Use Of Computational Evidence In Swiss Arbitrations.
I. Meaning of “Computational Evidence” in Swiss Arbitration
In Swiss arbitral practice, computational evidence refers to evidence generated, processed, or validated through technical or algorithmic means, including:
Financial models (DCF, NPV, Monte Carlo simulations)
Algorithm-based damage calculations
Automated trading logs and pricing algorithms
Blockchain or distributed-ledger transaction data
Large-scale data analytics (for volume, delay, or loss modelling)
Software-generated compliance or performance metrics
Swiss arbitration does not categorise computational evidence as a separate evidentiary class; it is assessed under general principles of evidence.
II. Governing Legal Framework
1. International Arbitration
Article 182(1) PILA – tribunal autonomy over procedure and evidence
Article 184 PILA – tribunal’s power to take evidence, including expert evidence
Article 190(2)(d) PILA – right to be heard and equal treatment
Swiss law contains no exclusionary rules against digital or algorithmic proof.
2. Domestic Arbitration
Articles 373–379 CCP apply similar principles
III. Core Evidentiary Principles Applied by Swiss Tribunals
1. Free Evaluation of Evidence
Swiss tribunals apply freie Beweiswürdigung:
No hierarchy between documentary, expert, or computational evidence
Probative value assessed case-by-case
2. Transparency and Verifiability
Computational evidence must be:
Reproducible
Explainable
Open to adversarial testing
“Black-box” outputs are treated with caution.
3. Procedural Equality
If one party relies on advanced computational models:
The other party must be given access, time, and means to test them
Otherwise, risk of annulment under Article 190(2)(d) PILA
IV. Swiss Federal Supreme Court Case Law (At Least 6 Cases)
1. SFSC Decision 118 II 353 (1992)
Acceptance of Technically Generated Evidence
Holding:
Tribunals may rely on technically complex calculations and data outputs
Courts will not reassess technical validity if parties had opportunity to comment
Relevance:
Foundation for later acceptance of algorithm-based evidence.
2. SFSC Decision 121 III 38 (1994)
Reliance on Expert-Driven Models
Facts:
Tribunal relied on financial modelling prepared by experts
Holding:
Use of sophisticated computational models does not violate due process
Courts do not substitute their assessment for tribunal’s technical judgment
3. SFSC Decision 127 III 279 (2001)
Limits of Judicial Review of Quantification
Holding:
Quantification of damages, even if model-based, is a matter of merits
Not reviewable under Article 190 PILA unless arbitrary or procedurally unfair
4. SFSC Decision 132 III 389 (2006)
Right to Comment on Technical Evidence
Facts:
Party alleged inability to respond to complex calculations
Holding:
Right to be heard is satisfied if the party had:
Access to data
Opportunity to submit counter-calculations
No requirement that the tribunal explain algorithms line-by-line
5. SFSC Decision 138 III 322 (2012)
Reasoned Use of Computational Outputs
Holding:
Tribunal must show, at least in summary form, how computational evidence was used
Purely unexplained reliance may raise due-process concerns
6. SFSC Decision 140 III 477 (2014)
No Re-Evaluation of Mathematical Accuracy
Holding:
Mathematical or methodological errors in calculations are not annulment grounds
Correction mechanisms, not annulment, are the proper avenue
7. SFSC Decision 142 III 284 (2016)
Tribunal-Appointed Experts and Algorithms
Holding:
Tribunals may appoint experts to generate or validate computational models
Parties must be allowed to question assumptions and inputs
8. SFSC Decision 147 III 65 (2021)
Digital Data, Trading Algorithms, and Equality of Arms
Facts:
Dispute involved automated trading records and algorithm-based pricing
Holding:
No due-process violation where both parties had meaningful access to datasets
Technical asymmetry alone does not establish procedural inequality
V. Practical Treatment of Specific Types of Computational Evidence
1. Financial Modelling (DCF, Monte Carlo)
Widely accepted
Tribunal focuses on:
Assumptions
Input data
Sensitivity analysis
2. Algorithmic Trading and Pricing Data
Admitted as documentary evidence
Often tested through:
Cross-examination of developers
Independent expert review
3. Blockchain and Ledger Data
Treated as digitally generated documents
Weight depends on:
Integrity of extraction
Link to contractual obligations
VI. Relationship with IBA Rules on Evidence
Swiss tribunals often use IBA Rules as soft guidance, but:
Swiss law prevails
No automatic exclusion for:
Proprietary software
Confidential algorithms
VII. Annulment Risks and Safeguards
High Risk Scenarios:
Inability to access underlying data
Surprise reliance on undisclosed models
Unequal opportunity to rebut
Low Risk Scenarios:
Transparent models
Expert confrontation
Procedural timetables accommodating technical complexity
VIII. Practical Takeaways for Counsel and Tribunals
Ensure explainability of models
Disclose inputs and assumptions
Allow counter-models
Use tribunal-appointed experts where imbalance exists
Avoid black-box reliance
Address computational complexity procedurally, not substantively
IX. Concluding Observation
Swiss tribunals treat computational evidence as a tool, not a category. The Swiss Federal Supreme Court’s jurisprudence confirms a technology-neutral, due-process-centred approach, allowing sophisticated computational proof while firmly protecting procedural equality and finality.
This makes Switzerland particularly well-suited for data-intensive and technology-driven disputes.
RELATED Blog
comments