15. Doctoral Research and Future Governance Frontiers

Introduction: Governance as a Research Problem

Corporate governance and business ethics are not static bodies of rules. They are evolving fields concerned with how authority is allocated, how decisions are justified, how organizational power is constrained, and how firms affect stakeholders and society. At doctoral level, governance should therefore be treated not merely as a compliance topic but as a field of theory construction, empirical investigation, and normative reasoning.

Research in this area must explain variation. Why do some boards challenge executives while others become symbolic bodies? Under what conditions does stakeholder governance improve long-term performance? When do ownership structures reduce agency costs, and when do they create entrenchment? How do artificial intelligence, geopolitical conflict, decentralized organizations, and systemic risks alter traditional assumptions about accountability?

These questions require methodological pluralism. Archival data may reveal relationships among ownership, board structure, disclosure, and performance; experiments can identify causal effects of framing or incentives; surveys can measure perceptions and ethical climate; qualitative research can uncover processes that are invisible in numerical data; and mixed methods can connect breadth with explanation. The central doctoral challenge is not selecting the most fashionable method, but aligning research questions, theory, evidence, inference, and ethical responsibility.

Learning Objectives

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

1. Formulate theory-driven research questions in corporate governance and business ethics.

2. Distinguish major research paradigms and select methods appropriate to different questions.

3. Evaluate measurement validity, causal identification, endogeneity, replication, and generalizability.

4. Design archival, econometric, experimental, survey, qualitative, comparative-case, and mixed-method studies.

5. Apply research ethics to studies involving organizations, employees, investors, and sensitive data.

6. Analyze emerging governance problems involving geopolitical fragmentation, decentralized organizations, AI-enabled firms, systemic risk, and corporate purpose.

7. Develop defensible research contributions for scholarship, policy, and managerial practice.

From Governance Concepts to Researchable Claims

A doctoral study begins with a problem, but a problem is not yet a research question. “Board diversity matters” is a broad proposition. A researchable question specifies the relationship, context, mechanism, and outcome: How does professionally diverse board membership affect cybersecurity oversight in highly regulated firms, and through what monitoring mechanisms?

A strong research design generally connects five elements. First is the theoretical construct, such as agency conflict, stakeholder salience, legitimacy, institutional complexity, stewardship, or organizational identity. Second is the mechanism through which an effect is expected to occur. Third is the observable indicator used to represent the construct. Fourth is the empirical setting and unit of analysis. Fifth is the inferential claim: description, association, explanation, prediction, or causal effect.

The distinction between constructs and measures is fundamental. “Ethical culture” is a construct; employee responses to validated climate items, whistleblowing rates, and disciplinary records are possible indicators. No single indicator perfectly captures a complex construct. Researchers must therefore justify operationalization, consider measurement error, and explain whether the measure reflects the concept itself or merely one consequence of it.

Theory should do more than decorate an introduction. It should guide the selection of variables, specify boundary conditions, identify mechanisms, and generate propositions that could be contradicted by evidence. A useful theoretical contribution may extend an existing theory, reconcile apparently conflicting findings, identify a new mechanism, or demonstrate that a relationship depends on institutional conditions.

Theoretical Lenses for Governance Research

Agency theory emphasizes conflicts between principals and agents, including shareholders and managers, controlling and minority shareholders, or citizens and public enterprises. It supports research on incentives, monitoring, ownership, executive compensation, and disclosure. Its limitation is that it may understate intrinsic motivation, professional norms, and collective responsibilities.

Stakeholder theory treats the firm as a system of relationships involving employees, customers, suppliers, communities, creditors, regulators, and investors. Research informed by this perspective may examine stakeholder participation, materiality, human rights due diligence, or the relationship between stakeholder treatment and resilience. A central challenge is specifying how competing interests should be prioritized.

Stewardship theory proposes that managers may identify with organizational purposes and act as responsible stewards rather than opportunistic agents. It invites research into trust, purpose, professional identity, and non-financial motivation, but it should not be used to assume benevolence without evidence.

Institutional theory explains how firms respond to regulatory, cultural, and normative pressures. It is particularly relevant to symbolic compliance, sustainability reporting, global standards, and differences among national governance systems. Institutional conformity may create legitimacy without substantive change, making decoupling an important empirical issue.

