Problem:
I'm focusing on risk assessment across various assets in a Uni Finance group. Each person provides a risk score and a confidence score for multiple metrics (like competition, dependencies, team, etc.) associated with each asset. The risk score ranges from 1 to 10, with 1 indicating low risk and 10 indicating high risk. The confidence score ranges from 1 to 3, where 1 means low confidence in the risk assessment and 3 means high confidence (i.e., its a way to measure how confident someone is in their assessment - as one individual may be closer/know more on that asset than another individual).
Objective:
We aim to develop a formula that computes an Adjusted Risk Score (ARS) for each metric of each asset, considering both the risk score and the confidence score provided by each person. The ARS should give more weight to risk scores with higher confidence and less weight to risk scores with lower confidence. The ultimate goal is to aggregate these ARS values across the group to produce a comprehensive risk assessment for each asset.
Challenges:
- High Confidence, Low Risk: If a person is highly confident and gives a low-risk score, the ARS should reflect a low-risk assessment but be "amplified" due to the high confidence.
- High Confidence, High Risk: If a person is highly confident and gives a high-risk score, the ARS should reflect a high-risk assessment and be "amplified" due to the high confidence.
- Low Confidence, High/Low Risk: If a person has low confidence, regardless of the risk score, their assessment should have a diminished impact on the ARS.
- Interpretability: The ARS should be interpretable as a risk measure, meaning higher ARS values should unambiguously indicate higher risk and vice versa.
Previous Approaches:
We have tried various mathematical approaches to solve this problem, such as logarithmic transformations and sigmoid functions, to model the influence of the confidence score. However, I haven't yet arrived at a satisfactory solution that meets all the objectives and challenges outlined above.
End Goal:
To find a mathematical formula that accurately computes the ARS by balancing both the risk score and the confidence score in a way that aligns with our objectives and solves the challenges we face.
