Ethical Theories and Responsible Business Decision-Making

Introduction

Business analytics enables organisations to collect, combine, predict and act on information about customers, employees, suppliers and communities. These capabilities create value, but they also create ethical questions. A credit-scoring model may improve lending efficiency while excluding applicants from disadvantaged communities. A retailer may personalise prices in ways that increase revenue but undermine fairness. An employer may use productivity analytics to identify operational problems, yet create intrusive surveillance and psychological pressure.

Data governance provides structures for data quality, accountability, access, security and compliance. Ethical reasoning addresses a different but complementary question: what ought an organisation to do when several legitimate interests conflict? Legal compliance is necessary, but it is not always sufficient. An action may be lawful yet deceptive, exploitative, discriminatory or inconsistent with an organisation’s responsibilities to affected people.

This chapter introduces seven perspectives for analysing data-intensive business dilemmas: utilitarianism, deontology, rights-based ethics, justice ethics, virtue ethics, stakeholder theory and social-contract theory. Each highlights different features of a decision. Responsible business decision-making does not involve selecting one theory mechanically. It involves identifying affected parties, clarifying duties and rights, assessing consequences, examining fairness, and deciding whether the practice can be publicly justified.

Learning Objectives

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

1. Explain the central claims of seven ethical perspectives.
2. Distinguish legal compliance from ethical responsibility.
3. Apply ethical theories to data collection, analytics, artificial intelligence and automated decision-making.
4. Compare competing ethical arguments and identify conflicts among values.
5. Use a structured process to recommend responsible business action.
6. Evaluate the governance, professional and organisational implications of ethical decisions.

1. Ethics, Data and Business Decisions

Ethics is the systematic study of principles concerning right action, human welfare, duties, character and justice. Business ethics applies these principles to decisions made by organisations and their members. Data ethics focuses particularly on the creation, collection, analysis, sharing and use of data, including the effects of automated systems.

A data-intensive decision usually has at least four dimensions:

• Purpose: Why is the data being collected or used?
• Power: Who controls the data, model or decision process?
• Impact: Who benefits, who bears risks, and who may be excluded?
• Accountability: Who can explain, challenge and correct the decision?

For example, a bank’s fraud model may be technically accurate overall but produce more false accusations for a particular demographic group. The issue is not only model performance. It also concerns explanation, due process, equal treatment, privacy, institutional responsibility and the distribution of errors.

Ethics, law and compliance

Law establishes enforceable minimum standards. Ethics asks whether conduct is defensible even when no rule clearly prohibits it. Corporate policies and professional codes may exceed legal requirements, but they are only meaningful when supported by incentives, oversight and organisational culture.

A responsible analyst therefore asks at least three questions:

1. Is the action legally permitted?
2. Is it ethically justified?
3. Can the organisation explain and defend it to affected stakeholders?

These questions may produce different answers. Consent obtained through a long, confusing notice might satisfy a formal requirement while failing to provide meaningful understanding. Similarly, a predictive model may be statistically effective yet ethically unacceptable if people cannot contest high-stakes decisions.

2. Utilitarianism: Evaluating Overall Consequences

Utilitarianism judges an action by its consequences. In its classical form, an action is morally preferable when it produces the greatest overall balance of benefits over harms for all affected parties. Benefits may include safety, access, convenience, health and economic value. Harms may include discrimination, privacy loss, financial injury, anxiety and loss of autonomy.

A utilitarian analysis should not count only immediate financial returns. It should consider:

• the number of people affected;
• the magnitude and duration of benefits and harms;
• the probability of each outcome;
• whether harms fall disproportionately on vulnerable groups;
• direct and indirect effects; and
• short-term and long-term consequences.

For instance, sharing anonymised health data may support valuable medical research. A utilitarian assessment would consider research benefits, re-identification risks, effects on public trust and whether safer alternatives are available.

Strengths and limitations

Utilitarianism encourages organisations to consider broad social consequences rather than focusing only on shareholders or immediate gains. It is useful for public policy, resource allocation and risk prioritisation. However, a purely aggregative approach can justify sacrificing a minority if doing so produces a larger total benefit. It may also treat privacy, dignity or freedom as variables that can simply be exchanged for money or efficiency. Consequences are often uncertain, and organisations may overestimate benefits while discounting diffuse harms.