Resource dependence theory views boards as providers of information, legitimacy, relationships, and access to critical resources. It directs attention beyond monitoring toward board networks, political connections, expertise, and interorganizational dependence.

Legitimacy and political theories examine how corporations seek social acceptance and exercise influence. They are useful for studying lobbying, political spending, public commitments, crisis communication, and the boundary between private authority and public governance.

These perspectives are complementary but not interchangeable. Combining them requires theoretical discipline. A study should explain which assumptions are retained, where theories conflict, and whether the proposed integration produces novel predictions.

Choosing Evidence: A Methodological Architecture

Archival and Econometric Research

Archival research uses existing records such as annual reports, regulatory filings, ownership databases, executive biographies, litigation records, media archives, sustainability reports, and market data. It is valuable for large-sample analysis and historical comparison. Yet availability does not guarantee validity. A disclosure index may measure reporting practice rather than actual conduct; a board-interlock measure may capture visibility rather than influence.

Econometric analysis can estimate associations and, under credible assumptions, causal effects. A basic panel model might relate governance practice to organizational outcomes while controlling for firm and year effects. Fixed effects can address time-invariant unobserved characteristics, but they do not solve time-varying omitted variables or reverse causality. Difference-in-differences designs compare changes between affected and unaffected groups following a policy shock. Their credibility depends on parallel trends and the absence of simultaneous differential shocks. Instrumental-variable designs require an instrument that affects the explanatory variable while influencing the outcome only through that variable, an assumption that must be defended rather than asserted.

Endogeneity is a central threat. Firms with strong ethics may voluntarily adopt better governance, making governance appear effective even when underlying culture drives both adoption and outcomes. Possible responses include natural experiments, lagged specifications, matching, instrumental variables, regression discontinuity, sensitivity analysis, qualitative process tracing, and triangulation. No statistical technique automatically creates causality.

Researchers should report effect sizes, uncertainty intervals, model specifications, missing-data decisions, alternative measures, and robustness tests. Statistical significance alone does not establish practical importance. A small estimated effect may be highly consequential at systemic scale, while a large coefficient may be economically irrelevant in practice.

Experiments and Behavioral Governance

Experiments are useful when the research question concerns judgment, decision processes, incentives, or causal reactions to information. In a vignette experiment, participants might evaluate a manager who discloses an algorithmic error under different accountability conditions. A laboratory or online experiment can manipulate board monitoring, ethical framing, or compensation design while randomly assigning participants to conditions.

The strength of random assignment is internal validity. Its limitations include artificiality, demand effects, limited samples, and uncertain transferability to organizational settings. Researchers should preregister hypotheses where appropriate, justify sample size through power analysis, distinguish exploratory from confirmatory analyses, and avoid treating a student sample as evidence about all executives or investors.

Field experiments may improve realism but create additional ethical and operational challenges. Deception, unequal treatment, and effects on employees or customers require careful review. Experimental findings should be interpreted as evidence about the tested mechanism and population, not as universal laws.

Surveys and Measurement Models

Surveys can examine ethical climate, trust in boards, perceived fairness, stakeholder voice, moral disengagement, or attitudes toward corporate purpose. High-quality survey research requires construct clarity, suitable scales, translation procedures, pilot testing, attention checks used judiciously, and strategies to reduce common-method bias.

Reliability is not validity. A scale can produce consistent responses while measuring the wrong construct. Researchers should consider content validity, convergent validity, discriminant validity, criterion validity, and measurement invariance across countries, languages, or organizational groups. If a measure functions differently across groups, comparisons may be misleading.

Survey results are also vulnerable to social desirability, nonresponse, selection bias, and retrospective distortion. Anonymous administration, neutral wording, multiple data sources, and temporal separation of measurements can help. Ethical research requires that participants understand the study, risks, voluntary nature of participation, and data-use arrangements.

Qualitative Inquiry and Comparative Cases

Qualitative research is especially appropriate when governance processes are poorly understood, meanings are contested, or context is central. Interviews, observations, documents, meeting records, digital traces, and participant accounts can reveal how formal rules are interpreted and enacted.

A rigorous qualitative study makes its sampling logic explicit. Purposive sampling may seek variation in ownership, regulation, or organizational purpose; theoretical sampling may proceed as emerging concepts guide further data collection. Coding should be systematic, reflexive, and connected to the research question. Credibility may be strengthened through triangulation, negative-case analysis, member reflection, audit trails, and transparent positionality statements.

