Higher Education & ResearchIssue 376 records
Student Success
For 2026-10-13, student-success evidence centres on how campuses are supplying AI and AI literacy, with little outcome measurement. UEA began an optional StudyStash pilot that turns academics' own course materials into revision aids, and Washington State is opening free Gemini to all students with data-loss alerts but, by IT's account, nothing to stop AI completing whole assignments. A U-M blog relayed a Harvard educator's claim of 62% higher learning gains with a Collage AI tutor versus a prior semester; she is the vendor's chief learning officer. USC launched three optional AI Essentials courses, UTPB won a $3 million federal grant that includes AI student-success tools, and a peer-reviewed survey of 131 Australian engineering students found use varied by task, with verification the top challenge. Operating takeaway (interpretation): pair sanctioned tools with controlled evaluation and discipline-embedded verification skills. Limits: announcements, one student-newspaper report, one uncontrolled vendor-linked claim and one survey.
- Evidence records
- 6
- Cross-source patterns
- 2
- Evidence classes
- 3 vendor claim2 independent reporting1 academic research
- Outcomes
- 4 emerging1 cautionary1 mixed
- Source freshness
- 6 new this fortnight
- Research completed
- 2026-10-14
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
Campuses are choosing sanctioned or course-grounded AI tools, but integrity safeguards and outcome evidence lag behind
On 12-13 October, UEA began an optional StudyStash pilot that generates revision aids only from academics' materials, WSU opened free Gemini access to all students as an institutionally supported alternative to consumer tools, and a U-M blog relayed a vendor-affiliated claim of 62% higher learning gains from an instructor-configured Collage AI tutor. UEA published no pilot metrics, WSU said nothing prevents AI from completing entire assignments, and the Collage figure is an uncontrolled cross-semester comparison. Interpretation: grounding and institutional control are becoming the default design choices, but they do not substitute for controlled evaluation and course-level integrity rules. Limits: one announcement, one student-newspaper report and one second-hand claim.
Operating questionBefore scaling a sanctioned AI tool, has the institution defined how learning effects and integrity risks will be measured in the first term?
Supporting evidenceUniversity of East Anglia (Pro-Vice-Chancellor for Student Education and Experience); StudyStash (vendor)Washington State University (Information Technology Services)University of Michigan Center for Academic Innovation (host); Kelly Miller, Harvard SEAS; Collage AI (vendor)
AI literacy is being supplied as optional courses and grant-funded capacity, while student evidence points to task-specific needs
USC launched three optional 2-unit credit/no-credit AI Essentials courses open to all majors, and UTPB won a $3 million federal grant for AI credentials, training hubs and AI tools in student-success services. A peer-reviewed survey of 131 Australian engineering students published 12 October found use varied sharply by task and cohort (code generation 46.60% vs 7.14%) and that verifying accuracy was the top challenge, leading the authors to recommend discipline-specific literacy embedded in coursework. Interpretation: elective and capacity-level offerings expand access but may miss the verification skills students need in their own courses. Limits: two announcements without outcome measures and one self-report survey.
Operating questionDoes the institution's AI literacy plan include discipline-embedded verification tasks, or only elective courses and general training?
Supporting evidenceUniversity of Southern California (Office of the Provost; Dornsife, Marshall, Viterbi and Stevens schools)University of Texas Permian Basin (Provost's office; Grants and Sponsored Programs); U.S. Department of Education (funder)University of Technology Sydney and Southern Cross University (authors Lidfors Lindqvist and Quince)
Evidence ledger
Full record · every source keeps its link and limitations
University of East Anglia pilots StudyStash, an optional Blackboard-integrated AI revision tool grounded in course materials
UEA has begun an optional pilot of StudyStash, an AI platform available through Blackboard in selected courses across three schools, which turns lecture notes and other module materials into flashcards, practice quizzes, study plans and audio content and tracks student progress. UEA says the tool draws only on content created by its academics, complements lectures, seminars and academic guidance, and leaves students responsible for their own learning and assessment. UEA says the platform is already used at Birmingham, Loughborough and City St George's, and that findings from the pilot will shape the vendor's product development.
Why it matters, evidence and limitations
- Why it matters
- Published 13 October, this is a fresh example of a university choosing a course-grounded, LMS-embedded AI study tool rather than general chatbot access, framed as a pilot. It is relevant to student-success teams weighing revision support for gateway and international students.
- Evidence and measured results
- Institutional news release with pro-vice-chancellor and vendor CEO quotes and an FAQ. No pilot size, comparison group, evaluation metrics or timeline are stated; benefits are anticipated, not measured.
- Limitations and uncertainty
- Self-reported announcement; vendor will use pilot findings for product development; no data on use, learning, confidence or integrity effects yet.
Washington State University rolls out free Gemini to all students with data-loss alerts but no safeguards against AI completing assignments
WSU will give all students free access to Google Gemini starting 19 October, after faculty and staff from 12 October, under a no-cost tier limited to 25 pro-model prompts every four hours, 300 fast-model prompts a day and 50 Gemini Notebook questions a day. IT leaders said Google will not use WSU data for training, that data-loss controls alert security staff when FERPA, HIPAA or other regulated data enters the model but are 'not perfect', and that nothing currently prevents students from having the AI complete entire assignments. ITS will encourage faculty to use AI-percentage checks.
Why it matters, evidence and limitations
- Why it matters
- Reported 12 October with student access opening 19 October, this shows a campus-wide AI access decision being made mainly as an IT and data-protection rollout, with academic integrity and learning measures left to faculty.
