2026年11月2日（月）マイページ
← 記事にもどる 中学生版高校生版 Word

「印刷する」→ 送信先で「PDFに保存」を選ぶと PDF になります。解答は最後のページです。

原田英語の時事英語ジュニア｜英語ニュース ワークシート（高校生レベル）

Class ____ No. ____ Name ____________________

AI家庭教師で成績が下がった？約2400人の実験、質問の約7割は答えや説明

College Students Offered an AI Tutor Ended Up With Lower Grades

音声・ゲームはこちらから

1University students who were given access to an AI tutor earned lower final grades than others taking the same courses, a large experiment in the United States has found. A team led by Jing Liu ran the trial at the University of Maryland.

2The trial took place during the fall 2025 semester and involved 2,379 undergraduates and 30 instructors. The instructors were randomly divided into two groups. Only students in one group could use the AI tutor, which ran on GPT-4o inside each course website. It answered questions using course materials that the instructor had picked.

3When classes of the same course were compared, final grades in classes with the tutor were lower by 4.12 points on a 100-point scale. Participation in online class activities also fell sharply. Homework scores did not improve. When all the courses were looked at together, the difference in grades was not clear. Students who were the first in their families to go to college lost more than twice as many points as other students.

4Because the groups were chosen by chance, the study measures the effect of offering the tutor. However, only about 15 percent of the students in the AI classes actually used the tutor. The grades came from the courses themselves rather than from a separate test. The paper is an early version that may still be revised.

5The researchers think the AI may have replaced reading course materials and asking instructors questions. Instructors could set the tutor to give direct answers or to lead students to the answer. Most of them kept the default setting, which gave direct answers. Liu said the results do not show that AI tutors designed for learning are unhelpful.

6The team also sorted 3,460 requests that students sent to the tutor. About 74 percent of them asked for answers, explanations or other information. Around 11 percent were for test preparation or practice. Just 0.7 percent of them asked for feedback on the students' own work.

Words　単語

accessB1名利用できること、使える権利the chance or right to use something
undergraduatesB2名大学生（学部生）university students who have not finished their first degree
instructorsB1名教員、指導者people who teach a class or a skill
randomly外副無作為にby chance, without any plan
participationB2名参加taking part in an activity
sharplyB2副大きく、急にby a large amount and quickly
revisedB1動修正される、改訂されるchanged to make better or more correct
replacedA2動～にとってかわったtook the place of something else
default外形初期設定の、最初から決まっているused unless someone changes it
preparationB1名準備、対策the work you do to get ready

単語のレベル：CEFR-J（『CEFR-J Wordlist Version 1.6』 東京外国語大学投野由紀夫研究室. （URL: https://www.cefr-j.org/download.html より 2026年8月ダウンロード））

1. Reading Quiz　本文の内容に合うものを選ぼう

  1. How were the instructors divided into two groups?

    A. By the size of their classesB. RandomlyC. By how much they liked AID. By the students' grades

  2. What happened to homework scores?

    A. They rose by 4.12 points.B. They doubled.C. They did not improve.D. They were not counted.

  3. Which students lost more than twice as many points as other students?

    A. Students who were the first in their families to go to collegeB. Students who used the tutor every dayC. Students in math classes onlyD. Students who never asked questions

  4. Why should readers be careful about the results?

    A. Because the AI tutor was never usedB. Because the study had only 30 studentsC. Because the grades came from a separate national testD. Because only about 15 percent of the students in the AI classes used the tutor

  5. What setting did most instructors keep?

    A. A setting that only gave hintsB. The default setting, which gave direct answersC. A setting that turned the tutor offD. A setting that wrote homework for students

2. True or False　正しければ T、違っていれば F

  1. (     ) The AI tutor answered questions using course materials chosen by the instructor.

  2. (     ) The grades came from a separate test made by the researchers.

  3. (     ) Liu said the results show that AI tutors designed for learning are unhelpful.

  4. (     ) Around 11 percent of the requests were for test preparation or practice.