A responsible utilitarian analysis should therefore include safeguards for vulnerable people, uncertainty analysis and non-negotiable constraints derived from rights and duties.

3. Deontological Ethics: Duties, Rules and Respect

Deontological ethics evaluates actions according to duties, principles and rules rather than only outcomes. Immanuel Kant’s influential formulation holds that people must be treated as ends in themselves, not merely as means to an organisation’s objectives. In data governance, this means that individuals should not be manipulated, deceived or used merely as sources of exploitable information.

Relevant duties may include:

• telling the truth about data practices;
• keeping promises and agreements;
• protecting confidential information;
• avoiding deception and coercion;
• applying rules consistently; and
• accepting responsibility for foreseeable actions.

Suppose an application collects location data for navigation but secretly uses it to infer medical conditions and sell targeted advertising. Even if the practice increases revenue and users receive convenient services, it may violate duties of honesty and respect because the organisation has exceeded the understood purpose of collection.

Strengths and limitations

Deontology protects individuals from being sacrificed for aggregate benefits. It supports clear professional obligations, such as confidentiality and non-maleficence. Its difficulty is that duties may conflict. A company may have a duty to protect customer confidentiality and a duty to disclose information that could prevent serious harm. Rules can also be rigid when exceptional circumstances arise. Practical decision-making therefore requires interpretation, prioritisation and documentation of competing duties.

4. Rights-Based Ethics: Protecting Human Claims

Rights-based ethics asks whether a decision respects legitimate claims held by individuals or groups. Rights may be legal, moral or constitutional, depending on the context. In data-intensive business, relevant rights include privacy, equality, freedom of expression, security, access to information, due process and the ability to challenge significant decisions.

A rights-based approach differs from utilitarianism because it treats some protections as constraints on what may be done, even when violating them could create aggregate benefits. It also differs from deontology in emphasis: deontology focuses on duties attached to action, while rights-based reasoning focuses on the claims and protections of affected persons.

Important questions include:

• Whose rights are affected?
• Is the interference necessary for a legitimate purpose?
• Is it proportionate to that purpose?
• Is the person informed and able to exercise meaningful choice?
• Can the person access, correct or challenge the relevant data or decision?

In automated recruitment, an applicant’s right to equal treatment may be threatened by historical data that encodes gender or racial disadvantage. A right-based response may require removing inappropriate variables, testing outcomes, providing notice and establishing human review.

Rights are not always absolute. Privacy may be limited in carefully justified circumstances, such as preventing serious fraud. The ethical issue is whether the limitation is necessary, proportionate, transparent and subject to safeguards.

5. Justice and Fairness: Distribution, Procedure and Recognition

Justice ethics examines whether benefits, burdens and opportunities are distributed fairly. It also considers whether decision procedures are impartial and whether social identities and experiences are properly recognised.

Three dimensions are especially relevant:

1. Distributive justice: Who receives benefits and who bears costs?
2. Procedural justice: Are decisions made through consistent, transparent and contestable procedures?
3. Recognitional justice: Do systems respect the identity, circumstances and lived experiences of different groups?

An insurance algorithm might reduce premiums for customers predicted to be low risk. If the model uses neighbourhood data correlated with historical underinvestment, residents of certain communities may face higher prices despite having similar individual behaviour. This raises distributive concerns. If the company refuses to explain or review the result, it also raises procedural concerns.

Fairness is not identical to treating everyone identically. Equal treatment can reproduce disadvantage when people begin from unequal conditions. Conversely, treating groups differently may be justified when it corrects systematic barriers or addresses relevant differences. Analysts must define the fairness principle being used and recognise that statistical fairness criteria can conflict. For example, equal false-positive rates and equal positive predictive values may not both be achievable when groups have different base rates.

Justice analysis should therefore be explicit about the affected population, the relevant comparison, the allocation rule and the remedy for errors.

6. Virtue Ethics: Character and Organisational Culture

Virtue ethics focuses on the character and practical judgment of the decision-maker rather than only rules or outcomes. Important virtues in analytics include honesty, fairness, courage, humility, prudence, empathy, accountability and respect.

