Super Models Available Here
Book Review:
Gabriel Weinberg and Lauren McCann, "Super Thinking: Upgrade Your Reasoning and Make Better Decisions with Mental Models", Penguin Business, 2019.
If you hang around the same corners of the internet as me, there is a chance you've encountered the Farnam Street blog's "latticework of mental models" or the site Conceptually. Both promote the idea that so-called cognitive tools or mental models can enhance one's understanding of the world and help one make better decisions, by importing concepts from various fields and applying them more broadly. Economics, psychology, philosophy, physics, etc. all have their own sets of frameworks, shortcuts and models that help their practitioners explain things and cut through complexity. Yet by taking a multidisciplinary approach and analyzing a problem from different perspectives, we gain a more well-rounded understanding and reduce our blind spots. And with a "latticework of models" we move beyond just memorizing isolated facts. This is the approach that has famously been advocated by Charlie Munger, the long-time investing partner of Warren Buffett at Berkshire Hathaway.
Charlie Munger is also quoted in Super Thinking, the new book by Gabriel Weinberg and his wife Lauren McCann. Some of said Munger quotes:
"You've got to have models in your head. And you've got to array your experience -- both vicarious and direct -- on this latticework of models." (p. viii)
"And the models have to come from multiple disciplines -- because all the wisdom of the world is not to be found in one little academic department." (p. x)
"You don't have to know it all. Just take in the best big ideas from all these disciplines. And it's not that hard to do." (p. xi)
"Once you have the ideas, of course, they're no good if you don't practice. If you don't practice, you lose it. So I went through life constantly practicing this multidisciplinary approach." (p. 318)There you have the inspiration for this book. As for the authors, Gabriel Weinberg is the founder of search engine DuckDuckGo and Lauren McCann is a statistician who works on drug trials. Both are MIT alumni. Essentially, they wrote the book they wish they had when they were younger -- one that taught broadly useful concepts, or "super models", that unlock the ability to think at higher levels. (Yes, I know, it sounds somewhat cheesy.) Keeping in line with the title, the book's cover is designed to match the bright blue-yellow-red color scheme of Superman:
While I read the UK version, the American version has a different subtitle: "The Big Book of Mental Models". Beyond that I don't think there are any differences. The book is divided into an introduction, nine chapters, and a brief conclusion. It has an index but no bibliography. Each chapter has a unifying theme and the mental models are connected by narrative, but ultimately the book is just a long list of concepts. I could simply repeat the list of concepts here, but I think a lot of the book's added value lies in the way it organizes them -- so I'll summarize the most important takeaways from each chapter. (And besides, the book's website has a list anyway.)
***
I.Chapter 1 is curiously titled "Being Wrong Less", which alludes to LessWrong (although I'm not sure if it was intentional). This chapter is about solving problems while avoiding the many biases our minds have evolved. To begin, you should try to keep things simple by making only the most fundamental assumptions and building from there -- this is arguing from first principles, which helps you better understand the subject and avoid traps of conventional wisdom. You can test those assumptions (de-risking) by prototyping a minimum viable product, for instance. This helps you avoid doing too much work using mistaken assumptions (which is called premature optimization). One source of mental traps is the availability bias, which is when recent information or experiences distort your perception -- it can arise from your frame of reference or filter bubble, among other things.
A related model is the fundamental attribution error: attributing people's actions to their internal motives rather than external situation. To overcome it, try to increase your empathy with other people (e.g. by looking for a third story that an impartial observer would see). Other traps include the confirmation bias (interpreting information in a way that confirms your preexisting beliefs), and the optimistic probability bias (believing that something is likely, because you want it to be true). Challenge your intuition by seeking out a variety of perspectives (e.g. by taking the Devil's advocate position), try thinking gray (rather than black-or-white), and look for the root cause of an event.
II.
The next chapter continues the theme on problem-solving and planning but with more focus on avoiding unintended consequences. The title is a reference to Murphy's law, which pithily states that "anything that can go wrong, will go wrong". Unintended consequences are often seen in the case of market failure; for example, when individual fishermen follow their rational self-interest by catching more fish, the stock of fish will get collectively depleted and everyone suffers -- this is an example of the tragedy of the commons. It is likely to happen to public goods (e.g. national defense, air quality etc.) due to people using a resource without paying for it (the free rider problem). When a factory pollutes the surrounding air (spillover effects), residents experience negative consequences that they didn't ask for; in economics these are called externalities. One way to ameliorate this is to apply the Coase theorem to internalize the externality, for instance through a cap-and-trade scheme. Finally, markets may fail due to asymmetric information: or when one side of a transaction has more information than the other. This is the case for real estate agents who know the market better than you, but might not get the best price for you unless it also maximizes their commission (the principal-agent problem). Beyond market failure are more general cases of perverse incentives. For instance, Goodhart's law says that when a measure becomes a target, it will cease to be a good measure (e.g. because people focus only on the measure, often in unintended ways).