3. Fill in the Blanks　下の語から選んで空らんをうめよう（形を変えずに）

accessinstructorsrevisedundergraduatesParticipationreplaced

  1. University students who were given _____ to an AI tutor earned lower final grades than others taking the same courses, a large experiment in the United States has found.

  2. The trial took place during the fall 2025 semester and involved 2,379 _____ and 30 instructors.

  3. _____ in online class activities also fell sharply.

  4. The paper is an early version that may still be _____.

  5. The _____ were randomly divided into two groups.

  6. The researchers think the AI may have _____ reading course materials and asking instructors questions.

4. Word Order　語を並べかえて正しい文にしよう

  1. ( randomly / The / into / instructors / two / were / divided / groups ) .

  2. ( not / did / Homework / improve / scores ) .

  3. ( Around / test / were / percent / 11 / practice / or / for / preparation ) .

5. Think and Write　英語で答えてみよう

  1. Should schools let students use AI tutors? Give your opinion and reasons.（学校は生徒にAIの家庭教師を使わせるべきだと思いますか？意見と理由を言いましょう。）

  2. How could an AI tutor be set up so that students learn more, not less?（生徒の学びが減らずに増えるようにするには、AIの家庭教師をどう設定すればよいと思いますか？）

Answers　解答

1. (1) B　(2) C　(3) A　(4) D　(5) B

2. (1) T　(2) F　(3) F　(4) T

3. (1) access　(2) undergraduates　(3) Participation　(4) revised　(5) instructors　(6) replaced

4.

  1. The instructors were randomly divided into two groups.
  2. Homework scores did not improve.
  3. Around 11 percent were for test preparation or practice.

全文訳

1AIの家庭教師を使えるようにしてもらった大学生は、同じ授業をとるほかの学生より最終成績が低かったことが、アメリカの大規模な実験でわかりました。ジン・リウさんが率いるチームが、メリーランド大学でこの実験を行いました。

2実験は2025年の秋学期に行われ、2,379人の大学生と30人の教員が参加しました。教員は無作為に2つのグループに分けられました。片方のグループの学生だけが、AIの家庭教師を使えました。それはGPT-4oで動き、授業ごとのウェブサイトの中に組みこまれていました。AIは、教員が選んだ授業の資料を使って質問に答えました。

3同じ授業のクラスどうしでくらべると、AIを使えるクラスの最終成績は、100点満点で4.12点低くなっていました。オンラインでの授業の活動への参加も大きく減りました。宿題の点数は上がりませんでした。すべての授業をまとめて見ると、成績の差ははっきりしませんでした。家族の中で初めて大学に進んだ学生は、ほかの学生とくらべて2倍以上も点数が下がりました。

4グループは偶然に任せて決めたので、この研究は「AIを使えるようにしたこと」の影響を調べたものになります。ただし、AIを使えるクラスの学生のうち、実際に使ったのは約15％だけでした。成績は、別のテストではなく、それぞれの授業の成績でした。この論文は、あとで直されるかもしれない途中の版です。

5研究者たちは、AIが、授業の資料を読むことや教員に質問することにとってかわったのかもしれないと考えています。教員は、AIをすぐに答えを教える設定にするか、答えへ導く設定にするか選べました。ほとんどの教員は、最初から決まっていた「答えを教える設定」のままにしていました。リウさんは、この結果は「学ぶために作られたAIの家庭教師が役に立たない」ことを示すものではないと話しています。

6チームは、学生がAIに送った3,460件の質問も分類しました。そのうち約74％は、答えや説明など、情報を求めるものでした。約11％は、テスト対策や練習のためのものでした。学生が自分でやったものについて意見を求めた質問は、わずか0.7％でした。

文法ポイント：過去完了（had + 過去分詞）

過去のある時点より前に終わっていたことを表す。本文の course materials that the instructor had picked は「（AIが答えた時点より前に）教員が選んでおいた授業の資料」という意味。

出典：IZA Institute of Labor Economics、EdWorkingPapers、EdTech Innovation Hub ほか（記事は原田英語の時事英語ジュニアのオリジナル）｜https://www.haradaeigo.com/jijieigo-junior/news/2026-10-26-ai-tutor-lower-grades/