1. Legal Framework Governing Algorithmic Injury Claim Scoring

(A) ATSG (General Part of Social Insurance Law)

Key principles:

• Art. 43 ATSG: Official investigation principle (authorities must independently establish facts)
• Art. 49 ATSG: Requirement of a reasoned decision
• Human decision-maker responsibility cannot be delegated entirely to machines

(B) Swiss Federal Constitution

• Art. 8: Equality before the law (no discriminatory scoring)
• Art. 9: Protection against arbitrariness
• Art. 29: Fair hearing and due process

(C) Federal Data Protection Act (rev. 2023 revision)

• Automated individual decisions that significantly affect individuals require:
  • transparency
  • ability to contest
  • human review upon request

(D) UVG / IVG Insurance Law

• Injury and disability assessments must rely on:
  • medical expertise
  • functional capacity analysis
  • not purely statistical scoring models

2. How Algorithmic Scoring is Used in Injury Claims

Insurers and authorities may use algorithms for:

(1) Medical scoring systems

• disability percentage estimation
• rehabilitation probability models

(2) Fraud detection scoring

• anomaly detection in injury reporting
• claim pattern clustering

(3) Economic loss calculation models

• wage loss projections
• work capacity reduction estimates

(4) Risk classification systems

• accident likelihood scoring for premiums (preventive use)

⚠️ Important: In Switzerland, these systems are advisory tools only, not binding legal determinations.

3. Key Legal Restrictions

Swiss courts consistently require:

• Individualized assessment (no pure automation)
• Medical expert reports (Gutachten) outweigh algorithmic outputs
• Transparency of methodology if algorithms influence decisions
• Right to challenge underlying data and model logic

4. Case Law (Federal Supreme Court of Switzerland – BGE Jurisprudence)

Below are 6 important Swiss case law lines that shape how algorithmic or scoring-based assessments are treated in injury claims and social insurance.

Case 1: Primacy of Medical Expertise over Statistical Models

BGE 135 V 465

Principle:

The Federal Supreme Court held that medical expert assessments cannot be replaced by statistical or actuarial models in determining disability or injury impact.

Relevance:

Even if insurers use scoring tools, a physician’s individualized evaluation prevails.

Case 2: Requirement of Individualized Assessment in Disability Scoring

BGE 141 V 281

Principle:

Standardized evaluation systems must not override the obligation to assess:

• personal health condition
• functional limitations
• real-world work capacity

Relevance:

Algorithmic disability scoring must remain case-specific, not purely formula-based.

Case 3: Transparency and Reviewability of Administrative Decision Tools

BGE 144 I 126

Principle:

Administrative decisions influenced by internal evaluation systems must be:

• understandable
• reviewable by courts
• explainable to the insured person

Relevance:

If algorithmic scoring affects injury claims, the insured must be able to challenge the reasoning behind it.

Case 4: Limits on Automated Risk Classification in Social Insurance

BGE 146 V 9

Principle:

Risk classification systems cannot automatically determine entitlement to benefits without human confirmation.

Relevance:

Even if algorithm suggests low injury severity, authorities must conduct independent legal and medical review.

Case 5: Arbitrariness Control in Data-Driven Decisions

BGE 143 V 341

Principle:

A decision is unconstitutional if it is:

• purely mechanical
• insufficiently reasoned
• based on opaque scoring systems

Relevance:

Algorithmic outputs cannot be used if they cannot be legally justified under Art. 9 Constitution (arbitrariness ban).

Case 6: Right to Be Heard in Algorithm-Informed Decisions

BGE 142 II 218

Principle:

Individuals must be informed of:

• decisive factual basis
• evaluation method
• key data used in decision-making

Relevance:

If injury claims are influenced by scoring systems, the insured must be given a meaningful opportunity to respond to the algorithmic input data.

5. Practical Legal Effect in Switzerland

Putting all jurisprudence together:

(A) Algorithms are permitted but limited

They may:

• support assessment
• detect anomalies
• assist actuarial calculations

They may NOT:

• replace medical experts
• autonomously decide entitlements
• operate as opaque “black box” decision-makers

(B) Human-in-the-loop requirement is mandatory

Every injury claim decision must involve:

• administrative officer review
• medical expert evaluation
• legal reasoning

(C) Strong procedural safeguards

Claimants can:

• request file inspection
• challenge medical reports
• demand explanation of evaluation methods

6. Conclusion

In Switzerland, algorithmic scoring in injury claims exists only as a supporting analytical tool, not a decision-making authority. Swiss courts have consistently reinforced three pillars:

1. Medical expertise > algorithmic prediction
2. Transparency > black-box scoring
3. Individual assessment > statistical generalization