There are also boiling frog problems: when a decision that seems good in the short-run turns out suboptimal in the long-run, for example the kind of short-termism that leads to technical debt (quick fixes that will require costlier repairs in the future) or unfavorable path dependence (when your decisions limit the set of paths available to you). Ideally, you'd be preserving optionality by making choices that preserve future options -- but having too many options and too much information is not ideal either (see: the paradox of choice and information overload). Avoid analysis paralysis by recognizing reversible decisions and not focusing too much on them. For irreversible decisions that involve risk, keep in mind the precautionary principle ("better safe than sorry").
III.
Chapter 3 is about how to spend one's time wisely and in a manner that furthers one's personal mission and vision (which the authors refer to as a "north star"). The first thing to realize is that you can't do everything -- there are trade-offs in spending one's time, and multitasking can lead to worse performance when the task requires high concentration. Keep one activity the top idea in your mind, so that your brain can generate creative solutions. But how do you know what is important? Consider the option with the lowest opportunity cost: what are the explicit and implicit costs of giving up the best alternative? In negotiations, don't accept an offer if it's worse than your BATNA (best alternative to a negotiated agreement). The concept of leverage can also be useful -- just as a mechanical lever can amplify a small force, certain activities have a much greater effect given the money or time you invest. The Pareto principle (or 80/20 rule) says that 80 percent of the results come from about 20 percent of the input. For example, 20% of patients accounted for 82% of U.S. healthcare spending in 2013. This principle can help you identify low-hanging fruit, but note that after a certain point, additional effort will produce less impactful results, as per the law of diminishing returns. In some cases you may eventually overdo things to the point of being counterproductive, which takes you to the point of negative returns.
Unfortunately, even if you choose the right activities you can still fall into the trap of procrastination, which can be caused by present bias (overvaluing near-term rewards). At other times, your loss aversion makes you reluctant to abandon projects that are on the wrong path, because admitting a loss is painful. However, when your costs are irrecoverable, including them in your decision-making leads to the sunk-cost fallacy. Try to evaluate your projects from an objective point of view. You can also use the power of commitment to overcome present bias (e.g. having penalties for breaking commitments), and the default effect to make your desired actions the "path of least resistance" (or opt-in by default). Finally, look for shortcuts to success in the form of design patterns. These are reusable solutions, like how doorknobs are set at a certain height and staircases are always wide enough. Design patterns can also take the form of step-by-step algorithms. When facing a hard problem, consider ways to reframe the problem, transforming it into a version that is easier to solve.
IV.
The fourth chapter in Super Thinking borrows many concepts from physics and biology, especially ones that relate to change and adaptation. Of course, natural selection is how species (and arguably societies) evolve, and the experimental mindset of the scientific method can help us adapt even better to our environment. The concept of inertia refers to an object's resistance in changing speed and direction, but as a metaphor it can also be applied to beliefs (e.g. confirmation bias) and organizations (e.g. the strategy tax of being locked into a strategic position). A related model is momentum, a function of mass and velocity. For example, a flywheel stores energy as it gains momentum, and once it is spinning its inertia makes it easy to keep it spinning. In your personal and professional life you want to create or join healthy flywheels, and when facing resistance to change you should look for potential energy and a catalyst (something that lowers the activation energy required to trigger a chemical reaction). This can take the form of a forcing function, which is like a prescheduled event that facilitates a desired action (e.g. regular one-on-one meetings).
Major change often happens after a process reaches critical mass. For new ideas and technologies, the tipping point of adoption occurs when the growth curve inflects upward. This can be seen in the technology adoption life cycle, a model that describes how early adopters help push new things past a tipping point. Unfortunately, the economy is a chaotic system that cannot be accurately predicted; this means you can at best work to stay adaptable and increase the probability of a successful outcome. Your luck surface area refers to how wide a net you cast to catch positive opportunities -- cultivate it by putting yourself in more diverse and unfamiliar situations and meeting people. Note that this technically increases your entropy (a measure of disorder), so you want to avoid too much entropy by choosing which events are worth attending (see Chapter 3). When making choices, avoid the black-and-white fallacy of thinking you have only two options, or assuming that the situation is zero-sum (someone else gains at your expense). Look for win-win options.