A virtuous analyst does not ask only, “Can we deploy this model?” but also, “What would a trustworthy professional do?” Humility is important because data are incomplete and models are uncertain. Courage may be required to challenge a profitable but harmful practice. Prudence involves balancing innovation with foreseeable risk. Integrity requires consistency between public commitments and actual conduct.

Virtue ethics highlights organisational culture. A company that rewards rapid deployment while punishing employees who report risks may have formal ethical policies but an unethical practical culture. Conversely, a culture that encourages challenge, records dissent and rewards responsible escalation can improve decisions even where policies are incomplete.

The limitation of virtue ethics is that virtues can be interpreted differently and may not provide a precise decision rule. It is most effective when combined with clear governance, evidence and accountability mechanisms.

7. Stakeholder Theory: Considering Affected Relationships

Stakeholder theory holds that organisations have responsibilities to groups that affect or are affected by organisational activity. Stakeholders may include customers, employees, suppliers, investors, regulators, communities, citizens and the natural environment.

A stakeholder analysis identifies:

• the interests of each group;
• the group’s power and vulnerability;
• the benefits and risks created by the decision;
• conflicts among stakeholder interests; and
• mechanisms for consultation, remedy and accountability.

Stakeholder theory corrects narrow shareholder-only reasoning. A platform introducing workplace-monitoring software, for example, must consider not only the employer’s efficiency goals but also employee dignity, job security, data security, managerial accountability and the effects on workplace trust.

Stakeholder theory does not mean that every demand must be accepted. It requires reasoned consideration and fair justification. Stakeholder participation should also be meaningful. Asking for opinions after a decision is final is consultation in form but not necessarily in substance.

8. Social-Contract Theory: Legitimacy and Public Justification

Social-contract theory asks what rules people would reasonably accept as the basis for cooperative social life. In business, organisations operate under an implicit social licence: society permits them to use resources, collect information and exercise economic power in return for responsible conduct.

A data practice may therefore be assessed by asking:

• Would affected people accept the practice if they understood it clearly?
• Could the practice be publicly disclosed without deception or embarrassment?
• Does it support fair cooperation or exploit dependence?
• Are the rules applied consistently to powerful and less powerful parties?
• What institutions are needed to monitor and correct abuse?

The “publicity test” is particularly useful: would the organisation be willing to explain its practice to customers, employees, regulators and an independent journalist? A negative answer does not automatically prove wrongdoing, but it is a strong signal that further scrutiny is required.

Social-contract reasoning is valuable for platform governance, public-private data sharing and artificial intelligence because these systems shape social opportunities beyond individual transactions. Its limitation is that “reasonable acceptance” may be contested, especially in pluralistic societies. Deliberation must therefore include diverse voices and protections against majority domination.

9. Comparing the Ethical Perspectives

These theories should be used as complementary lenses. A strong analysis identifies where they converge and where they conflict. For example, collecting detailed mobility data may generate significant public benefits under utilitarian reasoning, but rights-based and deontological reasoning may reject collection without meaningful consent. A responsible recommendation must address both the benefits and the constraints.

10. A Framework for Responsible Decision-Making

The following process can be used by analysts, managers and governance committees.

Step 1: Define the decision and purpose

State what is being proposed, why it is needed and what alternatives exist. Avoid vague claims such as “improve efficiency.” Specify the operational objective and the people affected.

Step 2: Map data, power and stakeholders

Identify data subjects, data owners, processors, decision-makers, downstream recipients and people who may be indirectly affected. Document data sources, sensitivity, retention and potential inferences.

Step 3: Identify duties, rights and constraints

Review law, contracts, professional obligations, organisational policies and legitimate expectations. Determine whether the proposal affects privacy, equality, autonomy, confidentiality, safety or due process.

Step 4: Assess consequences and fairness

Estimate likely benefits and harms, including uncertain and indirect effects. Test for disparate impacts and examine who bears errors. Consider reversibility: a mistaken recommendation may be correctable, while a leaked sensitive dataset may create permanent harm.

Step 5: Consult and challenge

Seek input from affected stakeholders, domain experts, privacy and security professionals, legal advisers and independent reviewers. Create channels for dissent and escalation.

Step 6: Select safeguards and alternatives

Use data minimisation, purpose limitation, access controls, encryption, retention limits, human review, model monitoring, explanation, appeal and remediation. Ask whether a less intrusive method can achieve the same purpose.