Comparative case research can identify patterns across countries, industries, firms, or governance regimes. Comparison is meaningful only when cases are selected for theoretical reasons. A most-similar design holds important conditions relatively constant while examining a key difference; a most-different design searches for a common mechanism across contrasting settings. Case studies do not become rigorous by accumulating detail alone. They require a clear chain of evidence and disciplined claims about transferability.

Mixed Methods

Mixed methods is more than placing qualitative and quantitative chapters side by side. It integrates different forms of evidence in a design such as explanatory sequential research, in which statistical findings are followed by interviews; exploratory sequential research, in which qualitative work informs scale development or hypothesis testing; or convergent designs, in which datasets are analyzed separately and then compared.

Integration should occur at the level of research questions, sampling, measurement, analysis, or interpretation. For example, a panel study may find that firms adopting stakeholder committees experience fewer controversies, while interviews reveal whether the committees exercise authority or merely improve public symbolism. Divergence is not a failure; it may identify boundary conditions or measurement problems.

Validity, Replication, and Responsible Inference

Doctoral researchers must distinguish internal validity, construct validity, external validity, and statistical conclusion validity. Internal validity concerns whether the observed relationship can be attributed to the proposed cause. Construct validity concerns whether variables represent the intended concepts. External validity concerns the settings and populations to which findings may transfer. Statistical conclusion validity concerns whether the evidence supports the claimed relationship given sampling, measurement, and analytic uncertainty.

Replication is essential in governance research because findings may depend on specific periods, countries, samples, coding rules, or model choices. A direct replication tests whether a result recurs under substantially similar conditions. A conceptual replication tests the underlying proposition using different measures or settings. Reproducible workflows, version-controlled code, data dictionaries, preregistration, and clear reporting of exclusions improve credibility.

Researchers should avoid p-hacking, selective reporting, HARKing (presenting post hoc explanations as if they were predicted), and overreliance on a single favored specification. Open data are desirable when lawful and ethically permissible, but confidentiality, commercial sensitivity, and participant protection may require controlled access or synthetic data. Transparency must not expose vulnerable individuals or breach contractual obligations.

Research Ethics and the Governance of Knowledge

Research ethics extends beyond obtaining an approval form. Participants must not be coerced, misled without justification, or exposed to unreasonable risks. In organizational research, employees may feel pressure to participate if supervisors distribute invitations. Confidentiality can be difficult when a small sample makes individuals identifiable. Researchers must manage conflicts of interest, disclose funding, preserve records, and report adverse findings honestly.

Corporate governance research also raises ethical questions about the use of public information. Legally accessible data may still contain sensitive personal details. Web-scraped material, employee reviews, leaked documents, and social-media posts require careful consideration of consent, expectation of privacy, platform terms, and potential harm. Researchers should minimize data collection, secure files, define retention periods, and explain who will access results.

Ethical responsibility includes consequences of publication. A study that labels a group of employees unethical, exposes a whistleblower, or publishes a firm-level accusation without adequate evidence can cause serious harm. Scholarly independence requires resisting sponsor pressure to suppress inconvenient results while recognizing legitimate confidentiality constraints.

Future Governance Frontiers

Geopolitical Fragmentation

Global firms increasingly operate across jurisdictions with conflicting sanctions, data rules, labor expectations, tax regimes, and political alignments. Fragmentation challenges the assumption that a single global governance model can be applied consistently. Research opportunities include cross-border board accountability, supply-chain due diligence, sanctions compliance, political risk, and the governance of strategic technologies.

The key analytical issue is not only whether firms comply with different rules, but how they translate conflicting norms into organizational decisions. Comparative institutional research can examine whether firms centralize standards, localize responsibility, or create layered governance. Researchers must avoid treating national categories as fixed and should account for intra-country variation and transnational private standards.

Decentralized Organizations and Distributed Authority

Platform cooperatives, decentralized autonomous organizations, blockchain-based ventures, and open-source communities distribute decision rights differently from conventional corporations. Token voting, smart contracts, reputation systems, and online deliberation can reduce dependence on hierarchical intermediaries, but they may also reproduce unequal power through wealth, technical expertise, or control of infrastructure.

Research should examine who can participate, who controls code and treasury assets, how disputes are resolved, and whether formal decentralization produces substantive accountability. Code is not neutral governance: design choices embed assumptions about membership, voting, privacy, and enforcement. Decentralized arrangements also face legal-personhood, fiduciary, consumer-protection, and money-laundering concerns.

