COGNITIVE LIMITATIONS
&
RISK MANAGEMENT
Laurens Doedes Breuning ten Cate
BA (hons) Business Administration (&finance)
NNBS x SNS
03.08.16
X
Background
Example 1
The mean IQ of the population of eighth graders in a city is known to be 100. You have a selected a random sample of 50 children for a study of educational achievements. The first child tested has an IQ of 150.
What do you expect the mean IQ to be for the whole sample?
Tom is of high intelligence, although lacking in true creativity. He has a need for order and clarity, and for neat and tidy systems in which every detail finds its appropriate place. He seems to feel little sympathy for other people and does not enjoy interacting with others. Self-centred, he nonetheless has a deep moral sense.
Example 2
Which is more probable?
A. Tom is an engineer
B. Tom is an engineer and is an avid player of video games
SNS example 1
Risk & Control Matrix (RCM)
SNS example 2
Scenario Analysis / Stress testing
Conjunction
P (A&B) > P(A) = wrong
P(House market down & Defaults up) > P(House market down)
Research Questions
“How can a retail bank reduce the impact of humans inherent cognitive limitations in risk
assessment & decision-making?”
How can Debiasing be effectively implemented in SNS-bank
Surveys
Solution design
- Top-down as optimal
- IARPA SIRIUS project inspired
- Single training intervention
Sample & Skill
- Bias training group 12% up
- Training had 'confidence effect'
- 44y old, 78.9% Male
- Bias group (n=19)
- Validity of sample
Cognitive Bias Results
1. H0: The performance of risk managers in avoiding cognitive biases has not improved by at least 10% compared to the control group after application of the proposed solution.
1. Extrapolation (MP vs SNS): 61.74% vs 62.6%
2. Training (Pre vs Post): 62.6% vs 19.8% (Δ=42.8%)
3. Control (Pre vs Post): 62.6% vs 57.3% (Δ=5.3% 'noise')
Total: 42.8% - 5.3% = 37.5% performance increase
p>0.05
Predictors & Correlations
- Prediction value check to find validity thru Multi-var linear regression
Independents: Pre-survey variables & 'group variable'
Dependent: Recoded Post-survey variables into 1 performance variable
Model
p<0.05
R = 0.831
GroupNum
R = 0.857
p<0.05
Recommendation
Debiasing training
(top-down applied)
IARPA x Debiasing
- Scalability & long-term peer-reviewed effects
- Single shot solution, minimal time investment
Debiasing - Morewedge et al. (2015)
IARPA - IARPA (2016)
Leidos/522prod/Cretecinc - IARPA (2016)
32% short-term performance increase
24% long-term performance increase
Reasons