Step 7: Decide, document and monitor

Record the ethical reasoning, assumptions, dissenting views and approval authority. Assign responsibility for monitoring outcomes. Establish trigger conditions for suspension or redesign.

Step 8: Provide remedy

Responsible governance includes correction, appeal, compensation where appropriate and learning from failure. An organisation that cannot remedy foreseeable harm should reconsider whether it is ready to deploy the system.

11. Illustrative Case: The Workforce Insight System

Background

A fictional logistics company, Northstar Distribution, proposes an analytics system that combines scanner records, delivery times, vehicle location, attendance data and employee mobile-device signals. Management claims that the system will reduce accidents, improve route planning and identify training needs.

Problem

The proposal would create a performance score for each driver. Low scores could affect schedules, bonuses and continued employment. Employees are told that the system is intended for safety, but the company also plans to use it for productivity ranking. Some workers are concerned that location monitoring will continue outside working hours and that the model will penalise drivers serving difficult routes or caring for family members.

Ethical analysis

Utilitarianism: The system may reduce accidents and improve service for many customers. However, benefits must be weighed against stress, privacy loss, inaccurate scores and possible job loss. The company should quantify expected safety benefits and examine less intrusive alternatives.

Deontology: If employees were told that monitoring was for safety, using the same data secretly for discipline violates honesty and purpose-related duties. The organisation also has duties to protect sensitive location information and investigate errors before imposing sanctions.

Rights-based ethics: Continuous location tracking may interfere with privacy. Employees also have a claim to fair treatment and an opportunity to challenge consequential decisions. Monitoring outside work would require particularly strong justification and safeguards.

Justice: Drivers on congested routes, serving rural areas or working with older equipment may receive lower scores for reasons unrelated to effort. A uniform score could distribute disciplinary burdens unfairly. The company should test group and route-level effects and avoid using unvalidated scores as the sole basis for employment action.

Virtue ethics: A trustworthy employer would be transparent, cautious and willing to acknowledge uncertainty. A culture that treats the score as objective merely because it is numeric would reflect poor professional judgment.

Stakeholder theory: Management, drivers, customers, safety officers, unions, regulators and communities have distinct interests. Drivers should participate in defining acceptable monitoring and reviewing the scoring system.

Social-contract theory: Workers may accept safety monitoring as part of a fair employment relationship, but they are unlikely to accept undisclosed surveillance or unreviewable decisions. The system’s legitimacy depends on clear limits and accountability.

Approach and solution

Northstar should separate safety analytics from disciplinary analytics, disclose each purpose, limit collection to working time, and prohibit use of phone data unrelated to the stated objective. It should conduct a privacy and equality impact assessment, test the model across routes and worker circumstances, and provide human review and appeal. Scores should initially support coaching rather than automatic punishment. An employee-management committee should review performance, complaints and unintended effects.

Outcome

In this illustrative case, the company’s revised design produces a narrower but more defensible system. It retains route and vehicle data needed for safety while removing off-duty location signals. Early monitoring finds that weather and route complexity strongly affect scores, so the model is adjusted and managers are trained not to treat predictions as facts. The company accepts lower short-term savings in exchange for greater trust, better validity and reduced ethical risk.

Learning points

1. A beneficial purpose does not justify unlimited data collection.
2. Transparency must cover purposes, consequences and decision processes.
3. Numerical scores do not eliminate human judgment or responsibility.
4. Fairness requires examining context, not merely comparing overall accuracy.
5. Stakeholder participation and remedy are components of responsible design.

12. Technology, AI and Contemporary Governance

Artificial intelligence intensifies familiar ethical problems because systems can operate at scale, adapt over time and obscure the relationship between data and decisions. Generative AI introduces additional concerns about confidential information, fabricated outputs, intellectual property, authorship and human oversight.

Responsible AI governance should include documented purpose, data provenance, suitability testing, security controls, bias assessment, explainability appropriate to the decision, human oversight and post-deployment monitoring. Human involvement must be meaningful: simply placing a person at the end of an automated process does not create accountability if the person cannot understand or challenge the output.

