A summarization of the main cognitive biases, relating to financial trading, combined with my own understanding.
Diminishing marginal utility: the first one is more valuable than the second and so on. For example, when drinking a cup of water, the first sip has higher utility than the last sip.
Corresponding to diminishing marginal utility, the happiness of gaining $200 is less than two times of happiness of gaining $100. The perception of gain is convex.
The same applies to pain. The pain of losing $100 two times is higher than losing $200 in one time.
The “gain/loss” is relative to the expectation (frame of reference). Different people have different expectations in different scenarios.
Expectation management is important. If the outcome is good but doesn’t meet the high expectation, it still causes disappointment. Vice versa.
The expectation can gradually change. People gradually get used to the new norm. This make people be able to endure bad environments, and not get satisfied after achievement.
Shifting baseline syndrome (boiling frog syndrome): If the reality keeps changing slowly, the expectation also tend to keep nudging, eventually move a lot without being noticed. This is also common in long-term psychological manipulation.
Relative deprivation: When people expect to have something that they don’t have, they think they lose that thing, although they don’t actually losing it.
Door-in-the-face effect: Firstly make a large request that will likely be rejected, then make a modest request. The firstly made large request changes expectation to make the subsequent modest request easier to accept.
譬如你说,这屋子太暗,须在这里开一个窗,大家一定不允许的。但如果你主张拆掉屋顶他们就来调和,愿意开窗了。
If you say, “This room is too dark; we need to open a window here,” everyone would definitely refuse. However, if you propose removing the roof, they would come to a compromise and agree to open a window.
— 鲁迅 LuXun
Protective pessimism: Being pessimistic can reduce risk of disappointment.
Be optimistic | Be pessimistic | |
---|---|---|
Result is good | Expected. Mild happiness. | Exceeds expectation. Large happiness. |
Result is bad | Large disappointment. | Expected. Mild disappointment. |
Procrastination is also related to protective pessimism. If you believe that the outcome will be bad, then reducing cost (time and efforts put into it) is “beneficial”.
In real life, some risks are hard to reverse or are irreversible, so avoiding risk is more important than gaining. In investment, losing 10% requires gaining 11.1% to recover, and losing 50% requires gaining 100% to recover. Keep staying in the game is important as it makes one exposed to future opportunities.
So, losses have a larger mental impact than gains of the same size. The pain of losing $100 is bigger than the happiness of gaining $100.
People don’t intuitively treat avoiding losing one dollar as the same as earning one dollar.
Note that loss aversion is not uniform in all scenarios. Some risk-takers may be not loss averse.
We prefer deterministic gain instead of risky gain. A bird in the hand is worth two in the bush.
Given 100% chance to gain $450 or 50% chance to gain $1000, people tend to choose the former.
The professions that face uncertain gain, like academic research, where it’s common that researching a problem for years without getting any meaningful result, are not suitable for most people.
We prefer having hope rather than accepting failure.
Given 100% chance to lose $500 or 50% chance to lose $1100, most people will choose the latter. The second one has “hope” and the first one means accepting failure.
In this case, “no losing” is usually taken as expectation. What if the expectation is “already losing $500”? Then the two choices become: 1. no change 2. 50% gain $500 and 50% lose $600. In this case, people tend to choose the first choice which has lower risk. The expectation point is very important.
Telescoping effect:
Vierordt’s law: Shorter time intervals tend to be overestimated. Longer time intervals tend to be underestimated.
Oddball effect: The time that have novel and unexpected experience feels longer.
It can be seen that we feel time length via the amount of memory. Novel and unexpected experiences correspond to more memory. Forgetting “compresses” time. When people become older, novel experiences become more rare, thus time feels faster.
The memory of feeling risk has higher “weight” (risk aversion), so time feels slower when feeling risk. In contrast, happy time feels going faster.
Reference: Time perception - Wikipedia
Hedonic treadmill: after some time of happiness, the expectation goes up and happiness reduces. The things that people gained will gradually be taken for granted, and they always pursue for more.
Do not spoil what you have by desiring what you have not; remember that what you now have was once among the things you only hoped for.
