectively quoted in a report. Evidence quality is therefore multidimensional rather than a single score.

A high-quality claim normally makes five elements visible:

• The proposition: what is being asserted.
• The evidence base: which observations or sources support it.
• The warrant: why those observations support the proposition.
• The boundary: where the claim may or may not apply.
• The uncertainty: what remains unknown or contested.

For example, “Employees who used the mentoring programme had higher retention” is descriptive. “The mentoring programme caused higher retention” is causal and requires stronger design, such as random assignment, a credible comparison group, or a well-justified quasi-experimental method.

Important distinctions

2. Primary, Secondary, and Tertiary Evidence

Primary evidence is produced directly from the phenomenon or investigation. In HR it may include employee interviews, an original engagement survey, attendance records, payroll data, performance assessments, observation, experiments, or an organization’s original policy documents. Primary evidence offers proximity to the question but is not automatically unbiased or valid. An internal survey may suffer from non-response, fear of reprisal, leading questions, or poor measurement.

Secondary evidence analyzes, summarizes, compares, or interprets primary sources. Examples include peer-reviewed systematic reviews, meta-analyses, government labour statistics, scholarly literature reviews, and professional reports. Secondary evidence can reveal patterns across settings and reduce the burden of interpreting individual studies, but its quality depends on search methods, inclusion criteria, analytical choices, and source transparency.

Tertiary information provides orientation or condensed access, such as encyclopedias, textbooks, databases, news summaries, and general websites. It is useful for developing vocabulary and locating original sources, but important claims should normally be traced to primary or authoritative secondary evidence.

The categories describe a source’s relationship to the underlying event, not its prestige. A government administrative dataset is primary evidence for a labour-market count, while a newspaper account of that count is secondary. A peer-reviewed empirical article is primary for its own study and secondary when it synthesizes earlier studies.

Matching evidence to purpose

Different decisions require different evidence. An HR manager assessing legal compliance should prioritize legislation, regulations, case law, and authoritative guidance. A manager estimating employee sentiment may use a validated survey supplemented by interviews. A strategic workforce forecast may combine internal headcount data, demographic statistics, economic forecasts, and scenario analysis. No single source type is universally superior.

The strongest practical approach is often triangulation: examining a question through multiple methods, sources, investigators, theories, or data types. For instance, turnover records may show that exits increased, interviews may illuminate perceived causes, and a benchmark study may indicate whether the increase is unusual. Convergence increases confidence, while divergence may reveal subgroup differences or measurement problems.

3. From Management Problem to Researchable Question

Information quality begins with question quality. A broad request such as “Find research on employee wellbeing” is inefficient. A more useful question specifies population, intervention or exposure, comparison, outcomes, and context. The PICO structure can be adapted for management:

• Population: Which employees or organizational units?
• Intervention or exposure: What policy, practice, or condition?
• Comparison: Compared with what, if anything?
• Outcome: What measurable result and over what period?

For example: “Among customer-service employees, does supervisor training compared with existing practice reduce voluntary turnover and reported emotional exhaustion over twelve months?” This question guides keywords, data needs, and appraisal criteria.

A search plan should include synonyms, related concepts, exclusions, date boundaries, geographic scope, and source-quality requirements. Search terms for flexible work might include “remote work,” “telework,” “hybrid work,” “distributed work,” “productivity,” “retention,” and “employee wellbeing.” Citation chaining—examining references cited by a useful article and later studies that cite it—can complement keyword searches.

A defensible research log records:

• databases and websites searched;
• exact search strings and dates;
• inclusion and exclusion criteria;
• duplicates removed;
• sources screened and rejected;
• reasons for exclusion;
• limitations in access or coverage.

This record improves reproducibility and reduces the risk of unconsciously selecting only convenient or supportive material.

4. Evaluating Information Quality

4.1 Authority and provenance

Authority concerns who produced the information, their expertise, institutional role, and accountability. Provenance concerns where the material came from and whether its history is traceable. A source should identify its author, organization, publication date, methods, data origin, funding, and version where relevant. Anonymous material may be useful for generating leads but should rarely carry a major recommendation without corroboration.

