Sample & method.
Anonymous Qualtrics survey distributed via program channels in April 2026. Counts on the cards below reflect the active filter; demographic counts include partial responses where the respondent gave campus or year but did not finish.
Why students engage and disengage.
Multi-select questions. Bars show count of respondents in the filtered group who selected each option; percentage is over respondents who answered the question.
Tools and tasks.
These three questions were shown only to respondents who use AI tools. Percentages denominate against AI users in the filtered subgroup.
The acceptance line.
For each task, respondents selected the maximum AI involvement they would accept from a fellow GSND student. Sorted by mean — most restrictive on top. 1 = No AI · 2 = Student creates with AI assist · 3 = AI generates with student direction.
Policy preferences.
Closed-response questions on how GSND should handle AI in coursework, capstones, and program services.
Career and expectations.
How students see AI in their own path, plus the gap between what they want for the industry and what they think will actually happen.
Group differences.
Three non-parametric tests on the per-respondent permissiveness score (mean of the 14 acceptance items, 1–3 scale). Mann-Whitney U is appropriate here: ordinal scale, n < 30 in subgroups, no normality assumed. Wilcoxon signed-rank is used for the within-subject ideal-vs-realistic comparison.
Limitations and next round.
A faithful reading of these results requires acknowledging what the instrument cannot tell us, and what it should ask next time.
- Small n. 35 total responses, 27 with full data; subgroup tests have limited power. The +0.34 difference between 1st- and 2nd-year permissiveness means is suggestive but not statistically distinguishable from chance at this sample size.
- Self-selection. A program-wide opt-in survey on AI almost certainly over-samples respondents with strong views in either direction. The polarity of the open responses is consistent with that.
- Boston-heavy. The Oakland subgroup is too small to analyze separately even with filtering.
- Scale ceiling on Q4b. "Worried about academic integrity" is endorsed by most respondents including most AI users — the item conflates personal concern with concern about peer behavior.
- Domestic / international student. A single closed item would let us check whether the AI-restrictive cluster is concentrated in any particular subgroup.
- Career goal: academic vs. industry. Q26 measures attitude toward AI; a separate item on intended path would let us test whether the attitudes track with destination.
- Specialization. Art / programming / design / production. Would clarify whether the dialogue/visual/voice-acting boundary is held by the whole program or driven by art-track students.
- Disclosure-grading interaction. Q18 and Q19 are scored independently; pair them to surface joint preferences.
Verbatim — what students want to learn.
Q17: "What would you like to learn about AI through the GSND program?" Open-ended; reproduced unedited. Filter applies.
Verbatim — open comments.
Q27: "Is there anything else you'd like to share about AI in game design education, or changes you'd like to see in how GSND handles it?" Reproduced unedited; line breaks preserved. Filter applies.