Move fast and break things: tech ethics

• A. Apple
• B. Facebook (Meta) ✓
• C. Microsoft
• D. Google

Why: Mark Zuckerberg popularized the phrase in Facebook's early years, and it appeared in the company's internal culture and early investor materials before being retired in favor of 'Move fast with stable infrastructure' around 2014.

• A. To guarantee bug-free software releases
• B. To prioritize rapid iteration and growth over perfection and caution ✓
• C. To comply with government regulations
• D. To reduce the number of engineers needed

Why: The ethos encouraged engineers to ship quickly, learn from real user feedback, and accept some errors as a cost of rapid innovation and market capture.

• A. Waterfall methodology
• B. Agile software development ✓
• C. Formal verification
• D. Big design up front

Why: The Agile Manifesto (2001) valued responding to change and delivering working software frequently, which aligned with the iterative, fast-shipping culture that later became associated with startups.

• A. The phrase was legally trademarked by another company
• B. Public criticism and the cost of breaking user trust made the phrase look reckless
• C. The company stopped releasing software updates
• D. Engineers requested fewer deadlines

• A. It makes software too expensive to maintain
• B. Products may be released before risks to users, privacy, or society are adequately considered ✓
• C. It eliminates the need for user feedback
• D. It requires too much documentation

Why: Critics argue that prioritizing speed can externalize harms such as privacy violations, misinformation, and discrimination onto users and society, which may only become apparent after widespread deployment.

• A. Platform capitalism
• B. Surveillance capitalism ✓
• C. Attention economy
• D. Data commons

Why: Shoshana Zuboff coined 'surveillance capitalism' to describe companies that extract behavioral data from users and convert it into prediction products sold to advertisers and others.

• A. Unused storage space on company servers
• B. Data collected beyond what is needed to improve a service, used to build predictive products ✓
• C. Profits a company earns from selling hardware
• D. Employee time wasted on unproductive tasks

Why: Behavioral surplus refers to the excess behavioral data companies gather beyond what is needed to provide a service, which is then repurposed to predict and shape user behavior for profit.

• A. Companies should collect as much data as possible to improve accuracy
• B. Only data that is necessary for a specific, stated purpose should be collected and retained ✓
• C. Data should be stored in the smallest file formats available
• D. Users should receive smaller amounts of targeted advertising

Why: Data minimization holds that organizations should limit collection and retention to what is directly relevant and necessary for a defined purpose, reducing privacy risks.

• A. Users rarely read them, and consent is not meaningful when the alternative is refusing an essential service ✓
• B. They are written in languages users cannot understand
• C. They are illegal in most countries
• D. They only apply to government agencies, not companies

Why: Long, complex policies are seldom read, and users often face a take-it-or-leave-it choice, so 'consent' may not reflect a genuine, informed choice.

• A. A bookstore recommending titles based on a customer's recent purchases in-store
• B. A free app tracking users' location and browsing habits across sites to sell highly targeted ad placements ✓
• C. A library lending books to patrons who return them on time
• D. A bank sending a statement by mail each month

Why: The free app monetizes continuous behavioral tracking by selling predictive ad targeting, which is the core mechanism of surveillance capitalism; the other options involve limited, direct transactions without covert behavioral extraction.

• A. A deliberate choice by programmers to favor certain users
• B. Systematic and repeatable errors in a computer system that create unfair outcomes for certain groups ✓
• C. The mathematical error that occurs when a model is trained on too much data
• D. A hardware failure that causes software to produce random results

Why: Algorithmic bias refers to systematic, repeatable errors that produce unfair outcomes, often affecting particular groups. It is not necessarily intentional and often arises from data, design choices, or the context in which a system is used.

• A. Using too few processing cores
• B. Training data that underrepresents certain groups of people ✓
• C. Writing code in a high-level programming language
• D. Running models on cloud servers

Why: When training data underrepresents certain groups, the model learns patterns that work well for the majority and poorly for others. Historical data that reflects past discrimination can also carry bias forward.

• A. Every user receives identical outputs regardless of their inputs
• B. The system is designed so that it never makes any errors
• C. There is no single definition; fairness can be defined in several competing ways, such as equal error rates or equal positive prediction rates across groups ✓
• D. The algorithm is open-source and publicly available

Why: Fairness has no single agreed-upon mathematical definition. Different fairness criteria can conflict with one another, so designers must choose which notion of fairness fits their context and values.

