AI Governance & Risk Scoring · Nigerian Legal Practice
Govern AI-Assisted Legal Work. Verify Before You Trust.
Lawluate

Help & Knowledge Centre

User guide, model selection guidance, and frequently asked questions.

Part A — User Guide

  1. 1
    Select Legal Domain and AI Task
    Choose the area of law and the intended use of AI.
  2. 2
    Review Evidence-Informed Suggestions
    See three recommended models based on benchmark data.
  3. 3
    Select an AI Model
    Choose any model; recommendations are advisory only.
  4. 4
    Complete the Duty of Inquiry Assessment
    Answer eight mandatory governance questions with Yes or No.
  5. 5
    Review Assessment Results
    View the Governance Risk Score, Mandatory Governance Actions (for each "No"), and Risk-Level Recommended Actions.
  6. 6
    Download PDF Audit Report
    Generate a professional governance audit report for your records.

Part B — Model Selection Guide

The Model Selection Advisor ranks the six supported AI models for the chosen legal domain, AI task, and practitioner priority. The composite recommendation score is:

Score = 0.50 × Global Model Weight + 0.35 × Domain Adjustment + 0.10 × Task Heuristic + 0.05 × Priority Heuristic

Lower scores rank higher. Global model weights and domain adjustments for Contract Law, Tort Law, and Commercial Law are derived from the NaijaLaw benchmark. Criminal Law has limited evidence (one case only), so its domain adjustment defaults to 0.00 with a "Limited evidence" notice. Procedural Law, Document Hierarchy, Land/Mortgage Law and Banking Law are not yet benchmarked and display a "Not yet benchmarked" notice.

  • Task and practitioner-priority adjustments are expert-informed heuristics — not benchmark measurements.
  • All six models remain selectable regardless of ranking; the advisor supports professional judgement, it does not replace it.
  • Whichever model is chosen, the Duty of Inquiry Assessment and Governance Risk Score are calculated identically.

Part C — Frequently Asked Questions

Part D — Validation and Research Integrity

Lawluate communicates the categories and purpose of its validation processes while retaining detailed benchmark-development methods as proprietary research and implementation information.

Accuracy Validation

Evaluates whether benchmarked AI outputs align with verified legal outcomes used in the research evidence. This helps ensure that recommendation values are based on demonstrated performance rather than subjective model preference.

Confidence Calibration

Evaluates whether a model's expressed confidence is proportionate to its observed performance. This helps distinguish dependable confidence from confidently presented error and reduces overreliance on raw accuracy alone.

Legal Reasoning Validation

Evaluates whether model outputs demonstrate legally supportable reasoning rather than merely reaching a plausible conclusion. This reinforces the requirement to verify authorities, jurisdiction, reasoning, statutory basis, and factual application.

Implementation Validation

Confirms that Lawluate applies the approved evidence configuration correctly, calculates recommendation scores consistently, displays the intended evidence status, and preserves the separation between model recommendations and governance-risk assessment.

Validation strengthens confidence in the evidence and its implementation, but it does not certify that every future AI response will be correct, complete, unbiased, current, or suitable for professional reliance. Every AI-assisted legal output must still be independently verified against authoritative legal sources.