Lower-performing students benefited the most, but even higher-performing students expressed support for the policy.
Restricting student access to smartphones is associated with significantly higher academic performance
randomized ITT estimate is positive, though the confidence interval begins near zero and uses an unexplained one-sided test
Among the prespecified subgroups examined, we detect larger and statistically significant effects relative to the control only among students with lower prior academic performance and first-year undergraduates
subgroup significance patterns lack reported interaction tests and multiplicity handling
Students who experienced the in-class phone bans became significantly more supportive of such restrictions
positive survey estimate from an unblinded subset with unclear arm-specific response rates
Derived from the full evaluation — not a separate score.
Strengths
The study randomized 420 department-grade units involving 16,683 students across 10 institutions and implemented the policy throughout a semester. This supplies causal evidence at a scale uncommon in the cited phone-ban literature.
↳ Methods, p. 179
The principal outcome is end-of-semester GPA, with 60% to 80% of course grades derived from externally set examinations evaluated off-site by graders blinded to treatment status.
↳ Methods, p. 179; Shedding light on a central question, p. 181
The paper identifies the higher-education sample, lecture-only policy, potential internal-grading pathway, and geographic setting as limits on interpretation and transfer.
↳ Shedding light on a central question, p. 181
Limitations
The estimated GPA increase is 0.073 SD with a 95% confidence interval from 0.003 to 0.144 and an unexplained one-sided P value. The direction is positive, but statistical support is fragile at the lower bound.
↳ Academic performance, p. 179
The claim that lower-performing and first-year students benefited more is based on significant-versus-nonsignificant subgroup patterns, without visible interaction tests or multiplicity handling. The standfirst elevates this pattern into a headline conclusion.
↳ Standfirst and Academic performance, pp. 178–179
The paper attributes gains to attention and classroom dynamics, yet perceived distraction did not change and treated students appeared more distracted during spot checks. The attention-allocation explanation is post hoc and the observations are explicitly exploratory.
↳ Students’ perspective and classroom dynamics, pp. 180–181
The cohort-randomized design and externally dominated examination outcome support the principal finding that the lecture-time ban modestly increased GPA. The 0.073-SD estimate is nevertheless borderline, with a confidence interval beginning at 0.003 and an unexplained one-sided test. Subgroup and mechanism claims receive less weight because interaction tests are not reported, the survey covers a subset, and exploratory spot checks found treated students appeared more distracted. The discussion handles major setting boundaries reasonably well, but the evidence remains specific to higher education in Odisha.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Sound3.7
Confidence mediumThe trial meaningfully extends a literature dominated by mixed observational and quasi-experimental evidence through random assignment at substantial scale. The advance remains bounded by a modest estimate, a higher-education setting, and unresolved subgroup and mechanism claims.
Our findings suggest that phone bans can improve students’ academic achievement
The study uses cohort-level randomization, intent-to-treat analysis, clustered standard errors, baseline GPA adjustment, and predominantly external examination grading. The one-sided primary test, unadjusted subgroup analyses, unclear survey response process, and inferable spot-check assignment reduce confidence in secondary inferences.
We conducted an intent-to-treat analysis
The paper clearly separates GPA, survey, and exploratory observational results and reports the primary estimate transparently. The title and standfirst state subgroup and attitudinal conclusions more strongly and broadly than the reported analyses establish.
Lower-performing students benefited the most
The discussion benchmarks the GPA effect and identifies higher-education, lecture-only, grading, and geographic boundary conditions. Comparison with the closest intervention literature is limited, while extrapolation to work, driving, social gatherings, and sleep exceeds the trial setting.
There are several important boundary conditions of this study.
Caveats4 of 4 checks
The reported primary estimate, confidence interval, and P value are numerically coherent, but several inferential claims warrant caution. These issues qualify interpretation rather than indicating data inconsistency.
IRB approval, preregistration, funding, AI assistance, and data availability are declared, with no conduct inconsistency identified in the supplied text. The absent registration identifier and uninspected supplementary material are assessability gaps rather than conduct concerns.
Flags: 4 declared / 5 total
15 references in manuscript 10 of 10 checkable references found in an index 5 have no canonical index record — counted, but not index-checkable 1 reference confirmed by manual review
References not verified automatically are resolved manually within 24 hours.
No retraction or expression of concern found in Crossref (includes Retraction Watch).
Sources: Crossref (incl. Retraction Watch) ✓
Where this paper’s evidence sits on the path from initial observation to real-world use.
The intervention was implemented for a semester in ordinary courses across 10 institutions using deposit boxes and teacher enforcement. Readiness to other settings still requires adaptation beyond this supported higher-education implementation.
This protocol was applied in all classes for the treatment arm throughout the semester
AI-generated, human-governed. Something look off? Contact us to request a review.
Keep reading
Other papers Nabu has evaluated blind, starting with the same field.
Kien Nguyen‐Trung, Susanne Friese · 2026 · Other Social Sciences
We propose a shift from ‘AI as coder’ to ‘AI as interlocutor’.
SupportedDr. P. Ponmurugan, Dr. J. Philip Robinson and Dr. S. Poornima · Plant production and plant biotechnology
Astrid Van Kampen, Matthew James Phillips, Scott Devenport · 2024 · Psychology
Due to the pervasiveness of this unsafety, young women had learned to understand the safety of the ‘self’ as their responsibility and modified and limited their self-representation accordingly.
Supported