Which claim is associated with Langdon Winner's phrase 'artifacts have politics'?
Choose an answer, then check it.
Philosophy & Ethics · Foundations
Technology Ethics
On this page 9 sections
In 30 seconds
Technology ethics Philosophical inquiry into moral questions that arise from the design, development, use, and governance of technologies. Full entry → asks moral questions that arise as technologies are designed, developed, and used. A central idea, associated with thinkers such as Langdon Winner and with design-ethics scholarship, is that technologies are not value-neutral: design choices embody values like privacy, safety, fairness, and access. The field also examines unintended consequences, including the Collingridge dilemma The problem that early in a technology's development society has more control but less knowledge of consequences, while later knowledge grows but control shrinks as the technology becomes entrenched. Full entry →, privacy and consent in data-collecting systems, bias in automated decisions, and who bears responsibility when technology causes harm.
Why this matters
Every technology carries choices someone made: what data to collect, whom a system helps or burdens, what happens when it fails. Technology ethics makes those choices visible and gives us language to evaluate them. It matters academically because it connects ethical theory to engineering, policy, and daily life; practically because designers, operators, and regulators are making these decisions now; and educationally because questions about surveillance, algorithmic fairness, and accountability are public controversies students will keep meeting. Studying technology ethics does not produce an automatic answer for every case, but it turns vague unease into questions that can be argued about with evidence and reasons.
The college version
A domain of applied ethics
Technology ethics brings philosophical analysis to the moral questions that technologies raise in their design, development, use, and governance. Like bioethics or business ethics, it is a domain of applied ethics: it brings general moral considerations — consequences, rights, fairness, responsibility — to a particular kind of human practice. The Stanford Encyclopedia of Philosophy characterizes computer and information ethics as an applied ethics that studies the social and ethical impacts of information and communication technology, comparable to established fields such as medical ethics and business ethics. An ordinary question illustrates the domain: should a messaging app collect location data? It is a moral question, but answering it requires understanding how the technology actually works.
Technologies are not value-neutral
A central idea in the field is that technologies are not value-neutral: design choices embody values. Langdon Winner argued that "artifacts have politics" (1980): some technologies embody specific forms of power and authority, or are strongly compatible with certain social and political relations, and technologies may become political through the value system and structure of power reflected in how they were designed and deployed. Value-sensitive design aims to integrate values of ethical importance into engineering design systematically. These claims are contested: the Value-neutrality thesis The contested view that neither technology as such nor individual technologies or artifacts incorporate values. Full entry → — that neither technology as such nor individual technologies incorporate values — has defenders, and the SEP presents the dispute as open. The point is not that every design decision is secretly political, but that neutrality cannot simply be assumed: a design can make privacy, safety, fairness, or access easier or harder to protect.
Unintended consequences and the Collingridge dilemma
New technologies produce consequences no one predicted, which makes governing them awkward. David Collingridge (1980) captured the problem in what is now called the Collingridge dilemma: early in a technology's development, society has more control over outcomes but less capacity to predict them; later, knowledge grows, but control shrinks because the technology is more strongly embedded in society and investors and developers have vested interests. Control is hard early because we do not know what to control, and expensive late because changing an entrenched system costs more. Collingridge proposed making innovation more corrigible and reversible, and making developers and policymakers more accountable for technological choices. Design ethics has its own limit here: critics note that the ethical ramifications of a technology depend not only on its design but on its use, operation, maintenance, and retirement, so no design phase can guarantee an ethical outcome.
Privacy, consent, and surveillance
Data-collecting technologies raise recurring concerns about privacy. The SEP's survey of privacy and information technology observes that such technologies can reduce individuals' control over their personal data and open the possibility of negative consequences. Consent is the standard protection, but it is imperfect: consent requests interfere with task flows, and users may simply click them away. Surveillance is a further documented concern, and the literature includes the term "surveillance capitalism" for business models built on monitoring and predicting behavior. Power asymmetries run through all of this: those who collect data hold power over data subjects.
