How algorithms shape our reality

• A. A type of computer hardware that stores large amounts of data
• B. A finite, step-by-step set of instructions for solving a problem or completing a task ✓
• C. A programming language used only for building websites
• D. A random process that produces unpredictable results

Why: An algorithm is a well-defined sequence of steps that takes an input and produces an output, designed to solve a specific problem. It is a concept, not hardware, a language, or randomness.

• A. Looking at the sky to guess the weather
• B. Following a recipe to bake a cake ✓
• C. Feeling inspired to write a poem
• D. Choosing a song by mood alone

Why: A recipe lists ordered, explicit steps that transform ingredients into a result, just like an algorithm transforms inputs into outputs. The other options involve intuition or unstructured judgment rather than defined steps.

• A. So they can be written in multiple natural languages at once
• B. Because computers can only carry out instructions that are clearly defined ✓
• C. So that they run more slowly and use less memory
• D. Because humans must understand every step before a computer can run it

Why: Computers follow instructions literally, without common sense or the ability to infer intent. Vague steps cannot be executed, so precision is essential.

• A. It ranks and selects content based on rules and patterns learned from data ✓
• B. It randomly shuffles posts so every order is equally likely
• C. It only displays posts in the order they were originally published
• D. It is the physical server where the posts are stored

Why: Feed algorithms use inputs such as your past interactions and post features to decide what to show and in what order. This ranking process is a central example of how algorithms shape what we see.

• A. Algorithms have no inputs; they always generate their own data
• B. Algorithms take inputs, process them through defined steps, and produce outputs ✓
• C. Algorithms only produce outputs and never receive any information
• D. Inputs and outputs are the same thing in every algorithm

Why: The core structure of an algorithm is input, processing through defined steps, and output. Understanding this flow helps explain how a seemingly simple system can influence the results people see.

• A. To display posts in strict chronological order
• B. To rank and select content predicted to keep a user engaged ✓
• C. To delete posts that contain spelling errors
• D. To send every post to every user equally

Why: Feed algorithms score and rank posts based on predicted relevance and engagement, so each user sees a personalized selection rather than everything in time order.

• A. The user's government ID number
• B. Past behavior such as likes, watch time, shares, and comments ✓
• C. The weather forecast in the user's city alone
• D. The user's password history

Why: Platforms learn from behavioral signals like what you click, how long you watch, and what you interact with to predict what you will find relevant.

• A. A security feature that blocks spam accounts
• B. A situation where personalization limits a user's exposure to diverse viewpoints ✓
• C. A type of advertisement that floats over videos
• D. A setting that hides all images from a feed

Why: A filter bubble occurs when personalized recommendations repeatedly show content similar to what a user already engages with, narrowing the range of perspectives they encounter.

• A. Algorithms are designed to ignore emotional content
• B. Emotional posts tend to generate more engagement, which the ranking system may reward ✓
• C. Emotional posts are always fact-checked first
• D. Platforms require users to post only neutral content

Why: Because ranking systems often optimize for engagement signals like comments and shares, content that triggers strong reactions can receive a boost in distribution.

• A. They decide what is shown or hidden based on rules, classifiers, and ranking criteria ✓
• B. They only remove content after a court orders it
• C. They store copies of every message on the user's device
• D. They make all users see identical feeds to ensure fairness

Why: Filters combine automated classifiers, policy rules, and ranking criteria to decide what appears, what is demoted, and what is hidden, which shapes what users perceive as their reality.

• A. Algorithms always produce more accurate results when given more data.
• B. If the training data reflects historical biases or unequal representation, the algorithm's outputs will tend to reproduce those biases.
• C. Bias is removed from a system once the output has been reviewed by a human.
• D. Only data collected by governments can introduce bias into an algorithm.

• A. The system will perform equally well on all skin tones because it learns general facial features.
• B. The system will likely have higher error rates for darker-skinned faces because they were underrepresented in the training data.
• C. The system will automatically correct for skin tone during processing.
• D. The system will become less accurate overall as it sees more light-skinned faces.

