Loading prices…
🔥BULLISH

AI Tools Boost STEM Exam Scores in 19-University Trial

The randomized study supports course-specific AI as a teaching aid, though it did not compare the tools with generic chatbots.

AI Tools Boost STEM Exam Scores in 19-University Trial
AI Tools Boost STEM Exam Scores in 19-University Trial

Students given access to course-specific AI tools scored higher on proctored exams than those in classes awaiting their rollout, a randomized trial across 19 universities found. The study enrolled 7,200 students, with 4,768 included in the experimental exam analysis. Researchers estimated a gain of 0.32 standard deviations, with a 95% confidence interval of 0.06 to 0.59.

Why it matters

The trial covered 55 applied STEM classes over three semesters. Researchers grouped classes by subject and academic level, then randomly assigned instructors to introduce their tools immediately or one teaching term later. Both groups received design and technical support.

Instructors built tutors for targeted practice and feedback, coaches that guided students through problems with questions, simulations and role-play exercises. All students in participating classes had access to the tools during the term before sitting the exams used to measure learning.

The exam analysis covered 44 course offerings and included only proctored, in-person tests, mostly on paper. Researchers added that screening rule after preregistration and excluded offerings where exams could allow AI assistance. The study found no statistically significant change in engagement or satisfaction.

Market impact

The results offer evidence that tailored AI tools can support learning in applied STEM courses, an area of interest for education technology providers and universities considering adoption. The estimated gain remains uncertain, as shown by the confidence interval.

The trial compared rollout timing, not course-specific tools against generic chatbots. It also leaves open which designs contributed most to the gain and whether the results extend beyond applied STEM.

Frequently asked questions

  1. How many students were included in the exam analysis?

    The experimental exam analysis covered 4,768 students across 44 course offerings. The broader trial enrolled 7,200 students.

  2. How large was the estimated improvement in exam scores?

    Researchers estimated a gain of 0.32 standard deviations, with a 95% confidence interval from 0.06 to 0.59 standard deviations.

  3. What kinds of AI tools did instructors build?

    Instructors built tutors for practice and feedback, question-led coaches, interactive simulations and role-play exercises.

  4. Were students allowed to use AI during the exams?

    The analysis included proctored, in-person exams and excluded course offerings where exams could allow AI assistance. Most included exams were on paper.

  5. Did the trial compare tailored AI tools with generic chatbots?

    No. The trial compared classes assigned to introduce their tools immediately with classes assigned to introduce them one teaching term later.

Source attribution
Aggregated from CryptoSlate · Verified · Last refreshed 23m ago
Open original →