Automated decisions and bias
Automated systems increasingly make or inform decisions about loans, housing, hiring, and benefits. Researchers at NIST distinguish three sources of bias in such systems: statistical bias, prevalent in large-scale datasets; human bias, which enters through data selection, curation, and annotation; and Systemic bias Bias rooted in institutional norms, practices, and history rather than in a dataset alone; also called institutional or historical bias. Full entry →, also called institutional or historical bias, rooted in institutional norms, practices, and history. Documented concerns about fairness are especially prominent in consumer finance, housing, and employment. Two points matter for ethics. First, bias need not be deliberate: a system can inherit skew from the data it was trained on or from the practices it was built to reproduce. Second, "the data" is not neutral either — someone chose which data to collect, which variables to keep, and which outcomes to optimize. The detailed ethics of artificial intelligence is its own topic; here the point is that automated decisions deserve scrutiny wherever they are deployed.
Responsibility and accountability
When an automated system harms someone, who is responsible? A common difficulty is the Problem of many hands The difficulty of assigning individual responsibility in collective settings where knowledge and control are distributed across many actors. Full entry →, a term coined by Dennis Thompson (1980): in collective settings where knowledge and control are distributed across many actors, assigning individual responsibility is hard. The designer chose the model, the operator chose the data, the deployer chose the deployment, and no one person did everything. Philosophers also discuss a "Responsibility gap Situations in which it may be difficult or impossible to hold humans blameworthy for the actions of automated learning systems (Matthias 2004). Full entry →" (Matthias 2004): for systems that learn and adapt, it may be difficult or impossible to hold any human blameworthy, though some philosophers argue responsibility is better understood as distributed across human-technology collaborations. One practical answer is professional: computing professionals apply codes of ethics and standards of good practice. Technology ethics does not settle who must pay in every case; it insists that the question be asked of every actor with a hand in the system.
The limits of technology ethics
Technology ethics clarifies questions without promising final answers, and the surrounding law is not keeping pace. NIST reports that there is currently no uniformly applied approach among regulators and courts to measuring impermissible bias, and the SEP describes declining clarity and agreement on privacy as creating problems for law, policy, and ethics. Regulations lag because technologies move faster than legislation, and different jurisdictions answer the same question differently. The practical lesson is that ethical analysis and institutional design belong at every stage — early, when change is still cheap, and late, when the stakes are visible — and that honest disagreement about values is part of the subject, not a failure of it.
Eli explains
The same idea, in plain words
Explain it like I’m 10
A smartphone is not just a piece of glass and metal. Every feature was chosen by someone: the app that asks for your location, the feed that decides what you see first, the software that scores a rental application. Technology ethics is the habit of asking who made those choices, what values they built in, and who gets helped or hurt. It starts from the idea that tools are not morally blank. A camera on every street corner is not neutral: it changes who feels watched and who feels safe. The tricky part is timing. When a technology is new, we can still shape it, but we do not yet know what will go wrong. Once it is everywhere, we know the problems, but changing it is expensive and hard. That is why ethics has to sit inside design and policy from the beginning, not arrive after the damage.
Picture it like this
Think of a new technology as a river. Early on, near the source, the water is shallow and a few sandbags can redirect it cheaply. Farther downstream it is wide, deep, and lined with houses, factories, and farms; changing its course then means moving a town. The Collingridge dilemma is the ethics version of that: easy to shape before you understand it, hard to reshape after you do.
Where the picture stops working
A river has no preferences, while technologies are designed by people with values, and people can deliberate and be held accountable. The river analogy also suggests one correct channel, but societies can legitimately disagree about which consequences matter most and how to weigh them.
Worked example
A landlord starts using an automated service that scores rental applications from historical rental and payment data. The system rejects an applicant who actually qualifies; the applicant suspects the score is wrong. The technology-ethics questions line up quickly. What values are built into the scoring design — speed and convenience, or fairness and access? Where could bias enter: the historical data (statistical), the choice of variables (human), or the industry's past practices (systemic)? And if the applicant is harmed, who answers for it — the landlord who deployed the service, the developer who built it, or the company that supplied the data? The Collingridge dilemma adds a governance question: the service is already in use across many properties, so changing it now is costlier than shaping it before launch. None of these questions has an automatic answer, but each is now on the table.
Key takeaway
Technologies are not value-neutral: they embody values, produce unintended consequences, and distribute responsibility across many actors, so ethical reflection belongs at every stage of design, use, and governance.
Quick check
3 questions here, of 5 in this lesson’s practice set. Answers stay hidden until you check.
According to the Collingridge dilemma, why is society's control over technological development difficult?
Choose an answer, then check it.
A university's automated admissions-review flag identifies applicants from one region far more often than others, and the skew tracks the institution's own past admissions practices and records. Which NIST bias category best describes this pattern?
Choose an answer, then check it.