• A. A sensor malfunctions and records incorrect temperatures for a week.
• B. A hiring dataset is built from past hiring decisions in a company that historically favored male candidates for engineering roles.
• C. A survey is accidentally sent to the wrong email list.
• D. A programmer makes a typo in a line of code.

• A. Because neutral datasets always contain random errors that cancel out.
• B. Because proxy variables such as zip code can correlate with protected characteristics like race, leading the model to discriminate indirectly.
• C. Because algorithms cannot process numerical data accurately.
• D. Because neutral datasets are always too small to be useful.

• A. Assume the algorithm is objective because it uses mathematics.
• B. Remove the algorithm from consideration and rely only on intuition.
• C. Audit the training data for representation gaps and test the model's performance across different demographic groups.
• D. Increase the model's complexity so it can detect bias on its own.

• A. Selling subscriptions to users for premium access
• B. Selling user attention to advertisers ✓
• C. Charging developers a fee to build apps on the platform
• D. Selling user-generated content to publishers

Why: Most social media platforms are free to users, and they earn revenue by selling advertisers access to users' attention, typically measured by impressions and clicks. The more time and engagement a platform captures, the more ad inventory it can sell.

• A. How accurately content matches a user's stated interests
• B. How much users like or trust the platform overall
• C. Observable actions like clicks, likes, comments, shares, and watch time ✓
• D. How many hours per day a user spends offline

Why: Engagement refers to measurable behaviors such as clicks, likes, comments, shares, and time spent watching or reading. Algorithms optimize for these signals because they are easy to quantify and correlate with ad revenue, even if they do not perfectly reflect user satisfaction.

• A. Emotional content is cheaper to produce than factual content
• B. Platforms are legally required to promote emotional material
• C. Strong emotions like anger or outrage tend to provoke more clicks, comments, and shares ✓
• D. Emotional content always has higher advertising rates

Why: Content that triggers strong emotions, especially anger or outrage, often prompts people to react and share quickly. Because the algorithm rewards these reactions with more distribution, provocative content can be amplified even when it is not the most accurate or useful content.

• A. An economic system where people are paid wages to watch advertisements
• B. A market in which human attention is a scarce resource that is bought and sold ✓
• C. A government policy that taxes time spent on screens
• D. A system where teachers are rated by students' focus levels

Why: The attention economy treats human attention as a limited resource that businesses compete for. Because people have only so many waking hours, capturing and holding attention becomes valuable, and companies build products and sell advertising around it.

• A. Users consistently see only content they already agree with, with no exposure to new ideas
• B. Users may spend more time on platforms than they intended, even if they feel worse afterward ✓
• C. Advertisers stop paying for placements because engagement drops to zero
• D. Content creators are guaranteed equal reach regardless of quality

Why: When engagement is the main goal, platforms can keep users scrolling even when the experience is unpleasant or unhelpful, since time on site still generates revenue. Other effects, like filter bubbles, are also discussed, but option B directly describes the gap between engagement and well-being.

• A. To guarantee that every applicant receives a loan
• B. To estimate the likelihood that a borrower will repay a debt based on patterns in historical data
• C. To determine the exact future income of an applicant
• D. To replace all human judgment in financial decisions

• A. Because algorithms always generate random outcomes
• B. Because the historical data reflects past human decisions, which may have included discriminatory patterns
• C. Because algorithms cannot process numerical data
• D. Because job applicants always lie on their resumes

• A. A system where police officers review their own performance reports
• B. A cycle in which police are sent to areas predicted to have crime, generating more recorded arrests there, which then reinforces the prediction
• C. A method of encrypting police communications
• D. A process for training officers using simulated crime scenarios

• A. Predictions have no effect because they are only estimates
• B. Predictions can shape who gets opportunities, scrutiny, or resources, thereby influencing the outcomes they claim only to forecast
• C. Predictions only matter when they are perfectly accurate
• D. Predictions are powerful only when made by government agencies

• A. To make the algorithm run faster on computers
• B. To check whether the system produces unfair error rates or outcomes across different groups of people
• C. To remove the need for any data at all
• D. To ensure the algorithm never changes over time

• A. A physical room designed to amplify sound for public speeches
• B. An environment where a person mainly encounters information and opinions that reinforce their existing beliefs ✓
• C. A feature that automatically repeats posts to increase their visibility
• D. A government-run platform for public debate

Why: An echo chamber is a social or informational environment where people are mostly exposed to views that match their own, so their beliefs are repeated and reinforced rather than challenged.