― Epicurus
If happiness can be predicted, some happiness moves earlier. For example, one is originally happy when eating delicious chocolate. Then one become happy just after buying chocolate before eating it, and the happiness of actually eating chocolate reduces. In future the happiness can move earlier into deciding to buy chocolate. This effect is also called second-order conditioning.
Material consumption can give short-term satisfaction, but cannot give long-term well-being (paradox of materialism). Long-term well being can better be achieved by:
Means-end inversion: one originally want money (means) to improve life quality (end). However, the process of making money can sacrifice life quality. Examples: investing all money and leave little for consumption, or choosing a high-paying job with no work-life balance (golden handcuffs).
We already walked too far, down to we had forgotten why embarked.
People tend to maintain their ego by self-serving bias:
People tend to be overconfident about themselves:
The overconfidence is useful:
When looking at past, people find past events (including Black Swan events) reasonable and predictable, although they didn’t predicted these events in prior.
In a complex world, one event can have two contradicting interpretations.
One recent example: (This is written in Janurary 2025) DeepSeek V3 released, the report says that DeepSeek V3’s training cost is $5.5M, significantly lower than other large model training. Subsequent open model DeepSeek R1 has strong performance. And DeepSeek’s API price is significantly lower than OpenAI and Anthropic’s, creating a price war of AI. There are two different interpretations:
Both of the two interpretations makes sense. But only one is correct afterwards (could also be that the first is correct in short term and the second is correct in long term). The incorrect one will be ignored when analyzing history.
People make execuses about their prediction failure, such as:
People tend to justify previous behavior, even if these behaviors was made randomly, or made by external factors that does not exist now.
Self justitication shows self-control and consistency, facilitating social collaboration.
This is related to Stockholm Syndrome. After experiencing pain in the past, people tend to justify their previous pain.
Ben Franklin effect: People like someone more after doing a favor for them.
Endowment effect: We value more on the things that we own. Investors tend to be biased to positive information of the stock they own.
Foot-in-the-door effect: One agreed on a small request tend to subsequently agree on a larger request.
Saying becomes believing: Telling a lie too many times may make one truly believe in it.
People want to have a self-image of high capability. But a failure can debunk their self-image. Self-handicapping is one way of protecting the self-image. It’s an extension of protective pessimism.
Try hard | Self-handicap | |
---|---|---|
Get good result | Shows a sign of common capability. | Shows a sign of great capability. |
Get bad result | Shows a sign of low capability. | Can blame failure to self-handicapping. |
An example of self-handicapping: playing videogames before exam instead of learning.
Self-handicapping can avoid questioning the capability from self. However, other people usually just judge from result, and see self-handicapping as low capability.
Other examples of self-handicapping:
Self-handicapping is also a way of reducing responsibility. This is common in large corporations and governments: intentionally create reasons of failure to reduce responsibility.
People tend to fight the things that oppose their desire. Examples:
Providing external reward may reduce internal motivation (overjustification effect).
Confirmation bias: People tend to seek and accept the evidences that confirm their beliefs, and reluctant to accept contradictory evidences.
Motivated reasoning: when they does not want to accept contradictory evidences, they may make up and believe low-probability assumptions to explain the evidence in a way that follows the original belief.
Example: In a failing relationship, a person might attribute their partner’s distant behavior to work stress or other external factors, rather than accepting the possibility that their partner’s feelings have changed.
With confirimation bias, more information increases confidence, but doesn’t lead to better understanding.
If you don’t have an opinion, resist the pressure to have one.
— N. N. Taleb, Link
Information cocoon (echo chamber): People tend to actively choose to digest the information source that they like, and make friends with the one having similar beliefs.
Another thing I think should be avoided is extremely intense ideology, because it cabbages up one’s mind. …
I have what I call an iron prescription that helps me keep sane when I naturally drift toward preferring one ideology over another. And that is I say “I’m not entitled to have an opinion on this subject unless I can state the arguments against my position better than the people do who are supporting it. I think that only when I reach that stage am I qualified to speak.”
— Charlie Munger
Belief bias: if the conclusion confirms people’s existing belief, then people tend to believe it, regardless of the reasoning correctness, vice versa.
