Philosophy › Section J · Applied ethics › Topic J.3

Business ethics and emerging technology

Companies decide what is made, who is hired and what data is collected, and engineers build systems used by billions. Business and technology ethics asks what these organisations and their members owe to owners, workers, customers and everyone else their choices touch.

Ethics inside organisations

A business sells goods or services for profit, within laws that governments set. But the law is silent or slow on many choices, so organisations and professions write codes of ethics. These cover concrete matters such as bribery, discrimination, conflicts of interest and whistleblowing. The ACM Code of Ethics asks computing professionals to avoid harm, be honest and respect privacy, and the Software Engineering Code of Ethics, adopted jointly by the ACM and the IEEE Computer Society, asks engineers to approve software only if they believe it is safe and meets its specifications. Codes still leave open what counts as harm.

Shareholder primacy

In a 1970 New York Times Magazine essay, the economist Milton Friedman argued that a firm's one social duty is to raise its profits, while staying within the rules of the game, without deception or fraud. Executives are agents of the shareholders who own the firm. If they spend company money on causes of their own choosing, they are in effect taxing owners, customers or workers without any democratic mandate. Social goals belong to individuals and to governments, which make laws. Critics reply that firms help write the rules they follow, and that law cannot foresee every harm a profitable choice may cause.

Stakeholder theory

R. Edward Freeman's Strategic Management: A Stakeholder Approach (1984) helped popularise the rival view. A stakeholder is anyone who affects or is affected by a firm: employees, customers, suppliers, local communities and shareholders. Managers, on this view, should balance all their interests, not serve owners alone. Supporters argue that workers and towns also invest in a firm, through skills, years of service and public infrastructure, and that a firm depends on their trust. Critics answer that balancing everyone gives managers no clear goal and makes them accountable to no one, since any decision can be defended as serving some stakeholder.

Workers: safety, meaning and global supply chains

Early factory work meant long hours, child labour and danger. Britain's Factory Act of 1833 began legal limits, and the US Occupational Safety and Health Act of 1970 set national safety standards. Adam Smith worried that a fine division of labour made work repetitive and dulled the mind. Norman Bowie (1998), drawing on Kant, argued that firms have a duty to provide meaningful work: freely chosen, paid enough to live independently, and respecting workers as rational agents. Outsourcing raises a related dispute. Critics call low-wage overseas factories exploitative; defenders, including many economists, reply that such jobs often pay better than workers' alternatives and that boycotts can leave them worse off.

Affirmative action

Affirmative action means positive steps to increase the representation of groups historically excluded from jobs and education, from outreach and training to giving weight to race or sex in selection. Defenders argue that it redresses past and continuing discrimination, as Judith Jarvis Thomson held, that it can make selection fairer where bias is entrenched, as Mary Anne Warren argued, and that diversity benefits institutions. Critics argue that it selects people for traits unrelated to merit, wrongs individual applicants who did not cause past injustice, and can stigmatise those it aims to help. In 2023 the US Supreme Court, in Students for Fair Admissions v. Harvard, ruled that Harvard's and North Carolina's race-conscious admissions violated equal protection; the moral debate continues.

Worked example: a supplier's safety report

A clothing firm learns that a cheap overseas supplier blocks fire exits, which local law does not clearly forbid. 1. Shareholder view: Friedman's rules of the game include no fraud or deception, and a deadly fire would threaten the firm's reputation and profits, so fixing the problem may also serve owners. 2. Stakeholder view: the workers are stakeholders whose safety must be weighed directly, not only through profit. 3. Kantian view: knowingly profiting from avoidable risk treats workers merely as means. 4. Compare: the views may agree on acting, but they differ on why, and they diverge when fixing the problem is costly and no one would find out.

Technology, privacy and attention

Technologies are not neutral: their design decides who is watched, what is easy and who benefits, and digital scale multiplies each choice. Helen Nissenbaum's theory of contextual integrity holds that privacy is breached when information flows violate the norms of the context in which it was shared, as when health data given to a doctor is sold to advertisers. The EU's General Data Protection Regulation, applied since 2018, gives people rights over their personal data. Many free services earn money by capturing attention. Critics argue that infinite scrolling and constant notifications manipulate users, especially children, and that dark patterns, such as hidden cancellation steps, trick people; defenders stress user choice and the value of free services.

Algorithms and fairness

Systems trained on historical data can reproduce past discrimination. In 2016 the newsroom ProPublica reported that a risk-assessment tool used in US courts, COMPAS, wrongly flagged Black defendants as high risk more often than white defendants; its maker replied that its scores were equally accurate for both groups. Both claims could be true, because researchers have since shown that when groups have different base rates, several natural definitions of fairness cannot all be met at once. Choosing one is a value judgement. Joy Buolamwini and Timnit Gebru (2018) found that commercial face-analysis systems made far more errors for darker-skinned women than for lighter-skinned men.

Artificial intelligence

John Searle (1980) distinguished weak AI, the view that computers are useful tools for studying the mind, from strong AI, the view that a suitably programmed computer would literally have a mind. His Chinese room argument held that running a program, however clever, is not understanding. Many reject the argument; whether AI systems could ever have interests or moral status remains open. Practical concerns include responsibility for harms, jobs and deepfakes. Nick Bostrom's Superintelligence (2014) warned that a system far more capable than humans might pursue goals misaligned with human values. Some critics reply that such speculative risks distract from present harms such as bias and surveillance. The EU adopted a risk-based AI Act in 2024.

Worked example: face recognition for police

A city considers cameras that match faces against a database of wanted people. 1. Benefits: faster identification of suspects in serious crimes. 2. Accuracy: if error rates are higher for some groups, innocent people from those groups face more wrongful stops, a justice concern. 3. Liberty and privacy: constant identification in public may chill lawful protest and break the norms under which people appear in public. 4. Safeguards: independent accuracy audits, limits to serious crimes, warrants and public oversight. 5. Verdict: a consequentialist weighs crimes solved against errors and chilling effects; a rights-based view may reject mass surveillance even if it works. Several US cities have restricted police use.

Common mistakes to avoid

First, saying Friedman approved of anything profitable: he required obeying the law and avoiding deception and fraud. Second, treating stakeholders as just another word for shareholders: stakeholders include anyone affected. Third, thinking a code of ethics settles hard cases: it states duties but not how to weigh them. Fourth, assuming an algorithm is fair because it ignores race: correlated data can still produce unequal outcomes. Fifth, confusing Searle's strong AI with powerful AI: strong AI is the claim that a program could have a mind.

Key takeaways

Codes of ethics guide conduct but leave hard judgements open. Friedman's shareholder primacy says business should maximise profits within the rules; Freeman's stakeholder theory says managers should balance all affected interests; each faces serious objections. Bowie argues for meaningful work, while outsourcing and affirmative action divide opinion. Contextual integrity explains privacy breaches; attention-driven design raises manipulation worries. Competing fairness definitions cannot all be met when base rates differ. Searle denied that programs understand; Bostrom warned of misaligned superintelligence, while critics stress present harms.

Flashcards

Flashcard

Code of ethics

An organisation's or profession's written standards of conduct

ACM; Software Engineering Code.

Test your understanding

Question 1 of 5

Which statement best matches Friedman's 1970 position?

Question 2 of 5

What is a standard objection to stakeholder theory?

Question 3 of 5

Health details shared with a doctor are sold to an advertiser. On Nissenbaum's theory, why is this a privacy breach?

Question 4 of 5

Why could both sides in the COMPAS dispute make true claims about fairness?

Question 5 of 5

In Searle's sense, what does strong AI claim?

← Previous topicNext topic →