4.8.1 - Computing's moral, social and cultural effects
Computing changes who can collect information, make decisions and reach an audience. Learn to explain its opportunities and risks for particular people and communities, and how designers' choices can amplify both.
Stakeholders and competing effects
A computing system can benefit one group while imposing costs on another, so its consequences depend on whose interests we consider.
Stakeholder
A person or group affected by a system or able to influence its development or use.
Users, people represented in the data, employees, owners, advertisers, communities and public authorities may all be stakeholders. A person does not have to use an app to be affected: a photograph uploaded by someone else may identify them, or an employer may make a decision about them using purchased data.
The specification distinguishes individual moral, social ethical, legal and cultural opportunities and risks. These are useful lenses, with overlapping concerns:
| Lens | What to examine | Example in an online recruitment system |
|---|---|---|
| Individual moral | What a person ought to do; honesty, autonomy and responsibility | Should a developer report a known error that wrongly rejects an applicant? |
| Social ethical | Fairness and wellbeing across groups and society | Does fast screening increase access to jobs, or systematically disadvantage a group? |
| Legal | Applicable rights, duties and prohibitions | How must applicants' personal information be handled? |
| Cultural | Languages, practices, values and ways of life | Does the form accept local names and different qualification systems? |
Moral and ethical are closely related terms; the table follows AQA's individual/social wording rather than claiming that they are universally separate subjects. A cultural effect is more specific than saying “people dislike technology”: identify which practice, language or expectation is affected and how.
Digital communication makes sending, searching and copying information fast and inexpensive compared with distributing paper records. A small community group can publish advice to a large audience without owning a printing press. People can collaborate across distance, attend lessons remotely and seek support outside their local area. Those same flows can spread a false claim, expose a private address or enable harassment. Storage and search make an old disclosure discoverable long after the original conversation.
Consider a council moving appointment booking online. Residents can book outside office hours, and staff may spend less time taking routine calls. Someone without reliable internet or the skills to use the form may lose access to the same service. A telephone or staffed booking route can reduce this exclusion, but it requires staff and funding. “Computers save time” misses the distribution of the benefit; “online booking reduces residents' waiting time but can exclude residents without access” explains it.
For a developed consequence, connect system feature → effect on a named stakeholder → why that effect matters. Add conditions when they change the outcome. Remote work may reduce commuting, for example, but depends on suitable connectivity and work that can be performed remotely; it may also weaken informal contact between colleagues. Benefits and risks are possibilities to analyse, not claims that every deployment has the same outcome.
Monitoring behaviour and profiling people
Computing can turn many small observations of a person's behaviour into a detailed profile that influences how they are treated.
Monitoring records activity, such as location readings, page visits, purchases or workplace actions. Devices can collect observations continuously, software can combine records from several services, and automated analysis can look for patterns across many people. Personal information concerns an identifiable person; removing a name does not necessarily prevent identification if a distinctive location trail or other information can reconnect the record to them.
Profiling uses information to infer characteristics, interests or likely behaviour. A shop might combine past purchases and browsing to suggest products. A transport operator might analyse journeys to plan capacity. These can improve useful recommendations or services. A profile can also expose sensitive circumstances or support unequal treatment, and an inference may be wrong. Frequent visits near a hospital do not establish a medical diagnosis: the visitor could work there or visit someone else.
Privacy
A person's interest in controlling access to information about themselves and in avoiding unwanted observation or intrusion.
Privacy and security are related but different. Encryption can protect a stored record from outsiders; it does not settle whether the organisation should collect the record or use it to make a particular decision. A securely stored, unnecessarily detailed tracking history can still intrude into private life.
In a hypothetical workplace, measuring completed tasks might reveal a bottleneck and help staff obtain better equipment. Ranking staff by keyboard activity might penalise someone who spends time planning or helping a colleague. The measure rewards what it counts, which can change behaviour: staff may generate activity rather than improve the quality of work. People who know that their movements or searches are watched may avoid legitimate activities. This chilling effect matters even when nobody has yet been punished.
Explain the mechanism rather than treating all data use as harmful. Keeping only aggregate counts may support capacity planning while reducing individual tracking. Restricting access reduces the number of people able to misuse records. A clear purpose and retention limit reduce the opportunity to reuse old observations for unrelated decisions. Checking inferred profiles and providing a way to challenge a decision can reduce harm from errors. Each safeguard has limits: aggregating a very small group may still reveal a person, and a formal appeal is ineffective if nobody has authority to correct the outcome.
The linked law lesson examines which protections are legal duties in a particular context. Here the question is the consequence: what can be learnt about a person, who can act on that knowledge, and what happens if the knowledge or decision is wrong?
Publication, participation and cultural inclusion
Cheap global publication allows more people to participate, but also changes the reach, persistence and cultural context of their information.