Bullshit asymmetry principle: Refuting misinformation is much harder than producing misinformation. To produce a misinformation and make it spread, you just need to make it follow people’s existing beliefs. But to refute a misinformation, you need to find sound evidences. This is also reversal of the burden of proof.
The good side of stubborness is to maintain diversity of ideas in a society, helping innovation and overcoming of unknown risks.
People tend to justify the groups they belong (group justification), and justify the society that they are in (system justification).
Examples:
People love to correct others and persuade others. Some ideas are memes that drive people to spread the idea. However, due to belief stability, it’s hard to persuade others, and people dislike being persuaded. This effect is common on internet social media.
The trouble with having an open mind, of course, is that people will insist on coming along and trying to put things in it.
— Terry Pratchett
Commitment can be a good thing. A lot of goals require continual time, efforts and resources to achieve.
However, there are investments that turn out to be bad and should be given up to avoid futher loss. All the previous investments become sunk cost. People are reluctant to give up because they have already invested a lot in them. Doing stop-loss signals failure. We want to have hope rather than accepting failure.
Examples:
Opportunity cost: if you allocate resource (time, money) to one thing, that resource cannot be used in other things that may be better. Opportunity cost is not obvious.
The difference between “good persistence” and “bad obstinacy”:
The persistent are like boats whose engines can’t be throttled back. The obstinate are like boats whose rudders can’t be turned. …
The persistent are much more attached to points high in the decision tree than to minor ones lower down, while the obstinate spray “don’t give up” indiscriminately over the whole tree.
— Paul Graham, The Right Kind of Stubborn
An environment that doesn’t tolerant failure makes people not correct mistakes and be obstinate on the wrong path (especially in authoritarian environments, where loyalty and execution attitude override honesty).
When you’re in the midst of building a product, you will often randomly stumble across an insight that completely invalidates your original thesis. In many cases, there will be no solution. And now you’re forced to pivot or start over completely.
If you’ve only worked at a big company, you will be instinctually compelled to keep going because of how pivoting would reflect on stakeholders. This behavior is essentially ingrained in your subconscious—from years of constantly worrying about how things could jeopardize your performance review, and effectively your compensation.
This is why so many dud products at BigCos will survive with anemic adoption.
Instead, it’s important to build an almost academic culture of intellectual honesty—so that being wrong is met with a quick (and stoic) acceptance by everyone.
There is nothing worse than a team that continues to chase a mirage.
— Nikita Bier, Link
Ignoring negative information or warning signs to avoid psychological discomfort.
Examples:
We can learn from the world in an information-efficient way: learning from very few information quickly. (On the other side, current deep learning technology is information-inefficient, as it requires tons of training data to get good results.)
The flip side of information-efficient learning is hasty generalization. We tend to generalize from very few examples quickly, rather than using logical reasoning and statistical evidence, thus easily get fooled by randomness.
The reality is complex, so we need to simplify things to make them easier to understand and easier to remember. However, the simplification can get wrong.
To simplify, we tend to make up reasons of why things happen. A reasonable thing is simpler and easier to memorize than raw complex facts.
When statistical analysis shows that A correlates with B, the possible causes are:
Examples of correlation of A and B are actually driven by another factor C:
Among my favorite examples of misunderstood fitness markers is a friend of a friend who had heard that grip strength was correlated with health. He bought one of this grip squeeze things, and went crazy with it, eventually developing tendonitis.
— Paul Kedrosky, Link
People tend to believe more from stories, anecdotes or individual examples, even if these examples are made up or are just statistical outlier. On the contrary, people are less likely to believe in abstract statistical evidences.
Examples:
Narrative fallacy is introduced in The Black Swan:
We like stories, we like to summarize, and we like to simplify, i.e., to reduce the dimension of matters.
……
The fallacy is associated with our vulnerability to overinterpretation and our predilection for compact stories over raw truths. It severely distorts our mental representation of the world; it is particularly acute when it comes to the rare events.
— The Black Swan
Narrative fallacy includes:
People tend to judge things by first impression. This makes people generate belief by only one observation, which is information-efficient, but can also be biased.
