Trust & Transparency

Student accuracy is calculated using the SUM/SUM methodology, which ensures consistent, fair measurement across all contexts:

This canonical formula is used identically across Teacher Dashboards, Admin Dashboards, and all reports. We do not average percentages—we aggregate raw counts first, then calculate a single percentage. This eliminates mathematical artifacts that could misrepresent student performance.

• Teacher and Admin dashboards show identical accuracy values (0% delta)
• All API endpoints use the same calculation module
• Automated regression tests verify parity continuously

The Data Confidence Score reflects how reliable the displayed metrics are, based on the volume and recency of student activity:

The Rigor Engine analyzes question difficulty using multiple factors to ensure fair assessment and appropriate challenge levels:

• Depth of Knowledge (DOK): Questions are tagged DOK 1-4, measuring cognitive complexity from recall to extended thinking.
• Text Complexity: Lexile measures, sentence structure, and vocabulary density are factored into difficulty ratings.
• Historical Performance: Aggregate success rates inform difficulty calibration without penalizing individual students.
• Standards Alignment: Questions map to state standards and NC EOG benchmarks for grade-appropriate rigor.

The Rigor Engine does not modify scores retroactively. It informs adaptive pathways and coaching recommendations.

All AI-generated content in Lit-Levels passes through multiple safety layers:

1. Content Filtering: Generated stories and feedback are screened for age-appropriateness and educational value.
2. Bias Detection: Automated checks flag potentially biased language or cultural insensitivity for human review.
3. Factual Grounding: AI tutoring responses cite specific passage evidence rather than generating unsupported claims.
4. Human Oversight: Teachers can review, edit, or reject AI-generated content before student exposure.

Lit-Levels undergoes continuous automated testing (TESLA Certification) to ensure data integrity:

Our systems continuously monitor for data anomalies that could indicate technical issues or integrity concerns:

• Statistical Outliers: Scores significantly above or below historical norms are flagged for review.
• Timing Anomalies: Unusually fast response patterns trigger verification checks.
• Duplicate Detection: Idempotency keys prevent accidental double-counting of rewards or progress.
• Cross-Reference Validation: Student data is validated against roster records and enrollment status.

Detected anomalies are logged in the Admin Integrity Panel for transparent investigation by authorized personnel.