Risk vs. Uncertainty:
Situations of risk involve known alternatives, consequences, and probability distributions. Situations of uncertainty involve a degree of unknown factors, and the optimal solution is often not calculable. Optimization models are typically designed for risk, not uncertainty. Simplification can be beneficial in uncertain situations to reduce estimation error and overfitting.

Examples of Heuristics:
Baseball players use simple heuristics to catch a ball, such as fixing their gaze on the ball and adjusting their running speed to keep the angle of gaze constant. Pilots use a heuristic to decide whether to land a plane safely on a river or return to the airport by fixing the airport tower through the cockpit windshield and observing its movement.

Benefits of Heuristics:
Heuristics can enable better and safer decision-making by ignoring unnecessary information and reducing estimation problems. Heuristics can be taught explicitly, and they can be used consciously or non-consciously.

Systematic Study of Heuristics:
Formalize heuristics: Describe them precisely to assess their performance accurately. Test heuristics competitively: Compare them against complex models to evaluate their effectiveness. Test heuristics in real prediction: Evaluate their ability to predict future outcomes, not just fit historical data.

Application to Customer Retention:
A company with 100,000 customers sends information to 10% of them, resulting in 1% responding. A heuristic suggests sending information to a random 1% of customers, resulting in a similar 1% response rate. The heuristic is simpler and less costly while achieving the same outcome.

Pareto Negative Binomial Distribution Model:
Complex model used in marketing to predict customer activity. Makes several assumptions and relies on parameter estimation from samples. Works well in stable and highly predictable environments with large samples.

Hiatus Heuristic:
Simple heuristic used by experienced managers to classify customers as active or inactive. If a customer hasn’t bought for a certain time interval (e.g., nine months), they are classified as inactive. Pays attention to only a single variable.

Comparison:
Professors of marketing conducted a test to compare the two methods. The hiatus heuristic outperformed the Pareto negative binomial distribution model in three companies.

Less is More:
Simple heuristics often outperform complex models in prediction and decision-making. This “less is more” effect occurs when reducing data and simplifying computations lead to better outcomes.

Heuristic Examples:
In airline, apparel, and CDNOW businesses, a simple heuristic outperformed the complex Pareto negative binomial distribution model. Harry Markowitz’s Nobel Prize-winning mean-variance portfolio model was less effective than the simple “one over N” heuristic in diversifying investments.

Ecological Rationality:
The choice between simple heuristics and complex models depends on the environment and level of uncertainty. In stable and predictable environments with few alternatives and ample data, complex models may be better. In uncertain environments with many alternatives and limited data, simple heuristics may be more effective.

Quantitative Analysis:
Simulations suggest that Markowitz optimization would require approximately 500 years of stock data to consistently outperform the simple “one over N” heuristic.

Banking and Investment:
Banks often promote complex investment strategies based on Nobel Prize-winning models, despite their limitations in uncertain environments.

General Understanding:
The choice between simple and complex approaches should consider the level of uncertainty, the number of alternatives, and the availability of data.

Heuristics and Accuracy:
Heuristics are often viewed as simple and efficient, but they may not always lead to accurate predictions, especially in uncertain environments.

Bias and Variance in Predictions:
In a world of risk, heuristics may be less accurate due to bias, a systematic error in predictions.

Bias-Variance Trade-Off:
In uncertain environments, the bias-variance trade-off comes into play. Variance refers to the sensitivity of predictions to specific sample properties, leading to variability in parameter estimates.

Example of Bias and Variance:
A dartboard analogy is used to illustrate bias and variance. A player with low bias but high variance throws darts all over the board, while a player with high bias but low variance consistently throws darts in a specific area.

Simple Heuristics and Variance:
Simple heuristics with no free parameters, such as the hiatus heuristic or 1 over N, have zero variance. This means that predictions made using these heuristics will only have error due to bias, not variance.

Bias-Variance Trade-Off in Decision-Making:
The goal in decision-making is not just to minimize bias but also to consider the trade-off between bias and variance. Simple heuristics with no free parameters can offer better predictions by reducing variance, even if they have a higher bias.

Simplicity in Modeling:
Overly complex models that try to account for every potential causal variable can lead to systematic mispredictions of the future. Simpler models, while not perfect, can provide more accurate predictions.

Example:
Fitting polynomials of different degrees to temperature data in London. A complex polynomial with 12 degrees of freedom fits the data better than a simpler polynomial with 3 degrees of freedom.

