영어 모의고사 (자동 편성)

Many large companies now rely on artificial intelligence ①to screen job applications before a human ever sees them. These systems promise efficiency, scanning thousands of resumes in seconds and ranking candidates according to patterns ②drawn from previous hiring decisions. However, this reliance on historical data creates a troubling risk. If a company's past hiring favored a particular group, the algorithm may learn to treat that pattern as a marker of success and quietly reproduce it, even when no one ③intending discrimination. Unlike a human interviewer, the system offers no explanation for its choices, which makes the bias difficult ④to detect until an outside audit reveals it. Some researchers argue that fairness cannot be added afterward; it must ⑤be built into the system's design from the very beginning.

Generative AI systems learn to write, draw, and compose by studying enormous collections of existing human work, much of ①which was created without the artists' knowledge or consent. Supporters of the technology argue that this process ②resembles how a human student learns by studying the masters, absorbing style and technique rather than copying specific pieces. Critics respond that the comparison ③breaks down at scale: a single AI model can absorb millions of works in days and then generate outputs that directly compete with the very artists whose work ④trained it. Courts around the world are now ⑤ask to decide whether training an algorithm on copyrighted material counts as fair use or as infringement.

Conventional economic wisdom holds that more options are always preferable to fewer, since a larger set of alternatives ①increases the likelihood that any individual will find one that precisely matches their preferences. When shoppers are presented with an extensive array of choices, they frequently experience heightened anxiety about making the wrong selection and, paradoxically, become ②less likely to make any purchase or decision at all. Furthermore, individuals who do eventually choose from an overwhelming array often report ③higher satisfaction with their selection than those who chose from a modest handful of options, since the sheer number of unexplored alternatives ④fosters lingering doubt about whether a better choice was overlooked. Businesses that recognize this dynamic have begun deliberately ⑤curating smaller, carefully selected assortments, betting that a narrower range of options can sometimes generate more sales and greater customer satisfaction than abundance.

Hospitals are increasingly adopting artificial intelligence tools that can scan medical images and flag potential tumors ①faster and, in some trials, more accurately than experienced radiologists. Patients understandably welcome any technology that catches disease earlier, yet the growing reliance on these tools raises an uncomfortable legal question that hospitals have only begun to address. When an AI system misses a diagnosis that a human doctor would likely have caught, responsibility becomes ②easy to assign. The doctor may argue that they simply trusted a tool the hospital required them to use, the hospital may argue that the software company guaranteed its accuracy, and the software company may argue that no algorithm can be expected to be ③perfect. Until legal systems ④clarify this chain of responsibility, patients harmed by an algorithmic error may find themselves with no clear party to hold ⑤accountable.

Metacognition, broadly defined as the capacity to monitor and evaluate one's own cognitive processes, is often treated as a reliable guide to what we actually know, yet a growing body of evidence suggests that the correspondence between subjective confidence and objective accuracy is far weaker than intuition would suggest. Individuals routinely overestimate how thoroughly they have understood a passage of text immediately after reading it, a phenomenon partly attributable to the fluency with which the material was processed rather than to any genuine grasp of its underlying structure. Because rereading familiar material tends to feel effortless, _______________, a misattribution that can lead them to abandon further study prematurely. This gap between perceived and actual mastery, sometimes labeled the illusion of knowing, has practical consequences for education, since students who trust their inflated self-assessments are less likely to seek out the more effortful, retrieval-based strategies that would otherwise consolidate their learning.

Labeling theory approaches deviance not as a fixed property of certain acts, but as a social process created when powerful groups attach a negative label to a person's behavior. According to this view, nearly everyone breaks minor rules at some point, yet only some individuals are singled out, caught, and formally branded as delinquents or criminals. Sociologist Howard Becker argued that this label often becomes a master status, a single identity that overshadows every other trait a person has. Once labeled, an individual may find legitimate opportunities closing off, pushing them toward _______________. This process, sometimes called secondary deviance, illustrates how social reaction can produce the outcome it claims merely to describe.

Even highly capable individuals can make disastrous collective decisions once they are absorbed into a tightly knit group, a phenomenon psychologist Irving Janis termed groupthink. Within cohesive committees, boards, and advisory councils, members often prioritize consensus and group harmony over a rigorous evaluation of alternatives, silencing doubts rather than risking the discomfort of dissent. Sociologists extend this analysis beyond small committees, noting that entire organizations can develop cultures that reward conformity and punish employees who raise inconvenient questions. Historical inquiries into policy failures and corporate disasters frequently reveal that critical information existed within the organization but never reached decision makers, not because it was hidden, but because _______________.

Facial recognition software has moved rapidly from airport security lines into shopping malls, schools, and even private apartment buildings. (A) Independent studies, however, have repeatedly found that many commercial systems perform less accurately on women and people with darker skin tones. (B) Proponents note that the technology can unlock doors, catch shoplifters, and speed up identity checks far faster than any human guard could manage. (C) Several cities have already reported cases in which an innocent person was detained based on a flawed match.

A cotton plant that manufactures its own insecticide sounds like science fiction, yet it has been growing in fields around the world for decades. (A) Supporters point to yield increases and reduced pesticide spraying documented across millions of hectares, while critics worry about resistance evolving over time, which is why regulatory agencies in most countries now require extensive testing before a genetically modified crop reaches the market. (B) Other genetically modified crops carry a gene that allows them to survive direct exposure to a particular herbicide, letting farmers spray a field to kill weeds without harming the crop growing among them. (C) Scientists achieved this by isolating a gene from a soil bacterium that produces a protein toxic to specific insect pests but harmless to humans, then inserting that gene into the plant's own genome so that every cell manufactures the protective protein.

Every one of those clones, remarkably, is a perfect genetic copy of the original. The immune system produces an almost limitless variety of antibodies, each shaped to latch onto one particular molecular target with remarkable specificity. ( ① ) Monoclonal antibody drugs harness this natural precision by cloning a single antibody-producing cell in the laboratory. ( ② ) The result is an army of identical antibodies built to recognize one target, such as a protein on a cancer cell or a molecule driving an autoimmune disease. ( ③ ) Because the antibody binds only to its intended target, these drugs tend to spare healthy tissue far more effectively than traditional chemotherapy, which attacks any rapidly dividing cell regardless of whether it is cancerous. ( ④ ) Manufacturing monoclonal antibodies remains far more complicated than producing a conventional small-molecule pill, since living cells must be cultivated in carefully controlled bioreactors for weeks before the antibody can be purified and prepared for use. ( ⑤ ) This complexity keeps monoclonal antibody treatments considerably more expensive than most conventional drugs, a cost that continues to shape debates over pricing and access.