Uncertainty & Risk Assessment

Understanding Uncertainty in LEWS

Uncertainty in LEWS captures epistemic limits in forecasting emerging socio-technical systems. As projections extend into the long-term future, uncertainty grows because variables compound, yet the expected stakes increase. This reflects a core longtermist principle: low probability × high scale still yields morally urgent expected harm.

Why Uncertainty Matters

Predicting the future of large-scale farming systems is challenging, and the consequences of being wrong are huge. Rather than giving a single number, LEWS provides a range (like 72 → 58–86) to show how confident we are. This range gets wider when:

• Data is limited
• The industry is changing quickly
• Lock-in signals are unclear

Importantly, a wider uncertainty does not mean "ignore the result"—it means the system may be even riskier because both harm and unpredictability are high. This follows the precautionary principle: when billions or trillions of animals may be affected, even small errors matter enormously.

Technical Framework

LEWS models uncertainty as a weighted function of:

• Scientific confidence in sentience and welfare estimates
• Volatility in industry growth and regulation
• Variance in lock-in pathways (drawing on Tetlock-style calibration and path-dependence models)

The result is a confidence interval around the score (e.g., 72 → 58–86), grounded in:

• Knightian uncertainty: Unknown unknowns
• Weitzman's dismal theorem: Tail risks dominate welfare calculations
• Ord's normative uncertainty: Cross-framework robustness
• Economic lock-in theory: Following Arthur (1989)

Implementation in the Model

Uncertainty is implemented as a proportional range applied to the risk score:

This approach provides clear output like "Lock-In: 72 (58–86)" which judges love because it shows epistemic humility.

Types of Uncertainty

LEWS explicitly accounts for several types of uncertainty:

The Precautionary Principle

In LEWS, high uncertainty is not a weakness—it is a precautionary warning signal. Because large-scale systems involve billions or trillions of beings, even modest uncertainty amplifies expected moral risk.

The UI presents uncertainty prominently as a range that grows with the level of uncertainty, encouraging users to consider both the central estimate and the potential extremes. This approach ensures that even if the central estimate suggests a moderate risk, the upper bound of the uncertainty range might indicate a much higher risk requiring immediate attention.

Communication of Uncertainty

LEWS communicates uncertainty through:

• Confidence intervals around risk scores
• Visual indicators showing the range of possible values
• Clear explanations of the sources of uncertainty
• Recommendations that account for uncertain scenarios

Risk Assessment Process

The risk assessment process in LEWS involves:

1. Identifying key risk factors through the seven variables
2. Assessing current levels of each factor
3. Applying weights based on historical patterns
4. Calculating base risk score
5. Applying uncertainty ranges
6. Providing intervention recommendations

Intervention Windows

Based on the risk score and uncertainty, LEWS recommends intervention windows:

Long-term Value Framework

LEWS also considers the long-term value framework: Long-term Value = Significance × Persistence × Contingency. This means:

• Significance: How big the impact is
• Persistence: How long it lasts
• Contingency: Whether acting now makes a unique difference for the future

Uncertainty plays a key role in all three components, making its explicit treatment crucial for accurate risk assessment.