People like to judge a decision by its immediate result. However, the real world is full of randomness. A good plan may yield bad result and a bad plan may yield good result. And the short-term result can differ to long-term result.
There is no perfect strategy that will guarantee success. Overemphasizing short-term outcomes leads to abandoning good strategies prematurely.
The quicker the feedback gives, the quicker people can learn (this also applies to reinforcement learning AI). But if the feedback delays 6 months, it’s hard to learn from it, and people may do wrong hasty generalization using random coincidents, before the real feedback comes, thus get fooled by randomness.
Reducing cost by removing safety measures usually does not cause any visible accidents in the short run, but the benefit of reduced costs are immediately visible. When the accident actually happened because of the removed safety measures, it may be years later.
The solution is to take notes and analyze past notes.
People crave quick feedback. Successful video games and gambling mechanisms utilize this by providing immediate responses to actions.
What’s more, for most people, concrete visual and audio feedback is more appealing than abstract feedback (feedback of working with words and math symbols).
The previously mentioned reverse psychology is also related to learning. Being forced to learn make one dislike learning it. Self-directed learning make one focus on what they are interested in, thus is more effective.
To summarize, most people naturally prefer the learning that:
It’s also hard to learn if the effect of decision is applied to other people, especially for decision-makers:
It is so easy to be wrong — and to persist in being wrong — when the costs of being wrong are paid by others.
— Thomas Sowell
People tend to simplify causal relationship and ignore complex nuance. If X is a factor that causes Y, then people tend to treat X as the only reason that caused Y, over-simplifying causal relationship.
Usually, the superficial effect is seen as the reason, instead of the underlying root cause.
Examples of causal oversimplification:
Oversimplification: “Poor people are poor because they are lazy.”
Nuanced view: Poverty is also influenced by factors such as education access, systemic discrimination, health disparities, job market conditions, and the Matthew effect.
Oversimplification: “Immigrants are the cause of unemployment.”
Nuanced view: Unemployment rates are also influenced by factors including manufacturing relocation, automation technologies, economic cycles, skill mismatches, and overall labor market conditions.
Oversimplification: “The Great Depression happened because of the stock market crash of 1929.”
Nuanced view: The Great Depression resulted from a combination of factors, including excessive financial risk-taking, lack of regulatory oversight, production overcapacity, wealth inequality, and international economic imbalances.
Oversimplification: “That company succeeded because of the CEO.”
Nuanced view: Corporate success also depends on various factors, including market conditions, employee contributions, incentive structures, company culture, government-business relationships, competitive landscape, and the cumulative impact of past leadership decisions.
We tend to simplify things. One way of simplification is to ignore the grey-zone and complex nuance, reducing things into two simple extremes.
Examples of binary thinking:
People’s evaluations are anchored on the expectation, and not meeting an exception could make people’s belief turn to another extreme.
Internet has indeed changed the world. But the dot com bubble burst. It’s just that the power of Internet required time to unleash, and people placed too much expectation in it too early.
Neglect of probability: either neglect a risk entirely or overreact to the risk.
Halo effect: Liking one aspect of a thing cause liking all aspects of that thing and its related things. (爱屋及乌)
Examples:
Horn effect is the inverse of halo effect: if people dislike one aspect of a thing, they tend to dislike all aspects of that thing and its related things. People tend to judge words by the political stance of the person who said it. Disaggrement on ideas tend to become insults to people, especially in a competitive environment.
Halo effect and horn effect are related to binary thinking.
People prefer definite answer, over ambiguity or uncertainty (such as “I don’t know”, “it depends on exact case”, “need more investigation”), even if the answer is inaccurate or made up.
This is related to narrative fallacy: people like to make up reasons explaining why things happen.
One day in December 2003, when Saddam Hussein was captured, Bloomberg News flashed the following headine at 13:01: U.S. TREASUERIES RISE; HUSSEIN CAPTURE MAY NOT CURB TERROISM. ……
As these U.S. Treasury bonds fell in price (they fluctuate all day long, so there was nothing special about that) …… they issued the next bulletin: U.S. TREASURIES FALL; HUSSEIN CAPTURE BOOSTS ALLURE OF RISKY ASSETS.
