What Does the Research Say?

Positive Findings

• AI tutoring effect sizes of 0.3–0.6 confirmed across 50+ studies, strongest in STEM and writing-intensive courses
• Students using AI-assisted tools completed problem sets 40% faster with no loss in retention (MIT, 2024)
• Adaptive platforms show 20–30% course-completion improvements in community college contexts
• AI-augmented courses outperform controls on transfer tasks — suggesting deeper learning (Stanford HAI, 2025)
• ESL students using AI writing support improved grades 18–27%; 24/7 availability disproportionately benefits working adults and rural learners

Concerns & Limitations

• "Cognitive offloading": heavy AI users show reduced ability to initiate problem-solving independently (Harvard, 2024)
• Heavy generative-AI reliance associated with reduced original ideation over a semester (Nature Human Behaviour, 2025)
• The "AI divide": higher-SES students are 3× more likely to use AI strategically vs. reactively (Educause, 2025)
• AI detection tools show false-positive rates up to 61% higher for ESL writers (Stanford, 2024)
• Traditional text-based assessments now considered unreliable without process documentation (AERA consensus, 2025)
• 71% of faculty feel underprepared to teach AI literacy; 58% report inadequate institutional PD

What Research Tells Faculty to Do

• Design AI-transparent assessments (not "AI-resistant"): oral exams, process portfolios with AI documentation, locally-situated problems, AI + human hybrid tasks, staged in-class checks
• Teach WITH AI: prompt engineering as literacy, SIFT framework adapted for AI outputs, AI red-teaming exercises, model your own AI use transparently
• Redefine learning objectives: Bloom's top tiers as the baseline, AI literacy as a program-level outcome, distinguish AI-augmentable vs. human-essential skills

What Research Tells Administrators to Do

• Establish clear, contextual AI policies — course-level and tied to learning objectives; blanket bans are ineffective and erode trust
• Adopt AI-use disclosure protocols and consider co-developed "AI use contracts" with students
• Invest in faculty development: hands-on, discipline-specific, ongoing — with time and course release for redesign