Peer review is a quality-control process, not a guarantee of truth. Published studies can contain methodological weaknesses, errors, conflicts of interest, or conclusions that later research challenges. Conversely, high-quality official statistics may not be peer reviewed but can be authoritative for narrowly defined administrative facts.

4.2 Relevance and applicability

A credible source may still be irrelevant. A study of technology firms in one country may not generalize to hospitals, small enterprises, or another legal system. Appraisal should consider population, setting, intervention, outcome, timing, and institutional context. Transfer is a reasoned judgement, not an automatic consequence of statistical significance.

4.3 Reliability

Reliability concerns consistency. A reliable measure produces similar results under consistent conditions, although it may consistently measure the wrong construct. Common forms include:

• Test–retest reliability: stability over time.
• Inter-rater reliability: agreement between observers or coders.
• Internal consistency: coherence among items intended to measure the same construct.
• Procedural reliability: consistency in data collection and coding.

Reliability is evidence, not a complete verdict. A measure can be reliable but invalid, such as a clock that is always ten minutes slow.

4.4 Validity

Validity concerns whether an interpretation or measure is justified. Important forms include:

• Construct validity: whether an instrument represents the intended concept.
• Content validity: whether relevant dimensions are adequately covered.
• Criterion-related validity: whether scores relate to an appropriate external criterion.
• Internal validity: whether observed effects can reasonably be attributed to the proposed cause rather than confounding factors.
• External validity: whether findings can transfer to other people, settings, or times.

Validity is not a permanent property of an instrument. A measure may be valid for one population and purpose but inappropriate for another. In HR, “performance” may mean sales, quality, customer ratings, teamwork, or supervisor judgement; each operationalization has consequences.

4.5 Transparency and reproducibility

A report becomes more trustworthy when readers can understand how evidence was obtained and analyzed. Transparency includes clear definitions, sampling procedures, missing-data treatment, coding rules, analytical methods, funding, conflicts of interest, and deviations from the original plan. Reproducibility means another competent analyst could repeat the process, subject to privacy and security constraints.

4.6 Currency

Currency matters differently by topic. A current regulation or labour-market figure is essential, while a foundational theory may remain influential for decades. Recent does not mean superior; recency should be balanced with methodological quality and historical significance.

A practical appraisal matrix can organize judgement:

5. Bias, Uncertainty, and Inference

Bias is a systematic tendency that distorts observation, analysis, or interpretation. It differs from random error, which varies unpredictably across observations. Large samples may reduce random error but do not automatically correct systematic bias.

Common forms include:

• Selection bias: the observed sample differs systematically from the target population.
• Non-response bias: people who do not respond differ from respondents.
• Measurement bias: wording, instrument, timing, or context systematically affects responses.
• Recall bias: participants remember past events inaccurately.
• Social-desirability bias: respondents give acceptable rather than truthful answers.
• Confirmation bias: analysts notice evidence supporting prior beliefs.
• Publication bias: positive or statistically significant results are more likely to be published.
• Survivorship bias: analysis includes only those who remain visible or successful.
• Conflict-of-interest bias: financial or institutional interests influence design or interpretation.
• Algorithmic bias: data, labels, objectives, or deployment conditions produce systematically unfair outcomes.

Bias should be analyzed as a process. Ask who defines the question, who is included, who is absent, which variables are measured, who benefits from the interpretation, and what consequences follow from error. Disaggregation by gender, race or ethnicity, disability, age, contract type, location, or other relevant categories can reveal effects hidden in aggregate results. Such analysis must respect privacy, sample-size limitations, and applicable law.

Association and causation

An association means variables vary together; it does not by itself establish that one causes the other. Causal interpretation requires attention to temporal order, plausible mechanisms, confounding, selection, and design. Randomized experiments can strengthen causal inference when assignment is genuinely random and implementation is sound. In organizational settings, quasi-experimental designs—difference-in-differences, interrupted time series, matching, or regression discontinuity—may be more feasible, but their assumptions must be stated and tested.