Prediction Performance:
When predicting future temperature values, the simpler 3-degree polynomial outperforms the complex 12-degree polynomial. Even a simple 1-degree polynomial (a line) has a smaller total error than the complex 12-degree polynomial.

Equal Weighting:
As an alternative to complex models, using equal weights for different variables can be effective. This approach avoids the need to estimate weights and can lead to simpler, more accurate models.

Key Points:
Situations of risk differ from those of uncertainty. Despite this distinction, most economic theories treat them similarly. Heuristics, rather than being a sign of human stupidity, are useful tools in a world where optimization is often unrealistic. Robustness, not optimization, should be the goal when dealing with uncertainty. Complex problems do not always require complex solutions. Simple solutions should be considered and valued.

Debunking Pseudo-Experts:
We often rely on experts who may appear impressive but lack practical knowledge and understanding. The “expert problem” involves individuals who claim to possess knowledge and expertise but are often incorrect or misguided. This issue has become more evident in recent years due to the availability of information and the ability to scrutinize claims through platforms like Twitter.

Scientific Papers and P-Values:
The belief in the reliability of scientific papers and p-values can be misplaced. P-values are statistical measures that indicate the likelihood of obtaining results by chance. Replication studies have shown that many scientific findings fail to hold up when tested, highlighting the limitations of relying solely on p-values. Experts often engage in selective reporting, presenting only the most favorable results while omitting those that do not support their claims.

The Problem of Survivorship Bias:
Researchers and experts tend to focus on successful experiments or outcomes while ignoring failures. This survivorship bias can lead to misleading conclusions and an inflated sense of reliability. Experts who conduct multiple experiments and report only the best results are engaging in a form of cherry-picking that distorts the true picture.

The Importance of Heuristics:
Heuristics are mental shortcuts that allow us to make quick decisions in complex situations. While heuristics can sometimes lead to errors, they are often necessary and effective in a world where complete information and perfect optimization are unrealistic. Heuristics can be passed down through generations, embodying practical wisdom and knowledge that may be superior to expert opinions.

Examples of Expert Failures:
Experts have made numerous erroneous predictions and decisions across various fields, including economics, politics, and science. Examples include flawed theories such as dynamic stochastic equilibrium modeling and the selfish gene theory. Even medical practices like lobotomies were once widely accepted despite their harmful effects.

Moving Beyond Experts:
The limitations of experts highlight the need for skepticism and critical thinking. We should not blindly trust experts but evaluate their claims and consider alternative perspectives. Robustness, rather than optimization, should be the goal when dealing with uncertainty. Simple solutions should be considered and valued, even if they appear less impressive than complex mathematical models.

Asymmetry in Rewards and Penalties:
When traders make money, they receive a bonus; however, when they lose money, there is no penalty or negative bonus. This asymmetry creates an environment where experts are not held accountable for their mistakes.

Examples of Lack of Accountability:
The New York Times and Thomas Friedman promoted the Iraq War based on false information about WMDs, but they did not face any financial consequences.

Skin in the Game as a Disciplinarian:
Having skin in the game, or personal financial stake, is the only effective disciplinarian in the financial world. Individuals with skin in the game are more likely to seek simplicity and avoid complexity in their decision-making.

Limited Understanding of Skin in the Game in Academia:
Only a small percentage of academics understand the concept of skin in the game, and even fewer explore its impact. This lack of understanding may lead academics to overlook the importance of simplicity in real-world decision-making.

Comparison of Markowitz and Thorpe’s Projects:
Markowitz’s approach, though intricate, often fails in practice but finds success in academia. Thorpe’s approach, employing simple heuristics, proved successful in blackjack and finance (via the Kelly Criterion).

Kelly Criterion:
Relies on knowing the future and precise joint distribution of returns at infinity. Practical implementation requires only expected return and acceptable loss, enabling continuous portfolio adjustments. Rooted in information theory and computer science, emphasizing bias-variance trade-offs.

Critique of Academic and Pseudo-Expert Culture:
The Journal of Finance, despite its impact on tenure, rarely publishes articles relevant to real-world applications. Pseudo-experts dominate academia, exemplified by the delayed adoption of simple yet effective stress-testing methods at the IMF. Resistance to straightforward solutions like Thorpe’s heuristic and GMO opposition exemplify this culture’s limitations.

Simplicity and Practicality in Finance:
Thorpe’s simple heuristic in blackjack and the Kelly Criterion in finance demonstrate the power of straightforward approaches. The financial world needs practical solutions, not overly complex theories that lack real-world applicability.