V.
Given that one of the co-authors is a statistician, it is unsurprising that there is a whole chapter dedicated to numeracy. But statistics and probability are key to grappling with uncertainty. Rather than relying on anecdotal evidence from personal experience, you should seek data from a randomized controlled experiment or A/B testing, where the factor under consideration is isolated. These methodologies try to account for the fact that correlation does not imply causation -- although correlations may help with hypothesis generation. Even better, look for a systematic review or meta-analysis, which combines results from several relevant studies in an area. This is a useful starting point, because individual studies often fail to replicate (see: the replication crisis). Studies may suffer from hidden biases, such as selection bias when people are not randomly assigned to different groups, response bias when people don't answer surveys truthfully, or survivorship bias when a sample only includes the population that "survived" (e.g. when you survey current but not former employees).
Whether an experiment shows statistical significance relates to its likelihood of being a false positive (type I error, denoted α) or false negative (type II error, denoted β). Most commonly, researchers choose a false positive rate of 5%, which is why they call a result statistically significant when its p-value is less than 0.05 (in other words, there is less than 5 percent chance that the null hypothesis, if true, would produce a result as extreme). However, setting a lower alpha goes hand-in-hand with a higher beta, or falsely giving a negative result, thereby lowering the power of the experiment to detect a true effect. This has to do with the bell-curve shape of the normal distribution, where most values cluster around the mean (average) but some are dispersed due to underlying variance. See the following figures from the book:
Because the two bell curves overlap, some observations are consistent with both the null and alternative hypotheses. Luckily, you can reduce one of the error rates without increasing the other by increasing the sample size -- this will decrease the overlap between the bell curves by making the curves narrower. That's a consequence of the law of large numbers (which, by the way, does not make it more likely that a streak of ten heads from a coin flip will be followed by a tail than another head; that's the gambler's fallacy). And we're assuming that the differences between experimental participants are normally distributed in the first place due to the central limit theorem: when taking numbers from a probability distribution (e.g. Bernoulli distribution) and averaging them, the result tends to approximate a normal distribution. This is how frequentist statistics uses a large sample to estimate the population mean to within a certain confidence interval (ideally represented by error bars if there is a graph). By contrast, Bayesian statistics lets us make probabilistic judgments even if there are very few data points, by starting from a prior and updating it with new information. The prior can reflect knowledge about base rates, for example the probability that a random driver is drunk before you administer a Breathalyzer test. Without accounting for the base rate, you may underestimate the chance that the driver is sober given that the test says they're drunk, which is the base rate fallacy. Eliezer Yudkowsky is an avowed Bayesian, but the authors of Super Thinking argue that both approaches are valid when done right.
VI.
Chapter 6 is all about decision-making, and goes beyond the simple pro-con list. The go-to solution here is a cost-benefit analysis, which asks you to attach weights to each pro and con item (perhaps in terms of dollars) and lay them out over time so that you can apply a discount rate to account for future uncertainty and inflation (i.e. a benefit today is worth more than the same benefit later). However, the discount rate you choose can wildly influence the estimate of your net benefit; that's why it's a good idea to run a sensitivity analysis on variations in the discount rate and any other input parameters that are based on uncertain assumptions. In addition to cost-benefit analysis, you can also use a decision tree to analyze different possible outcomes of decision points and chance points. For each branch on the tree, you can attach a probability estimate plus a value (either dollar prices or utility values) and multiply them to calculate the expected value of each choice. When doing this analysis, beware of black swan events (consequential events that are rare but still happen more often than you might expect) and unknown unknowns (things you don't know you don't know).
There are two techniques that can help you uncover such risks. The first is systems thinking, a way to think about the overall system and how its parts interact, such as through causal loop diagrams or simulations. A Monte Carlo simulation, for example, is like a dynamic sensitivity analysis, whereby a system is simulated many times with random initial conditions to gauge the probability of different outcomes. Systems thinking can also help you identify the best outcome, known as the global optimum, so you don't get stuck chasing a local optimum solution. The second model is scenario analysis, which asks you to think up several plausible future scenarios (e.g. "What would happen if major event X transpired?"). Scenario planning requires lateral thinking ("thinking outside the box") and divergent thinking (the opposite of converging on one solution). When brainstorming in groups, seek diverse points of view in order to avoid groupthink (for example, you can try to crowdsource ideas or use a prediction market).