Contemporary governance increasingly emphasises risk-based regulation, impact assessments, algorithm registers, independent audits and lifecycle accountability. Organisations should monitor regulatory developments in the jurisdictions where they operate. They should also distinguish between technical uncertainty and ethical uncertainty. A model may be statistically reliable while the underlying purpose remains unjustified.

13. Professional and Managerial Implications

Business analysts are not merely technical service providers. They help define objectives, select measures and translate organisational values into operational systems. Their responsibilities include:

• questioning unjustified data collection;
• documenting assumptions and limitations;
• avoiding misleading visualisations and claims;
• protecting confidential information;
• escalating material risks;
• testing for unequal effects;
• preserving audit trails; and
• communicating uncertainty to decision-makers.

Managers should align incentives with responsible conduct. Ethical review should occur early, not only after a controversy. Governance committees need sufficient independence, expertise and authority. Organisations should reward employees who identify risks rather than treating criticism as disloyalty.

Challenges and Limitations

Ethical theories do not produce automatic answers. Stakeholders may disagree about relevant harms, acceptable risk and the meaning of fairness. Consequences are uncertain, rights may conflict, and consultation can be costly. Ethical analysis can also be captured by powerful actors if affected communities lack representation.

Other limitations include automation bias, poor data quality, proxy discrimination, institutional inertia and the difficulty of measuring dignity or trust. A documented assessment is not evidence of ethical conduct if the decision was predetermined or if safeguards are ignored. Responsible decision-making is therefore iterative: organisations must revisit assumptions as evidence, technology and social expectations change.

Best Practices

1. Define the legitimate purpose before collecting data.
2. Collect only what is necessary and retain it for no longer than justified.
3. Separate compatible and incompatible purposes.
4. Use multiple ethical lenses rather than relying on efficiency alone.
5. Test outcomes across relevant groups and contexts.
6. Provide understandable notice, meaningful choice and effective appeal.
7. Keep humans accountable for high-impact decisions.
8. Document assumptions, uncertainty, approvals and dissent.
9. Monitor real-world outcomes after deployment.
10. Provide correction, remedy and a clear route for escalation.

Chapter Summary

Ethical theories provide complementary ways to evaluate data-intensive business decisions. Utilitarianism focuses on overall consequences; deontology on duties and rules; rights-based ethics on dignity and protected claims; justice ethics on fair distribution, procedures and recognition; virtue ethics on professional character and organisational culture; stakeholder theory on affected relationships; and social-contract theory on legitimacy and public justification.

No single perspective is sufficient. Responsible business decision-making combines legal and governance requirements with analysis of purpose, power, consequences, rights, fairness, stakeholder interests and accountability. In analytics and artificial intelligence, responsible practice requires transparency, proportionality, data minimisation, human oversight, monitoring and remedy. The central professional obligation is not simply to produce accurate outputs, but to help ensure that organisational decisions are defensible, fair and worthy of trust.

Key Terms

Ethics: Systematic study of right action, duties, values and human welfare.

Business ethics: Application of ethical reasoning to organisational conduct and decisions.

Data ethics: Ethical analysis of data collection, analysis, sharing and use.

Utilitarianism: View that actions should produce the greatest overall balance of benefits over harms.

Deontology: Ethical approach centred on duties, rules and principles.

Rights-based ethics: Approach that protects legitimate claims such as privacy, equality and due process.

Justice: Concern with fair distribution, procedures and recognition.

Distributive justice: Fair allocation of benefits, burdens and opportunities.

Procedural justice: Fairness and transparency of decision-making procedures.

Virtue ethics: Approach focused on character, practical judgment and moral virtues.

Stakeholder: Person or group affected by, or able to affect, an organisation.

Social contract: Implicit basis of legitimate cooperation between organisations and society.

Purpose limitation: Restricting data use to specified, legitimate purposes.

Data minimisation: Collecting no more data than is necessary for the purpose.

Proportionality: Requirement that an interference or risk be justified by the legitimate objective and not excessive.

Algorithmic bias: Systematic and unjustified disadvantage produced by data, models or their use.

Human oversight: Meaningful human responsibility and intervention in automated systems.

Remedy: Correction, appeal, compensation or other response to harm or error.