A responsible report uses calibrated language: “is associated with,” “is consistent with,” “may contribute to,” or “the evidence suggests,” rather than “proves” or “causes,” unless the design warrants that confidence.

6. Synthesizing Evidence

Synthesis is more than placing quotations side by side. It identifies patterns, differences, explanations, and implications across sources. A synthesis should compare study populations, designs, measures, outcomes, and limitations.

A simple evidence map may classify findings as:

A qualitative synthesis may use thematic coding. The analyst develops codes, applies them consistently, compares cases, searches for negative cases, and preserves an audit trail. A quantitative synthesis may compare effect estimates, confidence intervals, sample sizes, and heterogeneity. Statistical combination should not be undertaken merely because studies share a topic; measures, populations, and designs must be sufficiently comparable.

The synthesis should distinguish:

1. What is well established.
2. What is plausible but uncertain.
3. What is contested or inconsistent.
4. What has not been adequately studied.
5. What decision can be made despite uncertainty.

Uncertainty does not mean inaction. It means matching the scale, reversibility, monitoring, and safeguards of a decision to the strength of evidence. A low-cost pilot with predefined evaluation criteria may be preferable to either immediate large-scale implementation or indefinite delay.

7. Citation, Attribution, and Academic Integrity

Citation performs three functions: it gives credit, allows verification, and shows the intellectual basis of a claim. Citation is required for quoted words, paraphrased ideas, distinctive arguments, data, statistics, theories, images, tables, and material that is not common knowledge.

A paraphrase is not created by changing a few words. It restates the source’s meaning in genuinely independent language and still cites the source. A quotation reproduces exact wording, identifies the source, and normally includes a page or paragraph locator. A reference list must correspond to in-text citations, and bibliographic details should be checked against the original source.

APA 7th edition is one widely used standard. Typical patterns include:

• Parenthetical citation: (Author, Year).
• Narrative citation: Author (Year) argues that …
• Direct quotation: (Author, Year, p. 15).
• Organization as author: (International Labour Organization, 2023).

Institutions may require another standard, such as Harvard, Chicago, or OSCOLA. The essential principles are consistency, completeness, accuracy, and traceability. Citation managers can store metadata and generate references, but they can import errors and should not replace source checking.

Plagiarism is presenting another person’s words, ideas, data, structure, or creative work as one’s own. It includes verbatim copying without quotation and citation, patchwriting, uncited close paraphrase, self-plagiarism where prohibited, and presenting generated text as original scholarship. Fabrication means inventing data or sources. Falsification means manipulating materials, procedures, or results so that the record is misleading. Misrepresentation includes selectively reporting findings, omitting material limitations, altering a graph’s scale deceptively, or implying a source says more than it does.

A practical integrity workflow is:

1. Capture source details immediately.
2. Separate copied notes from personal analysis.
3. Mark quotations clearly.
4. Paraphrase only after understanding the passage.
5. Cite the original source where possible.
6. Verify every citation and quotation.
7. Compare the final report with the evidence, including contrary findings.

8. Managing Information Overload

Information overload occurs when the volume, speed, complexity, or contradiction of information exceeds the user’s capacity to process it for a decision. More information can reduce quality when it obscures priorities, encourages superficial reading, or delays action.

An evidence brief should therefore define the decision, audience, deadline, materiality threshold, and stopping rule. A stopping rule might be reached when additional searching produces no materially new evidence, when key stakeholder perspectives are represented, or when the remaining uncertainty cannot be reduced within available resources.

Useful reporting practices include:

• begin with a decision-relevant executive summary;
• distinguish evidence from interpretation and recommendation;
• use headings, tables, and short paragraphs;
• present the strongest counterevidence;
• report denominators, dates, and population definitions;
• avoid unnecessary precision;
• use one clear visual for one important message;
• place technical detail in an appendix;
• state confidence and limitations explicitly.

Visual integrity is part of evidence quality. Axes should be labelled, units and denominators shown, time periods defined, and truncated axes disclosed where used. A dramatic percentage increase from a very small base should be accompanied by the underlying counts.