• A. Overall accuracy always hides errors, so accuracy is never meaningful
• B. Accuracy averaged across all users can mask much higher error rates for a minority subgroup ✓
• C. High accuracy means the model must have been trained on synthetic data
• D. Accurate models cannot be audited for fairness

Why: An aggregate accuracy figure can look strong while a smaller group experiences substantially worse performance. Evaluating error rates separately for each subgroup is necessary to detect this kind of disparity.

• A. Removing all demographic data so the model cannot see any group membership
• B. Testing the system on diverse data and evaluating performance across different demographic groups ✓
• C. Increasing the model's size until bias disappears on its own
• D. Deploying the system quickly and relying on user complaints to find problems

Why: Disaggregated evaluation on diverse data reveals disparities that aggregate metrics hide. Simply removing demographic fields often fails because other features can act as proxies for those characteristics, and larger models do not automatically become fairer.

• A. A commitment to slow, heavily regulated product development
• B. A willingness to release products quickly and accept some harm or errors as a cost of innovation ✓
• C. A focus on hardware repair and sustainability
• D. A legal requirement for software testing before launch

Why: The phrase reflects a development philosophy that prioritizes speed and iteration over caution, accepting that some things will break in the process. In the context of misinformation, this mindset can mean harmful content spreads before platforms have time to address it.

• A. Content that is deliberately created to deceive or manipulate audiences
• B. False or inaccurate information shared without intent to deceive ✓
• C. Any opinion that a reader disagrees with
• D. Information that has been removed by a platform

Why: Misinformation is false or inaccurate information regardless of intent. Disinformation is the term typically reserved for false information deliberately spread to deceive, so the question's wording distinguishes unintentional error from intentional manipulation.

• A. They only show content from verified news organizations
• B. They tend to amplify content that provokes strong reactions, which can favor sensational or false claims ✓
• C. They are legally required to promote false content
• D. They prevent users from sharing any content with others

Why: Algorithms optimized for clicks, shares, and comments can reward emotionally charged content, and false or sensational claims often generate high engagement. This creates an incentive structure where misinformation can spread faster than corrections.

• A. Platforms are neutral conduits with no influence over what users see
• B. Platforms design the systems that determine what content is amplified and therefore shape its reach ✓
• C. Users cannot be trusted to read anything online
• D. Platforms have no economic interest in their content

Why: Because platforms design ranking, recommendation, and sharing features, they actively shape which content gets seen. This influence is the basis for arguing they bear some responsibility, in contrast to the view that they are passive hosts.

• A. Adding context labels, reducing algorithmic amplification, or prompting users to read before sharing ✓
• B. Deleting all user accounts that have ever posted an opinion
• C. Disabling all comment sections permanently
• D. Requiring every post to be approved by a government agency

Why: Interventions like fact-check labels, friction before sharing, and demoting unreliable content aim to limit harm while preserving free expression. They are often discussed as middle-ground options between doing nothing and blanket removal.

• A. Salaried staff
• B. Platform workers ✓
• C. Unionized employees
• D. Full-time benefits recipients

Why: Platform workers is the common term for people who earn income through apps and online marketplaces while typically being classified as independent contractors, which often means they lack employee protections such as minimum wage guarantees, overtime pay, and employer-provided benefits.

• A. Automation always reduces prices for consumers and eliminates all job losses
• B. Automation can increase productivity and create new kinds of jobs over time ✓
• C. Automation is illegal in most countries without government approval
• D. Automation only affects workers who have no formal education

Why: A frequent defense of automation is that it raises productivity, which can lower costs and generate new roles, such as jobs in maintaining systems or designing software. Critics note that these benefits are not guaranteed and that displaced workers may not easily transition into the new roles.

• A. Workers setting their own schedules and rates through negotiation
• B. Software systems that assign tasks, monitor performance, and sometimes deactivate workers with limited human oversight ✓
• C. A government agency that regulates wages across all industries
• D. Traditional HR departments using spreadsheets to track employee hours

Why: Algorithmic management describes how platforms use automated systems to allocate work, rate workers, set pay, and discipline or remove them. Ethical concerns arise because workers may have little insight into or ability to appeal these automated decisions.