— The Black Swan
People dislike uncertain future and keep predicting the future, while ignoring their terrible past prediction record (hindsight bias).
People like to wrongly apply a theory to real world, because applying the theory can give results. Example: assuming that an unknown distribution is gaussian even when it’s not.
Zeigarnik effect: People focus on uncompleted things more than completed things. When some desire is not fulfilled (gambling not winning, PvP game not winning, browsing social media not seeing wanted content, etc.), the desire becomes more significant.
Need for closure is also related to curiosity.
People may idealize the things that they are not familiar with:
People tend to idealize the distant past and forget the past misery. This helps people get out of trauma, and at the same time idealize the past things:
People may think that they deeply understand something, until writing it down. When writing it down, the “gaps” of the idea will be revealed.
Pure thinking is usually vague and incomplete, but people overestimate the rationality of their pure thinking.
The reason I’ve spent so long establishing this rather obvious point [that writing helps you refine your thinking] is that it leads to another that many people will find shocking. If writing down your ideas always makes them more precise and more complete, then no one who hasn’t written about a topic has fully formed ideas about it. And someone who never writes has no fully formed ideas about anything nontrivial.
It feels to them as if they do, especially if they’re not in the habit of critically examining their own thinking. Ideas can feel complete. It’s only when you try to put them into words that you discover they’re not. So if you never subject your ideas to that test, you’ll not only never have fully formed ideas, but also never realize it.
— Paul Graham, Link
Even so, writing the idea down may be still not enough, because natural language is vague, and vagueness can hide practical details. The issues hidden by the vagueness in language will be revealed in real practice (e.g. turning software requirement into code).
Having ideas is easy and cheap. If you search the internet carefully you are likely to find ideas similar to yours. The important is to validate and execute the idea.
According to predictive processing theory, the brain predicts (hallucinates) the most parts of perception (what you see, hear, touch, etc.). The sensory signals just correct that prediction (hallucination).
Body transfer illusion (fake hand experiment)
Survivorship bias means that only consider “survived”, observed samples and does not consider “silent”, “dead”, unobserved samples, neglecting the selection mechanism of samples.
A popular image of survivorship bias:
The planes that get hit in critical places never come back, thus don’t get included in the stat of bullet holes, forming the regions missing bullet hole in that image.
Other examples of survivorship bias:
A more generalized version of survivor bias is selection bias: When the sampling is not uniform enough and contains selection mechanism (not necessary sample vanishing), there will be bias in the result.
Some physicists propose Anthropic Principle: the physical laws allow life because the existence of life “selects” the physical law. The speciality of the physical laws come from survivorship bias.
Availability bias: When thinking, the immediate examples that come into mind plays a big role.
Example: If you recently saw a car crash, you tend to think that traveling by car is riskier than traveling by plane. However, if you recently watched a movie about a plane crash, you might feel that planes are more dangerous.
Vividness bias: People tend to believe more from vivid things and stories, over abstract statistical evidences. This is related to anecdotal fallacy and narrative fallacy.
The Italian Toddler: In the late 1970s, a toddler fell into a well in Italy. The rescue team could not pull him out of the hole and the child stayed at the bottom of the well, helplessly crying. …… the whole of Italy was concerned with his fate …… The child’s cries produced acute pains of guilt in the powerless rescuers and reporters. His pictures was prominently displayed on magazines and newspapers …..
Meanwhile, the civil war was raging in Lebanon …… Five miles away, people were dying from the war, citizens where threatened with car bombs, but the fate of the Italian child ranked high among the interests of the population in the Christian quarter of Beirut.
— The Black Swan
Enforcing safety measures is usually unappreciated. Because people only see the visible cost and friction caused by safety measures (concrete), and do not see the consequences of not applying safety measures in a parallel universe (abstract), until an incident really happens (concrete).
People are more likely to pay terrorism insurance than for plain insurance that covers terrorism and other things.
If people are given some choices, people tend to choose one of the provided choices and ignore the fact that other choices exist. This is also framing effect.