9. AI, Digital Sources, and Governance

Search engines, text-mining systems, reference managers, transcription tools, and generative AI can accelerate discovery and organization. They can also produce fabricated citations, reproduce training-data bias, omit minority perspectives, expose confidential information, and generate plausible but unsupported claims.

AI output should be treated as an unverified draft or retrieval aid, not as evidence. Users should verify claims against original sources, inspect methods, check quotations, preserve prompts and versions where relevant, and disclose material AI assistance according to institutional policy. Confidential employee records should not be uploaded to tools lacking approved privacy, security, retention, and access controls.

Governance should address data minimization, lawful processing, informed access, human oversight, explainability appropriate to the decision, auditability, contestability, and non-discrimination. High-impact HR uses—such as recruitment screening, performance evaluation, promotion, or termination—require particular caution. A technically accurate model may still be unacceptable if its objective encodes unfairness or if affected employees cannot challenge an outcome.

10. Applied Case: Evaluating a Hybrid-Work Proposal

The following is a fictional illustrative case.

Background

Northbridge Services, a 1,200-employee customer-support organization, proposes allowing eligible staff to work remotely three days per week. The executive sponsor claims that hybrid work will increase productivity and reduce turnover. The HR director must prepare a board report.

Problem

An internal pulse survey of 240 volunteers shows that 72% prefer hybrid work. A consultant report states that remote work increases productivity by 13%. Finance reports lower office-utilization costs, while two team leaders report slower collaboration. The evidence is inconsistent and the survey is not representative.

Analysis

The HR analyst first defines the decision: whether to conduct a six-month, monitored pilot for eligible roles, not whether hybrid work is universally beneficial. The analyst identifies weaknesses:

• volunteer survey respondents may be more positive toward remote work;
• the consultant’s population, outcome definition, and funding are unclear;
• productivity may be confounded by role, manager, workload, and seasonality;
• customer satisfaction and employee wellbeing are not yet measured;
• employees whose work requires physical presence may experience inequity.

The analyst then triangulates evidence: historical productivity and quality records, voluntary turnover by role, customer metrics, structured interviews with a purposive range of employees, relevant peer-reviewed research, and legal and occupational-safety guidance. Outcomes are defined before implementation: service-level attainment, quality scores, absenteeism, voluntary turnover intention, actual turnover, wellbeing, promotion access, and customer satisfaction.

Approach and solution

Northbridge uses a staggered pilot. Eligible teams are selected with attention to comparable workload and role mix; some teams begin later, creating a comparison period. The report avoids claiming randomization because managers participate voluntarily. Baseline measures are recorded, missing responses are reported, and results are disaggregated by role, disability accommodation, gender, career stage, and location where privacy permits.

The communication plan distinguishes facts, assumptions, and forecasts. It states that cost savings are estimated rather than realized, that productivity effects are uncertain, and that collaboration risks will be monitored. The board receives a decision rule: continue, modify, or stop the pilot based on predefined service-quality and equity thresholds, not on a single favourable indicator.

Outcome

After six months, service levels remain stable, voluntary turnover intention falls modestly among eligible employees, and promotion access is lower for remote employees who have less informal contact with managers. Office costs decline, but managers report uneven coordination. The organization extends the pilot with structured meeting norms, quarterly career reviews, and an equity audit rather than adopting an unrestricted policy.

Learning points

The case illustrates that evidence quality depends on design and interpretation, not the number of charts in a report. A preference survey cannot establish productivity effects. Triangulation can reveal both benefits and distributional risks. A reversible pilot, transparent uncertainty, and predefined decision criteria can convert incomplete evidence into responsible action.

Challenges and Limitations

Evidence appraisal cannot eliminate judgement. Time, access restrictions, privacy obligations, weak organizational records, small subgroups, and rapidly changing conditions may limit certainty. Some important outcomes, such as trust or dignity, are difficult to measure without reducing them to inadequate proxies. Quantitative indicators can create false confidence, while qualitative accounts can be compelling but not representative.