VII.
Next, Weinberg and McCann continue the theme of decisions, with more focus on situations of conflict. Some of the models here come from game theory, the study of decision making in adversarial "games". For example, the prisoner's dilemma is a situation where the dominant strategy for a selfish player is to always betray their partner in crime, even though the best payoff for both prisoners is reached when they cooperate on remaining silent. That's because of the Nash equilibrium: neither player can unilaterally improve their outcome by changing strategy. Only if the game is iterated or repeated is there room for reputation and retribution, allowing for cooperation via the tit-for-tat strategy. Many situations can be analyzed through a game-theory lens, including the arms race (escalating competition leading to wasteful spending), the ultimatum game (read more here), and the war of attrition (where the winner is the last to run out of resources). To get your desired outcome, you can try to influence the other players through Cialdini's six persuasion models: reciprocity (get them to return a favor), commitment (make them agree to something small first), liking (be likable), social proof (use their desire to fit in), scarcity (make an opportunity seem rare), and authority (appeal to authority figures). In addition, there are models that relate to framing, such as the perspectives of social norms versus market norms (i.e. communal favors vs. financial transactions), distributive justice versus procedural justice (fairness as equal distribution or fairness as transparent and objective procedures), and appeal to emotion. Be aware that these models may be used against you, for example in the form of dark patterns -- design patterns that get you to spend more money at the casino, or make it difficult to unsubscribe from a website.
In many situations it may be better to avoid direct conflict altogether; in the extreme case, stability can be reached through the threat of mutually assured destruction (MAD). But more productive alternatives include diplomacy, deterrence (e.g. the carrot-and-stick model), containment (e.g. quarantine), or even appeasement if necessary. You can try to change the game to your favor by using guerrilla warfare -- getting a lot of leverage out of your limited resources by focusing on unconventional, nimble tactics -- and punching above your weight by intentionally taking on a larger-than-expected role in the arena. In these kinds of situations, remember that generals always fight the last war, meaning that your opponent may be using outdated tactics and technologies simply because they worked in the past. Finally, know your best exit strategy: do you have a graceful way to leave the situation, or do you need to throw a Hail Mary pass and burn the boats? Consider the likely long-term outcomes.
VIII.
In a chapter relevant to leadership and human resources, the central question is how to unlock people's potential and create incredible, 10x teams (that aren't literally ten times as productive, but you get the idea). The first thing to keep in mind is that people are not interchangeable, so your people management should be adjusted to the characteristics, personality traits, strengths and goals of the individual -- this is called managing to the person. For example, there are introverts and extroverts, generalists and specialists, and foxes versus hedgehogs (thriving on complexity vs. one grand philosophy). Of course, you should put people in the right roles; this does not always mean promoting a good performer to a higher role, because such roles may involve different skills. The Peter principle says that "managers rise to the level of their incompetence". You want to counteract it by developing multiple career tracks and helping people learn and grow into new roles in which they can succeed. The roles and responsibilities within your organization should be clearly delineated using the model of the directly responsible individual (DRI), who is accountable for the success of each action item.
Help people reach their full potential by encouraging deliberate practice -- putting them in situations at the edge of their abilities and providing real-time feedback. Give feedback through weekly one-on-one standing meetings, wherein you use radical candor by directly challenging them while still showing personal care. Look for suitable learning opportunities with the consequence-conviction matrix, which categorizes things into high vs. low consequence and whether you have high vs. low confidence in your own opinion. For decisions that would have a low impact if they fail, delegate them to your team members, especially if you have low conviction in being right. You can further unlock people's potential by fighting psychological barriers like impostor syndrome (when people feel like frauds who are on the brink of failure) and the Dunning-Kruger effect (when people who know something about an area become overconfident in their ability). Find the balance between helping people recognize their skill gaps, and praising progress. Finally, ensure that your organizational culture is well-defined: what beliefs, norms, and behavioral patterns fit your vision? People will be more intrinsically motivated and loyal if you are winning hearts and minds. Show employees that you value their contributions, and put your boots on the ground by showing you are one of them.
IX.
