Emerging Risk Research
Exploring the frontiers of risk identification and analysis with innovative approaches.
Rethinking Impact Assessment
Our research explores the philosophical and practical challenges in applying traditional risk assessment approaches to emerging risks, representing a fundamental paradigm shift in risk analysis.
We focus on probabilistic networks, conditional relationships, and uncertainty characterisation rather than definitive impact quantification, requiring a recalibration of mental models and communication strategies.
Key Concepts
For decades, risk management has been anchored in the probability-impact framework:
- Assess likelihood of risk occurrence
- Assess severity if the risk materialises
- Multiply to get an expected value or plot on a risk matrix
- Prioritise based on combined scores or matrix position
This approach works well for mature risks with rich historical data and well-understood causal mechanisms.
Emerging risks fundamentally differ from mature risks:
- Deep uncertainty about causal mechanisms
- Unknown and potentially novel impacts
- Lack of statistical basis for estimation
- Potential for unprecedented outcomes
- Complex relationships between risks
- Cascade effects through systems
These characteristics make traditional impact assessment problematic and potentially misleading.
The Bayesian network approach with MCMC simulation offers several advantages:
- Explicitly represents uncertainty in probabilistic terms
- Updates beliefs as new information emerges
- Maintains full probability distributions
- Focuses on relationships and conditions
- Enables dynamic risk intelligence
- Supports continuous updating and scenario exploration
Detailed Analysis
Attempting premature quantification of impacts for emerging risks creates several problems:
- False precision: Creating an illusion of knowledge where significant uncertainty exists
- Anchoring bias: Early numerical estimates unduly influence subsequent thinking
- Misallocation of resources: Focusing mitigation efforts based on potentially arbitrary impact assessments
- Overlooking systemic effects: Missing complex interactions that may amplify or transform impacts
Rather than forcing premature impact quantification, our approach:
- Explicitly represents uncertainty in probabilistic terms
- Updates beliefs as new information emerges
- Maintains the full probability distribution rather than collapsing to point estimates
- Allows for multiple possible scenarios
The network structure emphasises what we often can assess better than impacts:
- Conditional probabilities between risks
- Influence strengths and mechanisms
- Cascade paths and critical nodes
- System leverage points
Traditional approach might attempt to quantify:
- Financial impact of AI incidents in dollars
- Reputational damage on an arbitrary 1-5 scale
- Regulatory fines based on current frameworks
- Business interruption costs using standard models
Our network approach instead focuses on:
- How AI risks influence and are influenced by other technology risks
- Conditional probabilities of risk materialisation given various triggers
- Evolution of risk factors over time as the technology matures
- Key indicators that would signal increasing risk probability
The latter provides more actionable intelligence for truly emerging risks, even without definitive impact quantification.
Implementation Challenges
Executives and board members often expect simple, quantified answers:
- "What's the potential financial impact?"
- "Where does this rank on our risk matrix?"
- "What's the worst-case scenario?"
- "What's our exposure in dollars?"
Our approach requires reframing these conversations to focus on system behaviour rather than isolated impacts.
To successfully implement this new approach while meeting organisational needs:
- Emphasise the different nature of emerging vs. mature risks
- Provide alternative value metrics like critical dependencies and early warning indicators
- Create qualitative impact categories when quantification isn't feasible
- Develop conditional scenarios that link to existing risk frameworks
Transforming Risk Analysis
Our approach represents more than just a new tool – it embodies a fundamental shift in how we conceptualise, analyse, and communicate about emerging risks.
"The greatest risk in times of turbulence is not the turbulence itself, but acting with yesterday's logic." – Peter DruckerExplore Our Prototypes