There are also ethical tensions. Linking datasets may improve analysis but increase re-identification risk. Disaggregating outcomes may expose small groups. Publishing negative findings may challenge powerful sponsors. A responsible professional protects participants, reports material limitations, and resists pressure to convert uncertainty into unwarranted certainty.

Best-Practice Checklist for HR Reports

Before submission, ask:

1. Is the decision and audience explicit?
2. Are key terms, populations, dates, and denominators defined?
3. Have primary claims been traced to original or authoritative sources?
4. Are source quality, methods, funding, and conflicts assessed?
5. Does the design support the strength of the causal language used?
6. Are contrary findings and relevant subgroups included?
7. Are quotations, paraphrases, data, and visuals correctly attributed?
8. Can another reviewer reproduce the search and analysis sufficiently?
9. Are privacy, fairness, legal, and governance risks addressed?
10. Is the recommendation proportionate to evidence strength and uncertainty?

Chapter Summary

Evidence-based business communication requires more than collecting information. It requires a clear question, an appropriate evidence base, systematic appraisal, careful synthesis, and honest communication of uncertainty. Primary evidence provides direct observations; secondary evidence interprets and integrates existing work; tertiary sources support orientation but should not carry major claims without verification. Reliability concerns consistency, while validity concerns the justification of an interpretation or measurement. Authority, relevance, transparency, currency, independence, and applicability must be assessed together.

Bias can enter through sampling, measurement, analysis, publication, organizational incentives, and algorithms. Triangulation and disaggregation can expose weaknesses and differences that aggregate results conceal. Causal claims require stronger support than descriptive claims, and professional reports should use calibrated language. Citation protects intellectual integrity and enables verification; plagiarism, fabrication, falsification, and selective reporting undermine both scholarship and managerial trust. Finally, information overload is controlled through decision-focused search strategies, stopping rules, transparent visuals, concise synthesis, and evidence governance. AI can assist research, but original-source verification, privacy protection, human judgement, and accountability remain essential.

Key Terms

Algorithmic bias: Systematic unfairness arising from data, model design, objectives, or deployment.

Association: A relationship in which variables vary together without necessarily implying causation.

Bias: A systematic distortion in sampling, measurement, analysis, or interpretation.

Causal inference: Reasoning about whether an intervention or exposure produced an outcome.

Construct validity: The extent to which a measure represents the intended concept.

Data: Recorded observations, responses, measurements, or documents.

Evidence: Information evaluated as relevant support for a claim.

External validity: The extent to which findings apply beyond the original study context.

Fabrication: Inventing data, results, sources, or research records.

Falsification: Manipulating research materials, procedures, or results to mislead.

Information overload: A condition in which information exceeds the capacity for effective processing and decision-making.

Internal validity: The credibility of an inference that an observed effect was caused by the proposed factor.

Plagiarism: Presenting another’s words, ideas, data, or creative work as one’s own.

Primary evidence: Evidence collected directly from the phenomenon or original investigation.

Reliability: Consistency of a measure, procedure, or coding process.

Secondary evidence: Analysis, interpretation, or synthesis of primary sources.

Synthesis: Structured integration of findings across sources to identify patterns, differences, and implications.

Triangulation: Use of multiple methods, sources, investigators, or theories to examine a question.

Validity: The degree to which an interpretation or measurement is justified for its intended purpose.

Transparency: Clear disclosure of methods, assumptions, data origins, limitations, and analytical decisions.

Review Questions

Short Answer Questions

1. Distinguish primary, secondary, and tertiary information, using one HR example for each.
2. Why does reliability not guarantee validity?
3. Identify four common sources of bias in organizational research.
4. What is the difference between association and causation?
5. State three functions of academic citation.

Long Answer Questions

1. Explain how an HR manager could evaluate the quality of an external research report before using it in a board recommendation.
2. Discuss the role of triangulation in resolving disagreement between employee survey results and administrative records.
3. Compare internal validity and external validity, and explain why an HR intervention may have one without the other.
4. Explain how plagiarism, fabrication, falsification, and selective reporting differ and why each damages professional communication.
5. Develop a comprehensive evidence-governance protocol for an HR department using generative AI in research and reporting.