- Evidence and measured results
- Student-newspaper report quoting the CIO, an ITS director and a communications director. Usage caps, start dates and control descriptions are as stated by ITS; no usage, outcome or incident data.
- Limitations and uncertainty
- Single student-newspaper report; Google agreement terms unpublished; the paper's editorial board opposes WSU's AI use; no data on adoption or effects.
USC launches three optional 2-unit credit/no-credit AI Essentials courses for all undergraduates, starting Spring 2027
USC announced three new AI Essentials courses for Spring 2027, each a 2-unit Credit/No-Credit course open to any major with no technical background: a Marshall course on professional success and ethics in AI-powered workplaces, a Dornsife course on AI's risks, rewards and human future, and a Viterbi and Stevens course explaining what AI is and what it can and cannot do. The provost's office framed the courses as improving AI literacy for all undergraduates, taken in addition to degree-applicable units, and directed students to their academic advisors.
Why it matters, evidence and limitations
- Why it matters
- Announced 13 October, this shows a large research university meeting AI literacy through optional, low-stakes cross-school courses rather than a requirement or embedded discipline work. It offers a contrast point for institutions debating literacy requirements.
- Evidence and measured results
- Provost letter listing course codes, units, grading basis and owning schools; a Dornsife story (12 Oct) adds dean quotes. No enrolment, outcomes or evaluation information.
- Limitations and uncertainty
- Self-reported announcement; voluntary uptake may reach mainly already-interested students; no learning measures stated.
UT Permian Basin wins $3M federal Strengthening Institutions grant to build AI learning environments, credentials and AI-enabled student-success services
UTPB has received a five-year, $3 million Strengthening Institutions Program grant from the U.S. Department of Education for Project AIPB, running October 2026 to September 2031. The university plans AI-enabled learning environments, equipped AI training and research hubs, faculty and staff development, new credit and non-credit AI credentials, and integration of AI tools into student-success services to improve persistence, retention and graduation rates. Leaders also cast the grant as strengthening UTPB's regional workforce role. No targets, baseline, tools or evaluation design were reported.
Why it matters, evidence and limitations
- Why it matters
- Reported 13 October, a day after a comparable $3 million Strengthening Institutions award to Bradley was covered, this suggests federal capacity-building grants are now a recurring funding route for AI in student-success services at regional universities.
- Evidence and measured results
- Local newspaper item with provost and president quotes. Grant amount, program, term and planned components stated; no outcome measures, targets or tool details.
- Limitations and uncertainty
- Announcement-stage; local outlet relays university statements; claimed effects on persistence and graduation are goals, not results.
Harvard physics educator reports 62% higher learning gains with Collage AI tutor versus a prior semester, in U-M keynote summary; presenter is the vendor's chief learning officer
In a U-M Center for Academic Innovation summary of her keynote, Harvard's Kelly Miller reported that in a flipped applied physics course where Collage AI generated pre-class formative assessments and served as a tutor, students had 62% higher learning gains than a previous semester and appreciated the AI feedback. She also recapped the earlier Harvard physics experiment in which learning gains were nearly double with a custom AI tutor. Miller is chief learning officer of Collage AI and is extending it to other courses, universities and K-12 classrooms.
Why it matters, evidence and limitations
- Why it matters
- Published 13 October, this is the newest public claim of large AI-tutor learning gains in higher education. It sits against recent controlled evidence showing harms or null effects (such as the archived Maryland trial), making the evidence standard for tutor claims a live procurement question.
- Evidence and measured results
- Blog summary of a talk. The 62% figure compares one course with a previous semester; sample size, assessment instrument, cohort comparability and statistical tests are not given. The earlier randomized crossover study is described only briefly.
- Limitations and uncertainty
- Vendor-affiliated, non-randomized cross-semester claim reported second-hand by a host blog; no methods or data published in the source; one course in one elite institution.
Peer-reviewed survey of 131 Australian engineering students finds GenAI use varies sharply by task and cohort, with accuracy verification the top challenge
A peer-reviewed survey of 131 engineering students at two Australian universities found GenAI use was broad but uneven: 35.71% of a 28-student first-year regional cohort reported no academic use versus 9.71% of a 103-student second-year metropolitan cohort, and code generation was reported by 7.14% versus 46.60%. Verifying accuracy was the most common challenge for users in both cohorts (66.67% and 77.42%), and over 95% agreed universities have a role in preparing them for AI. The authors recommend gradual, discipline-specific GenAI literacy integrated into coursework.
Why it matters, evidence and limitations
- Why it matters
- Published 12 October, this adds fresh peer-reviewed evidence that student AI use is task- and stage-dependent, which bears directly on whether generic AI literacy courses or embedded discipline-specific instruction will meet student needs.
- Evidence and measured results
- Cross-sectional survey, 11 questions, chi-square and Mann-Whitney tests with Bonferroni adjustment and 95% CIs; cohorts are about 30% of each class. Self-reported use and perceptions only.
- Limitations and uncertainty
- Exploratory; confounded cohorts; self-report; engineering at two Australian universities only; the study does not measure learning, competence or misconduct.
How to read this edition
Source findings, measured results and limitations come from the cited publications. Patterns, operating questions, role takeaways and implementation considerations are Lighthouse Advisory interpretation, stated as questions to validate locally rather than guaranteed outcomes. Vendor and operator claims are labeled as claims. Full research method.
- Vendor claim
- A supplier-provided assertion that has not been upgraded to independent evidence.
- Independent reporting
- Independent reporting with attributable sources but without a formal evaluation design.
- Academic research
- Research produced through an academic institution or peer-reviewed venue.