AI-Enabled Firms

Artificial intelligence changes both the object and process of governance. Firms use algorithms in recruitment, lending, pricing, fraud detection, compliance, forecasting, and board reporting. AI may improve consistency and analytic capacity, but it can also amplify historical discrimination, obscure responsibility, generate fabricated outputs, and create security vulnerabilities.

Research should move beyond asking whether an algorithm is accurate. It should investigate fairness across groups, explainability for affected stakeholders, contestability, human oversight, data provenance, model drift, and accountability when responsibility is distributed among developers, vendors, managers, and users. Governance systems should maintain inventories of high-impact models, document intended use, test for disparate effects, monitor performance after deployment, and provide escalation and remedy mechanisms.

Systemic Risk and Interdependence

Traditional governance often evaluates firms individually, whereas systemic risk arises from connections among firms, markets, infrastructures, and states. Common exposures, correlated strategies, concentrated service providers, cyber dependence, and climate-related shocks can transform private decisions into collective instability.

This frontier requires network analysis, stress testing, scenario research, and attention to non-linear effects. A governance practice that benefits one firm may increase fragility for the system, such as excessive reliance on a single cloud provider or synchronized asset sales. Boards therefore need information about dependencies beyond their immediate legal entity. Researchers should connect micro-level incentives with macro-level outcomes and examine whether regulatory oversight can address risks that no individual firm has an incentive to internalize.

Evolving Corporate Purpose

Corporate purpose has moved from a narrow emphasis on shareholder returns toward more explicit attention to stakeholders, sustainability, resilience, and social legitimacy. The research challenge is to distinguish substantive purpose from rhetorical adoption. Relevant evidence includes capital allocation, executive incentives, product decisions, political activity, workforce treatment, and responses to externalities.

Purpose can clarify strategic priorities, but it can also create accountability ambiguity if objectives are broad and trade-offs are undisclosed. A doctoral study may examine how purpose is defined, who participates in its formation, how it is embedded in controls, and whether stakeholders can challenge deviations. Normative research should address the legitimacy of corporate power, while empirical research should test whether purpose affects decisions rather than merely communications.

An Illustrative Case: Meridian Systems and Algorithmic Accountability

The following is a fictional illustrative case.

Background. Meridian Systems is a multinational provider of employment-screening software. Its platform uses machine-learning models to rank applicants for corporate clients. The company has expanded across jurisdictions with different employment, privacy, and algorithmic-accountability requirements.

Situation. An external researcher reports that applicants from certain educational and geographic backgrounds receive systematically lower scores. Meridian states that the model is based on historical hiring outcomes and that its overall predictive accuracy is high. The board’s technology committee asks management for assurance. Management provides an aggregate accuracy figure but cannot explain how training data were selected, how model drift is monitored, or who is responsible for appeals.

Analysis. The problem is simultaneously technical, governance-related, and ethical. Accuracy does not establish fairness. Historical outcomes may encode discriminatory practices. Responsibility is distributed among data suppliers, model developers, clients, compliance officers, and senior management. The board also faces an information asymmetry problem: it must oversee a system that it cannot itself engineer.

A doctoral research team could study the case through a sequential mixed-method design. First, archival analysis could examine model documentation, incident logs, client contracts, and regulatory correspondence. Second, statistical testing could evaluate error rates and disparate impacts across relevant groups, subject to lawful access and privacy safeguards. Third, interviews with engineers, compliance staff, clients, and affected applicants could trace how accountability operates in practice. Finally, a comparative analysis could examine whether firms with independent algorithmic oversight achieve more effective remediation than firms relying solely on management assurance.

Approach. Meridian establishes an independent algorithmic-risk committee, requires documented impact assessments before deployment, conducts subgroup performance testing, introduces human review and appeal procedures, and gives the board periodic reports on model drift, complaints, overrides, and remediation. It also revises vendor contracts to allocate responsibilities for data quality, audit access, and incident notification. The company pauses use of the model in jurisdictions where adequate safeguards cannot yet be established.

Outcome. The measures do not eliminate all bias, and some clients resist slower hiring processes. However, Meridian gains a clearer accountability structure and identifies that the most serious problem arose not from code alone but from unexamined historical labels and weak escalation procedures.