• A. Contractors always earn more money than employees
• B. The classification can strip workers of protections like minimum wage, overtime, and benefits even when the company controls much of their work ✓
• C. Independent contractors are legally barred from working more than one job
• D. The classification requires workers to pay income tax while employees do not

Why: When a company sets prices, schedules, and performance standards, the worker may be economically dependent in ways resembling employment. Misclassification can therefore deny workers legal protections that would otherwise apply, which is why it is a central issue in labor debates.

• A. Rapid innovation always benefits every worker equally
• B. Speed of deployment can impose costs on workers and communities that are not considered in the rush to scale ✓
• C. Companies should avoid making any changes to existing business models
• D. Labor laws should be updated once every decade

Why: The phrase encourages rapid deployment, but when companies prioritize speed, the harms to workers, such as job displacement, unstable income, and loss of protections, can be externalized onto people who had no say in the decision. This tension is central to tech ethics discussions.

• A. A technical limit on how many devices can connect to the internet at once
• B. The phenomenon where a product becomes more valuable to each user as more people use it ✓
• C. A government rule requiring companies to share their user data
• D. The decline in a company's value after a data breach

Why: Network effects occur when each additional user increases the value of a service for others, as with social media or messaging apps. They tend to favor large incumbents and make it hard for new competitors to gain traction.

• A. A commitment to slow, heavily tested product launches
• B. A prioritization of rapid growth and iteration that can externalize harms onto users and society
• C. A legal requirement that startups disclose their risks to investors
• D. A management technique focused on employee wellbeing

• A. Platforms charge users too little for their services
• B. A single firm can control a market, set terms for users and sellers, and suppress competition
• C. Platforms hire too many engineers, raising labor costs
• D. Platforms store data on servers located in multiple countries

• A. Copyright law
• B. Antitrust or competition law
• C. Data localization rules
• D. Consumer warranty law

• A. Free services always contain hidden subscription fees
• B. Users often pay with personal data and attention, which can be monetized in ways they did not fully agree to
• C. Free services are required by law to display advertisements
• D. Free services are slower than paid alternatives by design

• A. Releasing the feature to all users immediately so problems surface quickly
• B. Piloting the feature with a small group and monitoring for harmful outcomes before a full rollout ✓
• C. Skipping user testing because speed is the top priority
• D. Collecting as much personal data as possible in case it becomes useful later

Why: A staged rollout with monitoring lets teams catch harms early, before they affect large numbers of people. Immediate full release and skipping testing prioritize speed over the safety of users.

• A. A clearly labeled button to decline optional cookies
• B. A subscription cancellation process that requires multiple confusing steps and phone calls ✓
• C. A settings menu that lets users export their data
• D. A notification that explains why location access is needed

Why: Dark patterns are interface choices that trick or manipulate users into outcomes they did not intend. Making cancellation deliberately hard is a classic example, while the other options respect user choice and transparency.

• A. Unintended consequences rarely matter once a product has many users
• B. Products can cause harm to groups who were not considered during design, so anticipating these effects is part of responsible development ✓
• C. Only legal teams, not designers or engineers, need to consider consequences
• D. Unintended consequences are always positive and need no review

Why: Designers often build for an imagined average user, which can exclude or harm others. Anticipating misuse and unequal impacts before launch is a core responsibility of ethical product development.

• A. Users are presented with clear, understandable information about what data is collected and why, and can meaningfully agree or decline ✓
• B. Users are assumed to agree because they scrolled past a long terms-of-service document
• C. Consent is obtained once at account creation and never requested again
• D. Companies may share data with any partner as long as it is stored securely

Why: Informed consent requires that people understand what they are agreeing to and have a real choice. Buried, lengthy legal text or one-time blanket agreements do not meet this standard.

• A. Ignore it, since engagement metrics are improving
• B. Hide the finding from leadership to avoid slowing the product roadmap
• C. Investigate the harm, adjust the ranking goals, and communicate the change transparently ✓
• D. Blame users for choosing to engage with the content

Why: Ethical practice means acting on evidence of harm: investigating its scope, changing the optimization target if needed, and being transparent about the change. Ignoring, hiding, or blaming users avoids responsibility.