Analytical / Application Questions

1. An organization reports that employees who attend resilience training have lower absenteeism. Analyze at least four alternative explanations and propose a stronger evaluation design.
2. A workforce dashboard shows that promotion rates are equal across departments, but employees from one demographic group report lower access to developmental assignments. Explain how aggregation may conceal inequity and design a follow-up analysis.
3. You must advise whether to adopt a four-day workweek using one internal survey, two case studies, and a meta-analysis with mixed findings. Construct an evidence matrix and state what recommendation, caveats, and monitoring plan you would present.

Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes primary from secondary evidence?

A. Primary evidence is always more reliable than secondary evidence.

B. Primary evidence is collected directly, whereas secondary evidence analyzes or synthesizes existing sources.

C. Primary evidence is published, whereas secondary evidence is unpublished.

D. Primary evidence is quantitative, whereas secondary evidence is qualitative.

Correct Answer: B. Primary evidence is collected directly, whereas secondary evidence analyzes or synthesizes existing sources.

Question 2

A measure gives nearly identical results each time but does not actually measure employee engagement. This measure is best described as:

A. Reliable but invalid

B. Valid but unreliable

C. Neither reliable nor biased

D. Externally valid and causal

Correct Answer: A. Reliable but invalid

Question 3

Which practice most directly reduces confirmation bias in an evidence review?

A. Searching only highly cited articles

B. Excluding studies with inconvenient findings

C. Defining inclusion criteria in advance and actively searching for contrary evidence

D. Using the newest source available

Correct Answer: C. Defining inclusion criteria in advance and actively searching for contrary evidence

Question 4

Why is a volunteer employee survey potentially vulnerable to selection bias?

A. Volunteers may differ systematically from employees who do not participate.

B. Volunteers always provide inaccurate answers.

C. A survey cannot measure attitudes.

D. Selection bias occurs only in randomized experiments.

Correct Answer: A. Volunteers may differ systematically from employees who do not participate.

Question 5

Which wording is most appropriate when a cross-sectional study finds that flexible work and retention are related?

A. Flexible work proves higher retention.

B. Flexible work causes higher retention in all organizations.

C. Flexible work is associated with higher retention in the observed sample.

D. Retention has no relationship to flexible work.

Correct Answer: C. Flexible work is associated with higher retention in the observed sample.

Question 6

Which action is plagiarism?

A. Summarizing a source in original language with a citation

B. Quoting exact words with quotation marks and a citation

C. Changing a few words from a source while omitting attribution

D. Using common knowledge without a citation

Correct Answer: C. Changing a few words from a source while omitting attribution

Question 7

What is the principal purpose of triangulation?

A. To guarantee that all findings agree

B. To examine a question using multiple sources or methods

C. To replace primary research with opinions

D. To increase the number of pages in a report

Correct Answer: B. To examine a question using multiple sources or methods

Question 8

Which feature most strengthens the transparency of an evidence review?

A. A persuasive executive summary without methods

B. A record of search terms, dates, inclusion criteria, and exclusions

C. Removal of all limitations from the final report

D. Use of only sources supporting the recommendation

Correct Answer: B. A record of search terms, dates, inclusion criteria, and exclusions

Question 9

Why should AI-generated citations be checked against original sources?

A. AI systems cannot produce text quickly.

B. AI systems may generate inaccurate, incomplete, or fabricated references.

C. Original sources are never subject to error.

D. Citation checking is unnecessary in professional reports.

Correct Answer: B. AI systems may generate inaccurate, incomplete, or fabricated references.

Question 10

Which response best manages uncertainty in a proposed HR intervention?

A. Implement it permanently without measurement.

B. Reject it because evidence is never certain.

C. Use a monitored, reversible pilot with predefined outcomes and decision thresholds.

D. Report only the most favourable indicator.

Correct Answer: C. Use a monitored, reversible pilot with predefined outcomes and decision thresholds.