Artificial intelligence is now increasingly used to help make decisions that have a direct effect on people's lives. Computer systems already assist in deciding who receives a bank loan, which job applicants are invited for an interview, and even how patients are diagnosed in hospitals. Supporters of these systems argue that they can be faster, more consistent, and entirely free from the tiredness, moods, and personal prejudices that affect human judgment. As such tools spread ever more widely, however, important questions about how much we should really trust them are becoming harder and harder to ignore.
One of the main attractions of artificial intelligence is that it can find patterns hidden within huge amounts of data far more quickly than any human being possibly could. A system that has been trained on millions of past examples may notice subtle connections that people would simply miss. In medicine, for instance, such systems have already helped doctors to detect certain diseases in medical images at an early stage. Used carefully and wisely, AI can genuinely support human experts, handling routine cases efficiently and freeing people to focus their attention on the most difficult ones.
Yet these same systems also have serious and worrying weaknesses. Because they learn directly from data produced by human beings, they can quietly absorb and then faithfully repeat existing human biases. If past hiring decisions unfairly favored one group of people over another, a system trained on that data may end up doing exactly the same thing, all while appearing perfectly neutral and scientific. Many of these systems are also extremely difficult to understand, even for the experts who created them, so it can be very hard to know why a particular decision was actually made. When such a decision turns out to be wrong, the person affected may have no clear way to challenge it.
For all these reasons, many experts now argue that genuinely important decisions should never be left to machines alone. They call for keeping humans firmly "in the loop," so that a real person always reviews and takes final responsibility for the outcome. They also stress the urgent need to test these systems carefully for hidden bias and to be able to explain clearly how they work. The goal, they suggest, is not to reject artificial intelligence, which offers real benefits, but to use it wisely—as a tool to support human judgment, never as an excuse to abandon the responsibility that must come with it.
（1） 正解 2. Because it can be faster, more consistent, and free from moods.
第1段落に「より速く、より一貫し、疲れや気分・偏見に左右されない」とある。選択肢2。
（2） 正解 3. By helping detect certain diseases in images.
第2段落に「医療画像で特定の病気の早期発見を助けてきた」とある。選択肢3。
（3） 正解 1. They can absorb and repeat human biases.
第3段落に「人間の偏りを吸収し繰り返しうる」とある。選択肢1。
（4） 正解 2. Keeping humans "in the loop" and testing for bias.
第4段落に「人間を判断の輪に残し、偏りを検査する」よう求めるとある。選択肢2。
irrational：非理性的な
not based on clear or sensible thinking（明確で分別ある考えに基づかない）
panic：うろたえる・パニック
sudden uncontrollable fear（突然の抑えられない恐怖）
drawback：欠点
a disadvantage of something（ある物事の不利な点）
overcome：克服する
to deal successfully with a problem（問題にうまく対処する）
bargain：交渉する
to discuss terms in order to reach an agreement（合意に達するため条件を話し合う）
strike：ストライキ
a refusal to work as a protest（抗議として働くことを拒むこと）
bias：偏り・偏見
an unfair preference or dislike（不公平な好みや嫌悪）
consistent：一貫した
always behaving in the same way（常に同じように振る舞う）