• A. They show users a random selection of posts from across the platform
• B. They prioritize content predicted to keep users engaged, such as content similar to what they have already liked or clicked ✓
• C. They only display posts from accounts the user has never interacted with
• D. They present every post in strict chronological order without any ranking

Why: Many recommendation systems predict which content will hold attention and rank it accordingly, often favoring material similar to past engagement, which can narrow the range of viewpoints a user sees.

• A. Differences in the emotional tone of news broadcasts across countries
• B. Growing emotional hostility and distrust between people who hold opposing political identities ✓
• C. A scientific method for measuring the intensity of online arguments
• D. The tendency of people to change their political views every election cycle

Why: Affective polarization refers to increasing dislike and distrust between groups defined by political identity, which is distinct from simply disagreeing on policy issues.

• A. Because emotionally charged or divisive content often generates more clicks, comments, and shares ✓
• B. Because algorithms are legally required to show only extreme political content
• C. Because users always prefer neutral and balanced information
• D. Because engagement metrics are never recorded by platforms

Why: Content that provokes strong emotions such as anger or outrage tends to attract more interaction, so systems optimized for engagement may amplify divisive material over measured, balanced content.

• A. Following only the accounts that share your existing views to stay informed
• B. Deliberately seeking out reliable sources with differing perspectives and checking claims across them ✓
• C. Turning off all news sources to avoid being influenced
• D. Relying solely on the first search result for every topic

Why: Actively seeking diverse, credible sources and cross-checking claims helps counter the narrowing effect of personalized feeds, while avoiding news altogether or using one-sided sources does not address the problem.

• A. Keeping a platform's code hidden so competitors cannot copy it
• B. Making it clear to users how an algorithm uses their data and what factors influence its outputs ✓
• C. Allowing algorithms to make decisions without any human oversight
• D. Showing users every single calculation the algorithm performs in real time

Why: Algorithmic transparency means users and regulators can understand, at a meaningful level, how a system uses data and what drives its decisions. It does not require exposing every line of code or every calculation, and it is distinct from secrecy or removing human oversight.

• A. Feeds are randomly generated with no link to user behavior
• B. Platforms are legally prohibited from showing any personalized content
• C. Recommendation systems are optimized for engagement and often offer few visible controls ✓
• D. Users are always able to adjust every ranking factor directly

Why: Most recommendation systems are tuned to maximize metrics like watch time or clicks, and many platforms provide only limited or hard-to-find settings. This is why reclaiming agency often involves seeking out available controls or pushing for better ones.

• A. The General Data Protection Regulation (GDPR) in the European Union ✓
• B. The Sherman Antitrust Act in the United States
• C. The Geneva Conventions
• D. The Basel Accords on banking capital

Why: The GDPR, in force since 2018, includes rights related to automated decision-making, such as the right to meaningful information about the logic involved and certain protections against decisions based solely on automated processing. The other options address different domains.

• A. Turning off all internet access permanently
• B. Reviewing privacy and personalization settings, and using options like 'not interested' or chronological feeds where available ✓
• C. Sharing your password so the platform can better understand you
• D. Assuming the default settings are the best available choice

Why: Many platforms offer personalization controls, feed preferences, and feedback options that shape recommendations. Actively reviewing these settings and giving explicit feedback helps users steer what algorithms show them.

• A. It guarantees that algorithms will never make mistakes
• B. It allows outside experts to test whether systems produce harmful or biased outcomes without relying solely on the company's own claims ✓
• C. It replaces the need for any laws or regulations
• D. It makes algorithms faster and cheaper to run

Why: Independent audits provide an external check on a company's self-reporting. Researchers and auditors can test system behavior for bias, harm, or inconsistency, which supports accountability and informs better regulation.