Learning points. The case demonstrates the importance of construct validity, process tracing, mixed methods, stakeholder participation, and governance of the full data lifecycle. It also shows why responsible innovation requires contestability and remedy, not merely technical performance.

Common Research Failures and Best Practices

Frequent failures include beginning with available data rather than a meaningful question; treating correlation as causation; using a proxy without discussing its limitations; selecting cases opportunistically; assuming a validated scale works identically across cultures; reporting only supportive findings; and making universal claims from a narrow institutional setting.

Strong practice includes:

• Aligning theory, constructs, measures, design, and inference.
• Defining the unit of analysis and temporal order explicitly.
• Conducting power, sensitivity, and robustness analyses where appropriate.
• Combining statistical evidence with process evidence when mechanisms matter.
• Documenting coding rules, exclusions, missing data, and analytic decisions.
• Engaging affected stakeholders without surrendering scholarly independence.
• Separating descriptive, predictive, causal, and normative claims.
• Designing governance recommendations that specify responsibility, resources, monitoring, and remedy.

The most valuable doctoral work is not necessarily the most technically elaborate. It is work that makes a defensible contribution, recognizes uncertainty, and connects rigorous evidence with the realities of organizational power.

Chapter Summary

Doctoral research in corporate governance and business ethics requires disciplined movement from theoretical constructs to researchable claims and credible evidence. Agency, stakeholder, stewardship, institutional, resource dependence, legitimacy, and political theories offer different explanations of organizational power and accountability. Archival, econometric, experimental, survey, qualitative, comparative-case, and mixed-method approaches each answer different questions and carry distinct validity risks.

Endogeneity, measurement error, weak causal identification, selective reporting, and poor replication can undermine otherwise sophisticated studies. Research ethics requires attention to consent, confidentiality, data security, conflicts of interest, and the consequences of publication. Emerging frontiers—including geopolitical fragmentation, decentralized organizations, AI-enabled firms, systemic risk, and corporate purpose—demand research that crosses organizational, technological, institutional, and societal levels. The central standard is alignment: a meaningful question, an appropriate theory, a transparent method, responsible evidence, and claims proportionate to what the design can support.

Key Terms

Construct: An abstract concept that a study seeks to examine, such as trust or ethical climate.

Operationalization: The process of translating a theoretical construct into observable measures.

Endogeneity: A condition in which an explanatory variable is related to unobserved influences on the outcome, undermining causal interpretation.

Internal validity: The credibility of a study’s causal inference within its research setting.

External validity: The extent to which findings transfer to other populations, settings, or periods.

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

Measurement invariance: Evidence that a measure has comparable meaning across groups or contexts.

Archival research: Research using existing organizational, regulatory, market, or documentary records.

Difference-in-differences: A design comparing changes over time between affected and comparison groups.

Instrumental variable: A variable used to identify causal effects under specific relevance and exclusion assumptions.

Preregistration: Publicly recording hypotheses and methods before examining outcome data.

Replication: Repeating or re-testing research to assess whether findings are reproducible.

HARKing: Presenting hypotheses developed after results are known as if specified beforehand.

Triangulation: Using multiple sources, methods, or perspectives to strengthen interpretation.

Theoretical sampling: Selecting additional cases or participants based on emerging theoretical needs.

Mixed methods: An integrated design combining quantitative and qualitative evidence.

Algorithmic accountability: Processes that assign responsibility for automated systems and provide oversight, explanation, and remedy.

Systemic risk: The possibility that interconnected failures generate instability across a broader system.

Corporate purpose: The stated and enacted rationale for a corporation’s existence and responsibilities.

Research ethics: Principles governing participant protection, integrity, confidentiality, and responsible knowledge production.

Review Questions

Short Answer Questions

1. Why is a construct different from its measure in governance research?

2. What problem does endogeneity create for causal inference?

3. Distinguish direct replication from conceptual replication.

4. Why may high algorithmic accuracy fail to demonstrate fairness?

5. What is the purpose of theoretical sampling in qualitative research?

Long Answer Questions

1. Compare agency theory, stakeholder theory, and stewardship theory as foundations for doctoral governance research.

2. Explain how a researcher could use a difference-in-differences design to study a governance regulation.

3. Discuss the principal validity threats in survey-based research on ethical climate.

4. Explain how mixed methods can reveal mechanisms that a large-sample archival study might miss.

5. Analyze the governance implications of geopolitical fragmentation for multinational corporations.

Analytical / Application Questions

1. Design a research project examining whether board-level AI oversight improves accountability. Specify the theory, unit of analysis, measures, method, and principal threats to inference.

