Date
Publisher
Google
We conducted a preregistered randomized controlled trial in Sierra
Leone to evaluate the impact of the Guided Learning feature in
Gemini on student learning outcomes in mathematics. The trial
enrolled __ = 1763 junior secondary school students across 48
mathematics classrooms. We asked teachers to incorporate Gem-
ini into half of their weekly lessons for eight weeks, adding up
to a requested total of 12 hours. Guided Learning significantly
improvedstudentlearningoutcomes,withanintent-to-treateffect
of +0.258 SD (__ = 0.029). Uptake exceeded our expectations,
with 69.0% of students engagingwith GuidedLearning for atleast
12 hours over the trial. Reaching this threshold yielded a signifi-
cant treatment-on-the-treated effect of +0.380 SD (__ = 0.029).
In focus groups, teachers emphasized students’ enthusiasm for
GuidedLearningsessionsandcreditedGeminiwithhelpingreveal
new ways to teach familiar topics.
What is the application?
Who age?
Why use AI?