Study tools & related lessonsYou’ll learn to · Common mistakes · Easily confused · Key vocabulary · Related
You’ll learn to
- Define technology ethics as the domain of applied ethics concerned with the design, development, and use of technologies.
- Explain the claim that technologies are not value-neutral, attributing it to design-ethics scholarship such as Langdon Winner's 'artifacts have politics'.
- State the Collingridge dilemma and explain why it makes governing new technologies difficult.
- Describe documented concerns about privacy, consent, and surveillance raised by data-collecting technologies.
- Distinguish statistical, human, and systemic sources of bias in automated decision-making as categorized in NIST research.
- Analyze how responsibility for harm caused by an automated system can be distributed across designers, operators, and deployers.
Common mistakes
Assuming technology is neutral and only the people using it are good or bad.
Design choices embody values: Winner's 'artifacts have politics' and value-sensitive design treat the artifact itself as ethically significant, and the value-neutrality thesis is contested, not settled.
Waiting until a technology causes problems and then fixing it.
The Collingridge dilemma says later control is harder and costlier because the technology is entrenched and interests are invested; reflection belongs early as well as late.
Blaming 'the computer' or a single person when an automated system harms someone.
Knowledge and control are usually distributed across designers, operators, and deployers — the problem of many hands — so responsibility must be traced through all of them.
Treating bias in automated systems as always deliberate prejudice by programmers.
NIST distinguishes statistical, human, and systemic bias, and much of it enters through data and institutional history without anyone intending it.
Assuming one country's rules settle the ethics of a technology everywhere.
There is no uniformly applied international approach to questions like measuring bias, and regulation generally lags behind technology.
Easily confused
Technology ethics vs. Applied ethics in general
Technology ethics is a domain of applied ethics focused on the design, development, and use of technologies; applied ethics supplies the general method (question, facts, considerations, alternatives) that the domain uses.
Value-neutrality thesis vs. Values-in-design view
Contested positions: the thesis holds that technologies incorporate no values, while the critique (Winner; value-sensitive design) holds that artifacts embody power, authority, and the values of their design and deployment.
Statistical bias vs. Systemic bias
Statistical bias comes from patterns in large-scale datasets, while systemic (institutional or historical) bias is rooted in institutional norms, practices, and history that the system reproduces.
Assigning blame to one actor vs. Problem of many hands
Looking for a single responsible individual versus recognizing that knowledge and control are distributed across many actors, which complicates individual ascription of responsibility.
Key vocabulary
- Technology ethics
- Philosophical inquiry into moral questions that arise from the design, development, use, and governance of technologies.
- Value-neutrality thesis
- The contested view that neither technology as such nor individual technologies or artifacts incorporate values.
- Values in design
- Approaches such as value-sensitive design that aim to integrate values of ethical importance into engineering design in a systematic way.
- Collingridge dilemma
- The problem that early in a technology's development society has more control but less knowledge of consequences, while later knowledge grows but control shrinks as the technology becomes entrenched.
- Informed consent
- Agreement to data processing given after being informed; treated in the privacy literature as an imperfect protection because consent requests are often clicked away.
- Algorithmic bias
- Systematic skew in the outputs of automated decision systems; researchers distinguish statistical, human, and systemic sources.
- Systemic bias
- Bias rooted in institutional norms, practices, and history rather than in a dataset alone; also called institutional or historical bias.
- Problem of many hands
- The difficulty of assigning individual responsibility in collective settings where knowledge and control are distributed across many actors.
- Responsibility gap
- Situations in which it may be difficult or impossible to hold humans blameworthy for the actions of automated learning systems (Matthias 2004).
Sources & references
- Philosophy of Technology — Stanford Encyclopedia of Philosophy, Metaphysics Research Lab, Stanford University
- Privacy and Information Technology — Stanford Encyclopedia of Philosophy, Metaphysics Research Lab, Stanford University
- Towards a Standard for Identifying and Managing Bias in Artificial Intelligence (NIST Special Publication 1270) — National Institute of Standards and Technology, U.S. Department of Commerce
- Computer and Information Ethics (Summer 2020 Edition) — Stanford Encyclopedia of Philosophy, Metaphysics Research Lab, Stanford University
- Applied Ethics — Internet Encyclopedia of Philosophy
EliExplains lessons are original prose written from the open, credible references above. See Copyright & Licensing.
Researched 2026-08-21
Educational content only. It is not medical, legal or professional advice. Found an error? Tell us.