People tend to attribute one product to one public figure, or attribute a company to its CEO, because that’s the name that they know, and because of causal simplification tendency.
25 years ago today, SYSTEM SHOCK 2 was first released, designed by @levine at Looking Glass.
— The Game Awards, Link
“There’s a tendency among the press to attribute the creation of a game to a single person,” says Warren Spector, creator of Thief and Deus Ex.
People prefer familiar things. One reason is the availability bias. Another reason is that people self-justifys their previous attention and dedication. This is highly related to availability bias.
When making decisions, people tend to focus on what they already know, and ignore the aspects that they do not know or are not familiar with. We have already considered what we already know, so we should focus on what we don’t know in decision making.
This is related to risk compensation: People tend to take more risk in familiar situations.
Imprinting: At young age, people are more likely to embrace new things. At older age, people are more likely to prefer familiar things and avoid taking risk in unfamiliar things. (Baby duck syndrome).
- Anything that is in the world when you’re born is normal and ordinary and is just a natural part of the way the world works.
- Anything that’s invented between when you’re 15 and 35 is new and exciting and revolutionary and you can probably get a career in it.
- Anything invented after you’re 35 is against the natural order of things.
― Douglas Adams
Noticing something more frequently after learning about it, leading to overestimating its prevalence or importance.
Sometimes, one talked about something then sees its ad in social media, thus suspecting that their phone and social media app is recording voice for ad recommendation. Of course that possibility exists, but perception of that possibility is exaggerated by frequency illusion.
People tend to judge things by comparing it with examples (stereotypes) that come into mind, and tend to think that one sample is representative to the whole group.
Representative bias can sometimes be misleading:
Say you had the choice between two surgeons of similar rank in the same department in some hospital. The first is highly refined in appearance; he wears silver-rimmed glasses, has a thin build, delicate hands, measured speech, and elegant gestures. …
The second one looks like a butcher; he is overweight, with large hands, uncouth speech, and an unkempt appearance. His shirt is dangling from the back. …
Now if I had to pick, I would overcome my sucker-proneness and take the butcher any minute. Even more: I would seek the butcher as a third option if my choice was between two doctors who looked like doctors. Why? Simply the one who doesn’t look the part, conditional on having made a (sort of) successful career in his profession, had to have much to overcome in terms of perception. And if we are lucky enough to have people who do not look the part, it is thanks to the presence of some skin in the game, the contact with reality that filters out incompetence, as reality is blind to looks.
— Skin in the game
Note that the above quote just talks about a hypothetical scenario and should not be simplified to tell that the unprofessional-looking ones are always better.
When an event has occured frequently, people tend to believe that it will occur less frequently in the future.
Examples:
One related topic is the law of large numbers: if there are enough samples of a random event, the average of the results will converge. The law of large numbers focus on the total average, and does not consider exact order.
The law of large number works by diluting unevenness rather than correcting unevenness. For example, a fair coin toss will converge to 1/2 heads and 1/2 tails. Even if the past events contain 90% heads and 10% tails, this does not mean that the future events will contain more tails to “correct” past unevenness. The large amount of future samples will dilute the finite amount of uneven past samples, eventually reaching to 50% heads.
Actually, gambler’s fallacy can be correct in a system with negative feedback loop, where the short-term distribution changes by past samples. These long-term feedback loops are common in nature, such as the predator-prey amount relation. It also appears in markets with cycles. (Note that in financial markets, some cycles are much longer than expected, forming trends.) In a PvP game with Elo-score-based matching mechanism, losing makes make you more likely to win in the short term.
One related concept is regression to the mean, meaning that, if one sample is higher than average, the next sample is likely to be lower than the last sample, and vice versa. Example: if a student’s score follows normal distribution with average 80, when that student gets 90 scores, they will likely to get a score worse than 90 in the next exam.
The difference between gambler’s fallacy and regression to the mean:
Regression fallacy: after doing something and regression to the mean happens, people tend to think what they do caused the effect (hasty generalization). Example: the kid gets a bad score; parent criticizes; the kid then get a better score. It’s seen that criticizing makes the score get better, although this is just regression to the mean that can happen naturally.