2. A study finds that firms with formal corporate-purpose statements have fewer public controversies. Develop at least three competing explanations and propose evidence that could distinguish them.

3. Using the Meridian Systems case, recommend a governance structure that balances innovation, fairness, privacy, accountability, and organizational efficiency.

22. Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes a construct from an operational measure?

A. A construct is always quantitative, whereas a measure is always qualitative.

B. A construct is a theoretical concept, whereas a measure is an observable indicator of that concept.

C. A construct is a research result, whereas a measure is a hypothesis.

D. A construct is objective, whereas a measure is necessarily subjective.

Correct Answer: B. A construct is a theoretical concept, whereas a measure is an observable indicator of that concept.

Question 2

Why does a statistically significant association between strong ethics programs and performance not by itself establish causality?

A. Statistical significance cannot be used with archival data.

B. Performance can never be measured reliably.

C. Unobserved factors or reverse causality may influence both ethics programs and performance.

D. Ethics programs are always symbolic.

Correct Answer: C. Unobserved factors or reverse causality may influence both ethics programs and performance.

Question 3

Which condition is essential to a credible difference-in-differences design?

A. The treatment and comparison groups must have identical outcomes after treatment.

B. The groups should have plausibly followed parallel trends before the intervention.

C. The researcher must exclude all time effects.

D. The treatment must be randomly assigned by the researcher.

Correct Answer: B. The groups should have plausibly followed parallel trends before the intervention.

Question 4

A researcher interviews employees after a statistical study finds that ethics committees reduce controversies. What is the main value of this sequential design?

A. It eliminates the need for statistical analysis.

B. It can help explain the mechanisms behind the observed relationship.

C. It guarantees external validity.

D. It converts an association into a proven causal effect.

Correct Answer: B. It can help explain the mechanisms behind the observed relationship.

Question 5

Which practice most directly reduces the risk of HARKing?

A. Adding more control variables after reviewing results.

B. Preregistering hypotheses and analysis plans before examining outcomes.

C. Using only statistically significant variables.

D. Removing contradictory cases from the sample.

Correct Answer: B. Preregistering hypotheses and analysis plans before examining outcomes.

Question 6

In the Meridian Systems case, why is high predictive accuracy insufficient evidence of ethical algorithmic governance?

A. Accuracy is unrelated to all organizational decisions.

B. Accuracy may coexist with subgroup discrimination, weak accountability, or unremediated harms.

C. Machine-learning systems cannot produce accurate predictions.

D. Ethical governance requires eliminating all human review.

Correct Answer: B. Accuracy may coexist with subgroup discrimination, weak accountability, or unremediated harms.

Question 7

Which research approach is most suitable for examining how governance rules are interpreted and enacted inside organizations?

A. Qualitative interviews and process tracing

B. A market index alone

C. A single cross-sectional performance ratio

D. A purely mechanical voting model

Correct Answer: A. Qualitative interviews and process tracing

Question 8

What is a central governance concern in decentralized organizations?

A. Decentralization automatically eliminates power differences.

B. Formal distributed decision rights may still produce concentration through wealth, expertise, or infrastructure control.

C. Decentralized organizations cannot use technology.

D. All decentralized organizations are legally identical to conventional corporations.

Correct Answer: B. Formal distributed decision rights may still produce concentration through wealth, expertise, or infrastructure control.

Question 9

Which statement best captures systemic risk?

A. It concerns only the failure of an individual firm.

B. It arises when interconnections and common exposures transmit shocks across a broader system.

C. It is identical to ordinary operational risk.

D. It can be eliminated by maximizing each firm’s individual efficiency.

Correct Answer: B. It arises when interconnections and common exposures transmit shocks across a broader system.

Question 10

Which recommendation best reflects responsible AI governance?

A. Rely exclusively on vendor assurances.

B. Evaluate only aggregate model accuracy.

C. Maintain impact assessments, subgroup testing, human escalation, documentation, and remedy mechanisms.

D. Delegate all accountability to the algorithm’s developer.

Correct Answer: C. Maintain impact assessments, subgroup testing, human escalation, documentation, and remedy mechanisms.