From the College Athletics edition of September 24, 2026
MDPI review: AI injury-prediction ethics lag safeguards across collegiate and elite sport
MDPI / Academy of Physical Education Katowice et al. · Academic sports science / AI ethics · International (collegiate contexts included)
- Publisher
- MDPI AI
- Original publication
- November 1, 2025
- Source retrieved
- 2026-09-25
What happened
Narrative review of 24 studies on AI-driven injury forecasting finds recurring gaps in privacy, fairness, informed consent, athlete autonomy, and long-term data governance despite reported model performance tools (e.g., SHAP). HIPAA, FERPA, and GDPR are judged only partial and inconsistently applied; collegiate settings appear among under-protected contexts. Authors recommend enforceable athlete data rights, participatory oversight, explainability with contestability, and independent audits. Newly relevant now as college S&C assistants (e.g., FYTT Flex) and performance-tech stacks expand AI into programming and risk workflows without published ethics audits.
Why it matters
Independent academic caution for collegiate athletic departments deploying injury-prediction or load AI—constrains how Flex-class and wearable stacks should be governed.
Evidence and measured results
PRISMA-guided narrative synthesis of 24 studies; thematic findings and recommendation tables; explicit collegiate citations in governance discussion.
Limitations and uncertainty
Narrative review (search ~Jan 2025); not a new college RCT; heterogeneity limits quantitative pooling; Poland-based authorship synthesizing global literature.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-09-25; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Departments adopting injury/load AI lack clear athlete-data rights and contestability, creating compliance and trust risk alongside performance promises.
- Stakeholders
- Sports medicine / team physicians
- Compliance / privacy office
- Student-athlete representatives
- Athletics administration
- Discovery questions
- Can a student-athlete see and appeal an AI risk label?
- Which laws (FERPA/HIPAA/state biometric) does counsel map to wearable vendors?
- Do predictions influence scholarships, playing time, or NIL without human review?
- Value hypothesis
- Governance-first discovery reduces deployment risk by surfacing rights gaps before contracts lock in opaque vendor models.
- Potential engagement
- Ethics-and-controls assessment workshop mapping one proposed injury model against the review’s five concern axes before procurement.
- Unsupported claims
- That this review proves any specific vendor is unsafe
- Guaranteed legal compliance from reading the paper alone
- Transfer of pro-league CBA norms wholesale to NCAA without adaptation
Pre-sales engineering
Role takeaway
- Fit
- Cautionary requirements stack for injury-prediction and load-AI systems in collegiate environments—not a product SKU.
- Architecture and integration
- Human-in-the-loop medical override; explainability outputs; access controls separating coaching vs medical uses; optional federated learning patterns.
- Prerequisites
- IRB/ethics review when research-like
- Representativeness checks for women/youth/disability subgroups
- Documented injury definition standards
- Constraints
- Partial FERPA/HIPAA coverage
- Vendor secrecy limiting auditability
- Small collegiate sample sizes
- Security
- Minimize biometric retention; ban undisclosed secondary AI training use; encrypt and role-limit health-adjacent features.
- Proof of value
- Bias audit report before production
- Appeal-path tabletop exercise
- Calibration and false-flag review with team physicians
Delivery
Role takeaway
- Work
- Policy drafting, consent UX, ethics board standup, vendor questionnaire aligned to privacy/fairness/autonomy/governance axes.
- Dependencies
- University privacy/legal
- Faculty athletics representative
- Vendor transparency
- Ownership
- Sports medicine leads clinical use; privacy office leads data rights; AD sponsors policy.
- Skills and adoption
- Train coaches that AI flags inform—not replace—medical decisions; educate athletes on rights and appeal.
- Governance checkpoints
- Pre-procurement ethics review
- Go-live consent audit
- Semester fairness/privacy retrospective
- Acceptance criteria
- Written athlete appeal process live
- No automatic roster exclusion from AI score alone
- Vendor contract includes data-deletion and training-use limits
- Risks
- Automation bias benching healthy athletes
- Predictive labels leaking into NIL/scholarship talks
- Paper’s international lenses mis-applied without U.S. counsel
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?
Favor explainable models, human-in-the-loop gates, federated or minimized centralization where possible, and athlete-facing contestability interfaces—not opaque vendor scores alone.
Governance
Who approves, reviews and stays accountable for outcomes?
Institutions should codify data rights, ethics boards with athlete voice, and appeal paths before predictive labels affect playing time, scholarships, or medical decisions.
Security and privacy
What data, permissions and controls need testing?
Biometric and health-adjacent streams need dynamic consent, retention limits, and limits on secondary commercial use; FERPA/HIPAA alone are insufficient design specs.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Calls out underrepresentation of women, youth, and Paralympic athletes in training data—equity risk for collegiate multi-sport departments.
Procurement
What should contracts, pricing and exit terms secure?
Use review axes as RFP/security questionnaire requirements for injury and wearable AI vendors serving NCAA clients.
Operating model
Which teams own the service once it runs?
Pushes operating model from coach-only dashboards toward shared athlete-medicine-privacy governance with contestability.
Publication history
- 2026-09-24College Athletics · Issue 184 resources
Stable resource ID: mdpi-ethical-bias-ai-injury-prediction-2025