Review Questions

Short Answer Questions

1. How does ethical responsibility differ from legal compliance?
2. What is the central question asked by utilitarianism?
3. Distinguish distributive justice from procedural justice.
4. Why is meaningful human oversight important in high-impact automated decisions?
5. What is the purpose of a stakeholder analysis?

Long Answer Questions

1. Compare utilitarian, deontological and rights-based approaches to the use of customer data for targeted advertising.
2. Explain how virtue ethics and organisational culture influence responsible analytics.
3. Discuss why treating everyone identically does not always produce a just outcome.
4. Explain the relationship between social licence, public justification and stakeholder trust.
5. Describe and evaluate the eight-step framework for responsible decision-making.

Analytical / Application Questions

1. A university uses an algorithm to identify students at risk of dropping out, but does not tell students how the system works. Analyse the proposal using at least four ethical perspectives and recommend safeguards.
2. A retailer wants to use loyalty-card data to offer different prices to different customers. Identify the principal benefits, risks, rights and fairness concerns, then develop a responsible governance approach.
3. An employer proposes using generative AI to summarise employee communications and predict misconduct. Construct an ethical assessment covering purpose, stakeholder interests, privacy, consequences, procedural fairness and remedy.

Multiple Choice Questions with Answers

Question 1

Which ethical perspective evaluates an action primarily by its overall benefits and harms?

A. Utilitarianism

B. Deontology

C. Virtue ethics

D. Social-contract theory

Correct Answer: A. Utilitarianism

Question 2

A company uses location data for advertising after telling users it would be used only for navigation. Which concern is most directly raised by deontological ethics?

A. The model may be computationally slow

B. The company may have violated duties of honesty and purpose consistency

C. The data may have low commercial value

D. The company may have too many stakeholders

Correct Answer: B. The company may have violated duties of honesty and purpose consistency

Question 3

Which is an example of procedural justice?

A. Ensuring that benefits are distributed equally

B. Ensuring that affected people can understand and appeal a decision

C. Maximising total revenue

D. Rewarding managers for rapid deployment

Correct Answer: B. Ensuring that affected people can understand and appeal a decision

Question 4

What does data minimisation require?

A. Collecting all potentially useful data

B. Collecting only data necessary for a legitimate purpose

C. Deleting data immediately after collection

D. Allowing unrestricted internal access

Correct Answer: B. Collecting only data necessary for a legitimate purpose

Question 5

Which question best reflects rights-based ethics?

A. Will the action maximise aggregate profit?

B. What kind of character does the organisation display?

C. Whose privacy, equality or due-process claims are affected?

D. Which stakeholder has the greatest financial power?

Correct Answer: C. Whose privacy, equality or due-process claims are affected?

Question 6

Why can a purely utilitarian approach be problematic in data governance?

A. It never considers consequences

B. It may justify sacrificing a minority for a larger aggregate benefit

C. It rejects all forms of measurement

D. It requires every decision to be made by a regulator

Correct Answer: B. It may justify sacrificing a minority for a larger aggregate benefit

Question 7

Which practice best demonstrates meaningful human oversight?

A. Automatically approving every model output

B. Having a person present who cannot review or change the decision

C. Giving an accountable reviewer authority, information and time to challenge the output

D. Removing all documentation from the process

Correct Answer: C. Giving an accountable reviewer authority, information and time to challenge the output

Question 8

Stakeholder theory primarily requires an organisation to:

A. Consider only shareholder returns

B. Identify and responsibly address the interests of affected groups

C. Avoid all decisions involving disagreement

D. Replace managers with automated systems

Correct Answer: B. Identify and responsibly address the interests of affected groups

Question 9

Which proposal best satisfies proportionality?

A. Using unlimited surveillance for any possible future purpose

B. Using the least intrusive effective method for a legitimate objective

C. Ignoring privacy whenever a system is accurate

D. Collecting sensitive data because storage is inexpensive

Correct Answer: B. Using the least intrusive effective method for a legitimate objective

Question 10

In the Northstar case, why was separating safety analytics from disciplinary analytics ethically important?

A. It reduced the number of software vendors

B. It aligned data use with disclosed purposes and reduced unjustified employment harm

C. It guaranteed that every prediction would be correct

D. It eliminated the need for employee consultation

Correct Answer: B. It aligned data use with disclosed purposes and reduced unjustified employment harm