Dissemination means spreading information to an audience. A private message can become a public post, be copied to another service, appear in a search result and reach people far beyond its intended recipients. Deleting the original does not necessarily delete screenshots or copies. An accurate statement may cause harm when its audience changes: sharing a person's location with a friend is different from broadcasting it to strangers. A false statement can gain credibility through repeated sharing even though repetition supplies no new evidence.
Online publication can support community organising, emergency communication and access to education. A speaker of a minority language can share recordings and teaching with a dispersed community. Translation and accessibility tools can help more people participate. However, a platform that accepts only one script or misinterprets a language's expressions can exclude those users. A ranking system that rewards the most widely viewed content may give an already dominant language more visibility, making smaller communities harder to discover. These are cultural effects because they influence whose language and practices remain visible.
The digital divide is unequal access to the devices, connectivity, skills and accessible design needed to use digital services. Access is not simply owning a phone: data charges, unstable connections, disability, language and confidence can all matter. An image-only notice might reach sighted readers but exclude someone using a screen reader. Giving useful text alternatives addresses that mechanism; merely saying that the service is “available to everyone” does not.
A hypothetical community mapping project illustrates conflicting interests. Searchable street photographs could help visitors find businesses and residents give directions. They could also reveal a private meeting, a sensitive location or a building that residents do not want prominently advertised. Different residents may have different views; a community leader's approval does not establish every individual's agreement. Designers should consult people who are pictured, local businesses, minority groups and people with limited access, rather than assuming one culture or one representative speaks for everyone.
Possible safeguards include limiting what is published, obscuring identifying details, offering a workable correction/removal process and retaining an accessible alternative. Obscuring faces may reduce direct identification but leave a recognisable location or clothing; removing one image may leave copied versions. Consultation can reveal an unexpected risk before publication, but should inform a decision rather than serve as a ceremonial endorsement.
The same reasoning applies to work and the environment. Automation can remove repetitive tasks and create new skilled roles while disrupting particular workers' incomes; retraining requires time and access. Remote communication may reduce travel, while devices, manufacturing and data processing still consume resources. A balanced answer identifies the affected activity and conditions rather than asserting that digitisation always creates jobs or always reduces environmental harm.
Design values, responsibility and scale
An algorithm embodies choices about what matters, and deploying it widely increases the consequences of those choices.
Software is not value-free merely because its calculations are automatic. Designers choose the objective, the data, the categories, the thresholds, the defaults and the route for challenging an outcome. A recommendation system that maximises viewing time gives priority to engagement; a system that prioritises verified information or varied viewpoints makes different choices. A form that requires every person to have a family name assumes a particular naming convention. These embedded values can be explicit design decisions or unintended assumptions.
An algorithm can consistently apply an unfair rule. Treating everyone by the same formula does not guarantee fairness if its inputs systematically misrepresent some groups. Historical recruitment records may reflect earlier unequal opportunities; learning the pattern of previously successful applicants can reproduce that inequality. A postcode used as a convenient input might act as a proxy for a characteristic that the developer did not intend to assess. Explain the data-to-decision connection; the mere presence of an algorithm does not prove bias.
Consider an original hypothetical scholarship system. Ranking applicants by the number of voluntary placements rewards experience, but may overlook applicants who had caring responsibilities or could not afford unpaid work. Changing the criterion to recognise relevant achievements in context may improve access, yet requires reliable evidence and consistent decisions. There is no unique fair formula supplied by the computer: people must explain which aims and constraints the design serves.
Computer scientists and software engineers have power because their code can determine access, visibility and opportunities. Responsibility includes checking assumptions and foreseeable effects, consulting affected groups, testing relevant populations and failure cases, documenting limits, protecting information, and monitoring what happens after release. Operators and organisations also make choices about how to use a system. “The computer decided” hides those choices; it does not explain who can investigate or correct an error. Ethical responsibilities and a particular legal liability must be distinguished: not every recommended safeguard is automatically a statutory duty in every system.
Scale changes the stakes. A useful translation improvement can help people across many countries with little additional distribution cost. One faulty rejection rule can exclude many applicants before anyone notices. A small error rate can affect many people when there are millions of decisions. Reusing one algorithm worldwide may also export an unsuitable cultural assumption to communities for which it was never tested.
Testing before release, a limited pilot, monitoring group outcomes, a clear escalation route and the ability to pause or reverse a deployment can reduce risk. None guarantees perfection. Human review needs trained people, enough time and real power to change a result; a person automatically accepting the software's recommendation adds little protection. Updates should be checked because changing a rule can introduce a new problem at the same scale.
Connect the design choice to the decision it produces, then to the people affected. Explain how deployment scale amplifies the benefit or harm and how a specific safeguard interrupts that chain.