People tend to think that more specific and reasonable cases are more likely than abstract and general cases.
Consider two scenarios:
Although B is more specific to A, thus have a lower probability than A, people tend to think B is more likely than A. B implies a causal relationship, thus look more reasonable.
People tend to think that a story with more details is more plausible, and treat probability as plausibility. A story with more details is not necessarily more plausible, as the details can be made up.
Making a story more reasonable allows better information compression, thus making it easier to remember and recall.
People often assume that others know what they know. So people often omit important details when explaining things, causing problems in communication and teaching.
When learning a new domain of knowledge, it’s beneficial to ask “stupid questions”. These “stupid questions” are actually fundamental questions, but seen as stupid by experts, who already forgot the thinking when not knowing the fundamentals, under curse of knowledge. One benefit of AI is that you can ask “stupid questions” without being humiliated (but be wary of hallucinations).
Simplicity is often confused by familiarity. If one is very familiar with a complex thing, they tend to think that thing is simple.
Normalcy bias: Thinking that past trend will always continue. This is partially due to confirmation bias.
Although the market has trends, and a trend may be much longer than expected, no trend continues forever. Anything that is physically constrained cannot grow forever. Even with Matthew effect, the richest one’s wealth cannot grow exponentially forever, as the wealth is physically constrainted.
Most people are late-trend-following in investment: not believing in a trend in the beginning, then firmly believing in the trend in its late stage. This is dangerous, because the market has cycles, and some macro-scale cycles can span years or even decades. The experiences gained in the surge part of the cycle are harmful in the decline part of the cycle.
Overemphasizing recent events and performance, while ignoring long-term trends, causing overreaction.
People tend to
This is related to Amara’s law: we tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.
This is also related to availability bias, where the just-seen events are more obvious and easier to recall than old events and non-obvious underlying trends.
In financial markets, even when a market is in a trend, it still has fluctuations that temporarily deviate from the trend. Overly focusing on recent fluctuations may cause misunderstanding of the trend. Of course, the temporal fluctuation can also be an indication that the trend stops.
Normalcy bias means underreact to new events, but recency bias means overreact to new events, which is the opposite of normalcy bias. These two are actually not conflicting. Which one takes effect initially is related to actual situation and existing beliefs (confirmation bias). When one person does not believe in a trend but the trend continued for a long time, binary thinking may make that person turn their belief 180 degrees and deeply believe in the trend (but it’s usually too late).
People tend to make decisions based on how information is presented (framed) rather than objective facts.
There are many ways to frame one fact.
The content creator could emphasize one aspect and downplay another aspect, and use different wording or art style to convey different opinions. The people reading the information could be easily influenced by the framing subconsciously.
The name of a thing affects how people perceive it. Examples:
A loaded question is a question that contains an assumption (framing). Following that assumption can lead to a biased answer. Example: “Do you support the attempt by the US to bring freedom and democracy to other places in the world?”
Asking the right question requires the right assumption.
Mehrabian’s rule: When communicating attitudes and feelings, the impact is 7% verbal (words), 38% vocal (tone of voice), 55% non-verbal (facial expressions, gestures, posture). Note that this doesn’t apply to all kinds of communications.
Just looking confident can often make other people believe. This even applies when the talker is AI:
A friend sent me MRI brain scan results and I put it through Claude. No other AI would provide a diagnosis, Claude did. Claude found an aggressive tumour. The radiologist report came back clean. I annoyed the radiologists until they re-checked. They did so with 3 radiologists and their own AI. Came back clean, so looks like Claude was wrong. But looks how convincing Claude sounds! We’re still early…
— Misha Saul, Link
Anchoring bias: People’s judgement may be influenced by reference “anchors”, even if the reference anchor is irrelevant to decision making. Anchoring is a kind of framing. A salesman may firstly show customers an expensive product, then show cheap products, making customers feel the product being cheaper, utilizing anchoring bias.
The Anchoring Bias and its Effect on Judges.
Lie by omission: A person can tell a lot of truth while omitting the important facts, stressing unimportant facts (wrong framing), intentially causing misunderstanding, but at the same time be not lying in literal sense.
