From the Student Success edition of September 22, 2026
ASEE NSF IUSE WIP: human-guided GenAI outreach linked to higher STEM Performance Self-Evaluation
University of Nebraska–Lincoln / Lehigh University (NSF IUSE) · Higher education / academic research · United States (undergraduate engineering; transferable with limits)
- Publisher
- ASEE PEER
- Original publication
- 2026
- Source retrieved
- 2026-09-23
What happened
Human-guided GenAI pipeline (LLM engagement forecast + counseling-expert messages + GPT-4o personalization) associated with higher positive Performance Self-Evaluation across the semester (75.4% vs 56.5% historical baseline; d=0.90). Why now: ASEE 2026 NSF grantee paper not in student-success archive; addresses proactive non-cognitive support gap after tutoring RCTs.
Why it matters
Offers U.S. STEM student-success teams a responsible pattern—experts author, AI personalizes—for scaling weekly outreach before midterm early-alerts fire.
Evidence and measured results
Quasi-experimental comparison of Spring/Fall 2025 intervention cohort (N=41) to three prior semesters (N=96); Welch t-test on PSE; WIP explicitly non-causal.
Limitations and uncertainty
Historical controls; PSE self-report only; no grade/retention endpoints in this WIP.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-09-23; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- STEM programs react too late on grades while confidence and self-evaluation erode unnoticed.
- Stakeholders
- engineering student success
- counseling/wellness
- IR/analytics
- IT
- Discovery questions
- Do you collect weekly non-cognitive self-reports?
- Who authors outreach copy today?
- Will you randomize within a semester?
- Value hypothesis
- Bounded value: pilot human-guided personalized outreach and measure PSE before claiming retention impact.
- Potential engagement
- One-semester capped STEM pilot with PSE dashboard and RCT design workshop—not a retention guarantee.
- Unsupported claims
- Causal proof of retention gains
- Guaranteed grade improvements
- Replacement of human advisors/counselors
Pre-sales engineering
Role takeaway
- Fit
- Fits as proactive outreach layer atop LMS/survey pipelines with constrained LLM personalization.
- Architecture and integration
- Survey store → forecast service → template vault → LLM personalizer → human approve → notify.
- Prerequisites
- Weekly PSE/engagement instrument
- Counseling SME
- Approved GenAI vendor
- Constraints
- Historical-control evidence only
- Self-report bias
- WIP status
- Security
- Encrypt self-reports and messages; no consumer tools for identifiable outreach.
- Proof of value
- PSE lift vs contemporaneous control
- Message edit/audit completeness
- Downstream grade/persistence correlation
Delivery
Role takeaway
- Work
- Deploy survey cadence, message library, personalization workflow, and weekly review huddles.
- Dependencies
- Counseling capacity
- Notification channel
- Analytics
- Ownership
- STEM student-success office with counseling psychology partner.
- Skills and adoption
- Train staff to review AI-personalized copy for tone and accuracy.
- Governance checkpoints
- Weekly message audit sample
- Student opt-out path
- IRB if research-framed
- Acceptance criteria
- Documented PSE trajectory
- Signed RCT protocol for next term
- Risks
- Hawthorne effects
- Message fatigue
- Over-reliance on self-report
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Two-stage pipeline: forecasting model on longitudinal self-reports; personalization LLM constrained to expert message cores; human review before send.
Governance
Who approves, reviews and stays accountable for outcomes?
Keep counseling expertise in the loop for tone/safety; treat WIP as hypothesis-generating pending RCT.
Security and privacy
What data, permissions and controls need testing?
Self-report engagement data and personalized messages are sensitive education records; minimize retention and vendor training.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Non-cognitive outreach may support persistence for students with declining confidence; ensure accessible survey/notification channels.
Procurement
What should contracts, pricing and exit terms secure?
Procure forecasting and personalization as modules with human-approval gates; avoid unsupervised generative outreach SKUs.
Operating model
Which teams own the service once it runs?
Shift from midterm reactive alerts to weekly forecast-triggered, human-approved outreach cycles.
Publication history
- 2026-09-22Student Success · Issue 174 resources
Stable resource ID: hayat-human-guided-genai-stem-pse-asee-2026