The Problem with GMOs:
The first rule of medicine, “first, do no harm,” should be applied to GMOs. Instead of complex and expensive GMOs, a simple solution to world hunger is to solve the distribution problem. GMOs are riskier than natural crops because they involve jumping the “viability island” and have unknown consequences.

The Problem with Experts:
Experts often propose complicated solutions that are not necessarily better than simple ones. The lack of Taylor risk studies and the unsoundness of Markowitz risk metrics indicate that the risk assessment of GMOs is flawed. Experts trained in linear methods may fail to understand the complexity of systems, leading to incorrect conclusions.

Characteristics of Complexity:
Opacity: The inability to see what’s going on in a system. Interaction between components: The way components interact with each other. Fat tails: The dominance of rare events in determining the properties of a system.

Interaction and Generalization:
Generalizing from the behavior of a single fish to a school of fish is invalid due to the interaction between components. This invalid generalization has negatively impacted behavioral economics, as it assumes that market behavior can be predicted based on individual biases, which is not the case.

Biases:
Market behavior cannot be inferred from individual behavior; the collective outcome may differ significantly from individual actions.

Murray Gell-Mann and Complexity:
Murray Gell-Mann, a Nobel Prize-winning physicist and founder of the Complexity Works at the Santa Fe Institute, made a significant discovery.

Illusion of Average Returns:
Investors often believe that average market returns apply to their trading strategies, which is incorrect.

Best Dependence and Sequence:
Best dependence highlights the importance of sequence in decision-making. The order of actions can significantly impact the outcome, as demonstrated by the example of ironing and washing a shirt.

State Space Methods and Average Returns:
State space methods, commonly used in finance, calculate expected returns by considering all possible outcomes. However, these methods fail to account for the impact of specific events, such as a speculator going bankrupt, on the overall average return.

Ensemble Average vs. Time Average:
The ensemble average represents the average outcome over multiple repetitions of an experiment, while the time average is the average outcome over time for a single instance of the experiment. These two averages may not be the same.

Ruin and Ergodicity:
Any speculator subjected to ruin will eventually be ruined. Ruin is only a matter of time. Ergodicity is a condition where the time average and the ensemble average match. Kelly criterion is ergodic and allows for capturing the effective alpha of the market.

Valuation and Alpha Matching:
Classical risk theory considers that valuation based on market return may not apply to individuals who experience ruin or panic dips. Alpha matching with the market occurs only in the absence of ruin.

Murray Gell-Mann’s Critique:
Murray Gell-Mann criticized decision science and decision theory, stating that they are bogus and don’t work. He argued that analyzing risk-taking as one-shot experiments is flawed because individuals are exposed to negative probabilities over time.

Cumulative Risk-Taking:
Risk-taking should be analyzed over time, considering the cumulative effect of multiple risks rather than as static, one-time events. This approach acknowledges that individuals take multiple risks throughout their lifetime, and the cumulative impact of these risks can lead to ruin.

Irrationality of Precaution:
Some experts pathologize concerns about GMOs as irrational, but precaution can be rational when considering the cumulative impact of risks on life expectancy and the ecosystem.

Decision-Making and Tail Probabilities:
Tail probabilities matter overall, not just locally, and should be considered when making decisions. The only way to analyze risks effectively is by looking at how much it reduces life expectancy at different levels (self, family, gene pool, ecosystem).

Applying Gaze Heuristic to Financial Complexity:
Nassim Nicholas Taleb introduced the concept of the gaze heuristic to Andy Haldane, highlighting how dogs solve complex problems under high uncertainty. Haldane was intrigued by this idea and incorporated it into his Jackson Hole talk, titled “The Dog and the Frisbee,” advocating for simplifying the financial system to make it safer.

Skin in the Game as Evolutionary Filter:
Taleb emphasized the importance of skin in the game as an evolutionary mechanism for filtering out risky behaviors in high-impact professions, such as driving on highways. The lack of skin in the game among financial managers can lead to disastrous consequences, as seen in the case of Deutsche Bank.

Hedge Funds as Natural Filters:
Taleb noted that the risk-taking activity has shifted from banks to hedge funds, which typically have more skin in the game. The dynamic ecology of hedge funds, with frequent closures and new entrants, acts as a natural filter, weeding out riskier players.

Common Ideas: Complexity and Skin in the Game:
Both speakers agreed on the need to address complexity in the financial system and the importance of skin in the game for promoting responsible risk-taking.