Chapter 9 is about how to achieve a sustainable competitive advantage in business. Of course, you'll want to have a 10x team. But on top of that, you want to gain market power: the ability to raise prices without losing many customers. This entails that you be differentiated from the competition. Consider contrarian bets: what do you know that others don't? According to the consensus-contrarian matrix, the only way to get outsize returns on investment is to be right while being contrarian. Peter Thiel calls this a secret. In addition to having a secret, you must have the right timing, so ask yourself why now? To be really successful, your product should change the behavior of your customers -- you need to achieve product/market fit. A model that can help you with this is customer development, which refers to the practice of de-risking an idea by running experiments with customers using an MVP (see Chapter 1) and asking for feedback. As you run through the loops of feedback, you may find yourself navigating the idea maze, which contains many dead ends on the path to product/market fit and success. Here, you need to be like heat-seeking missiles that lock onto their target and adjust course if necessary. Look for a resonant frequency, like how the right sound waves can cause a wineglass to break (but in this context the sign of resonance is that the product is flying off the shelves and customers are actively demanding more). You should be able to see bright spots somewhere in what you're doing, otherwise you will likely need to pivot -- to do something else. For example, Twitter started as a podcasting network before changing strategic direction.
Once you have a powerful market position, you can defend it by building a moat. There are many types of moats, including protected intellectual property, specialized skills or business processes that take a long time to develop, exclusive access to relationships or resources, a strong and trusted brand, substantial control over a distribution channel, a team of uniquely qualified people, network effects (a model related to flywheels and critical mass, whereby each additional node in a network increases its value, and thereby also the switching costs), and a faster pace of innovation. Big moats create barriers to entry, but even the biggest moats don't last forever. Disruptive innovations can shake up an industry; a famous example is that of Kodak, which was a giant in analog photography but got disrupted by digital cameras. Kodak was complacent and didn't appreciate the saying that "only the paranoid survive". By the way, the digital camera manufacturers were also disrupted by smartphone cameras. So, you should monitor potential threats, especially when they seem to have a good chance of crossing the chasm from early adopter to early majority (recall the technology adoption life cycle); and be ready to pivot.
X.
Finally, the Conclusion has two more mental models: the idea of a cargo cult (when people superficially imitate what they see but don't really understand it) and the circle of competence (the areas where you can think effectively). The message is that you need to think carefully about whether and how a given mental model applies to a situation, get feedback from other people, and try writing about complex topics in order to clarify your thinking. This way, your circle of competence can expand over time.
***
As you can see, I've summarized quite many of the book's models, but there are still a bunch more -- including, but not limited to, Hanlon's razor, the Streisand effect, bike-shedding, the Shirky principle, observer-expectancy bias, Chatelier's principle, Potemkin village, the Pygmalion effect, and the OODA loop.
I would rate Super Thinking at 3 stars out of 5, because it is not bad, but it's not one of those really memorable books either. The main problem is that there is way too much content for a book this size, with the result that each idea is only superficially discussed. To illustrate my point, there are roughly 330 to 340 mental models in Super Thinking, spread across 318 pages of content -- which means that there is an average of (slightly more than) 1 new model introduced on every page. Now, there happens to be something called the "Feynman technique", which is a way of learning a concept by explaining it simply on a sheet of paper (you could even consider it a mental model). It's as if Weinberg and McCann tried that here, but also had to link the models together in a unifying narrative. Perhaps I'm being unfair. However, I'm not alone. Other reviewers on Goodreads said the following:
- User A: "Each mental model can be turned in to [sic] small books. But cramming 100s of them into a single book like this is just foolish."
- User B: "Yes, it certainly packs a lot of mental models, but is also pretty exhausting to read, and has to be digested in small bits, as you are introduced to a new idea on every page."
- User C: "The author has attempted to explain too much in very few pages, who actually leads to have very little added info on on [sic] many different things."
Personally, the new things I've learned from Super Thinking include the terms "unforced error", "technical debt", "forcing function", and "luck surface area" among others (although the underlying concept of luck surface area is something I encountered in Nassim Taleb's The Black Swan). I was also less familiar with Chatelier's principle, hysteresis, or Joy's law. But for the other concepts, I would have appreciated deeper explanations (as the basic definitions were already somewhat familiar), or perhaps further illustration by way of unique academic studies or surprising statistics. Because Super Thinking does not stand out in these areas, its strength lies primarily in being a sort of repository of advice, which you can check to make sure your decisions don't overlook key factors.
Ultimately, Super Thinking does not really change the way you think about the world in the way that the Sequences do, for example. However, it does provide an extensive list of important concepts that can be used as a springboard for further investigation -- a step on one's journey to becoming a better thinker. That's not too bad.
Comments
Post a Comment