Defensive Decision-Making in Risk Models:
Complex machine learning models for risk assessment have become prevalent, but Taleb warns against their use in favor of simpler rules. Experienced bankers often rely on a few simple factors for creditworthiness assessment, rather than complex models. Many managers use complex models, often without understanding them, to protect themselves from criticism or lawsuits (defensive decision-making).

Empirical Approach:
Taleb advocates for an empirical approach to determine the effectiveness of simple rules versus complex models in risk assessment.

Skin in the Game:
Skin in the game incentivizes decision-makers to make optimal choices, as they bear the consequences of their decisions. Defensive decision-making is prevalent, especially in the US, with studies showing that over 93% of doctors practice it to protect themselves from legal liability.

Healthcare Example:
Doctors may prescribe unnecessary treatments or tests to avoid being sued, even though they may harm patients in the long run.

Risk Definition:
The speaker questions the definition of risk as the probability of occurrence, suggesting that low uncertainty should correspond with high risk.

Understanding Uncertainty:
Uncertainty is a complex concept with many dimensions, including unknown alternatives, unpredictable consequences, and uncertain probability distributions. Optimization is impossible in situations with high uncertainty.

Heuristics as a Tool:
Heuristics can be useful in dealing with uncertainty when formal probability models are inadequate. However, heuristics need to be studied and taken seriously, not just dismissed as irrational deviations from ideal models.

The Misuse of Probability:
Critiques of neoclassical economics often fall short by blaming human behavior (homo simpsons) rather than acknowledging the limitations of the economic model (homo economicus). Probability in time and space are fundamentally different concepts, and using probability models in situations with rare events can lead to an illusion of certainty.

Addressing Rare Events:
In cases where rare events are so rare that reliable statistics are unavailable, using probability models can be misleading. Caution and modesty are necessary when dealing with uncertainty, and heuristics may be more appropriate than probability models in such situations.

Evidence of Absence vs. Absence of Evidence:
Many professionals, especially those with a PhD in finance, struggle to differentiate between evidence of absence and absence of evidence. This can lead to misinterpretations and incorrect conclusions.

Turkey Problem:
The Turkey problem illustrates how half knowledge can be dangerous. In 1999, most fund managers who had a PhD in finance and joined the hedge fund sector blew up (experienced a large drawdown) in 1998, while those who didn’t have a PhD or were mathematicians fared better.

Example of the Turkey Problem:
A study of fund managers who blew up in 1998 revealed that 38 out of 39 professors of finance experienced an LTCM-like drawdown, except for one. The one who didn’t blow up had a background in mathematics, while the others had half knowledge in finance.

Conclusion:
Half knowledge can exacerbate the Turkey problem, leading to incorrect assumptions and poor decisions in finance.

Basel Accords Increasingly Complex:
Complexity of banking regulations has been increasing: Basel 1 had 30 pages, Basel 2 had 300 pages, and Basel 3 has over 600 pages. Basel III includes the term “simplicity” in its papers, but whether that translates into actual simplicity is questionable.

Risk Computations Based on Risk, Not Uncertainty:
Value at risk computations are based on the world of risk, not uncertainty. Large banks must estimate thousands of risk parameters and a covariance matrix in the order of a million. These estimates are largely guesswork and border on astrology.

Ineffectiveness of Complex Models:
Complex models have not prevented or foreseen any crises. Alternatives to complex models exist, but there is resistance to adopting them due to the illusion that complexity is necessary to handle complex situations.

Simpler Models Offer Robust Solutions:
Simpler models are more robust and easier for regulators to monitor. Regulators can more easily identify when banks are gaming the system with simpler models.

Heuristics:
Heuristics are not inherently good just because they are heuristics. Some people use bad heuristics and end up making poor decisions.

Regulation and Skin in the Game:
Regulation without skin in the game can lead to problems.

Regulation Creates Opportunities:
Regulators create opportunities for traders to find ways to make money within the new rules.

Example of Regulation Creating Opportunities:
When shorting on a downtick was regulated, traders created structures that allowed them to do it anyway.

Arbitrage Opportunities from Regulation:
Regulation can lead to arbitrage opportunities for traders.

Regulators Finance Traders:
The speaker believes that regulators finance traders by creating opportunities for them to make money.

Japan as an Over-Regulated Market:
Japan is cited as an example of an over-regulated market where traders can do anything with a good lawyer.

The Hosts Get Nervous:
The speaker’s comments about regulators financing traders make the hosts nervous, leading them to want the speaker to leave.