The price chart is often drawn by making lowest price at the bottom and highest price at the top. The price range of the chart is also framing. If one stock already have fallen by 30%, the latest price is in the bottom of the chart, so the stock seems cheap, but it may actually be not cheap at all. Reverse also applies: a stock grown by 30% makes its latest price shown on the top but it may be not expensive at all.
“Shooting the messenger” means blaming the one who bring the bad news, even though the messenger has no responsibility of causing the bad news.
The same effect happens in other forms:
Imagine someone who keeps adding sand to a sand pile without any visible consequence, until suddenly the entire pile crumbles. It would be foolish to blame the collapse on the last grain of sand rather than the structure of the pile, but that is what people do consistently, and that is the policy error. …
As with a crumbling sand pile, it would be foolish to attribute the collapse of a fragile bridge to the last truck that crossed it, and even more foolish to try to predict in advance which truck might bring it down. …
Obama’s mistake illustrates the illusion of local causal chains—that is, confusing catalysts for causes and assuming that one can know which catalyst will produce which effect.
— The Black Swan of Cairo; How Suppressing Volatility Makes the World Less Predictable and More Dangerous
People tend to value scarce things even they are not actually valuable and undervalue good things that are abundant.
Examples:
People tend to value something only after losing it.
The correlation of overall samples may be contradictory to the correlation inside each sub-groups.
Examples:
Base rate fallacy: there are more vaccinated COVID-19 patients than un-vaccinated COVID-19 patients in hospital, but that doesn’t mean vaccine is bad:
When one person is in a small group with similar opinions, they tend to think that the general population have the similar opinions. When they encounter a person that disagrees with them, they tend to think the disagreer are minority or are defective in some way.
This effect is exacerbated by algorithmic recommendation of social medias.
We also tend to think other people are similar to us in some ways. We learn from very few examples, and that few examples include ourselves.
We don’t see things as they are. We see things as we are.
We use relations to efficiently query information in memory. The brain is good at looking up relations, in an automatically, unintentionally and subconscious way.
Being exposed to information makes human recognize similar concepts quicker. Examples:
Being exposed to information also changes behavior and attitudes. Examples:
The main moral of priming research is that our thoughts and our behavior are influenced, much more than we know or want, by the environment of the moment.
— Think, fast and slow
Note that the famous “age priming” effect (walk more slowly after reminding aging concepts) failed to be replicated.
The placebo effect is also possibly related with priming.
Spontaneous trait transfer: listeners tend to associate what the talker say to the talker, even when talker is talking about another person:
When making decisions, human tend to follow intuitions, which is quick and energy-efficient, but also less accurate.
Thinking, Fast and Slow proposes that human mind has two systems: System 1 thinks by intuition and heuristics, which is fast and efficient, but inaccurate and biased; System 2: thinks by rational logical reasoning, which is slower and requires more efforts, but is more accurate. Most thinking mainly uses System 1 while being unnoticed.
People tend to choose the default and easiest choice. Partially due to laziness, partially due to fear of unknown risk.
In software product design, the default options in software plays a big role in how user will use and feel about the software. Increasing the cost of some behavior greatly reduces the people doing that behavior:
Status quo bias: tend to maintain status quo. This is related to risk aversion, as change may cause risk.
A related concept is omission bias: People treats the harm of doing something (commision) higher than the harm of not doing anything (omission). Doing things actively bears more responsibility. In the trolley problem, not doing anything reduces perceived responsibility.
Action bias: In the places where doing action is normal, people prefer to do something instead of doing nothing, even when doing action has no effect or negative effects.
When being judged by other people, people tend to do action to show their value, productivity and impression of control:
It’s being empirically observed that, in personal financial trading, higher trading frequency is correlated with worse profit.
Action bias does not contradict with default effect. When one is asked to work and show value, doing action is the default behavior, and not doing action is more risky, as people tend to question the one that does not look like working.
It’s not the things you buy and sell that make you money; it’s the things you hold.
— Howard Marks
People like easy solutions and shortcuts. However, some seemingly easy solutions does not address the root cause, having negligible effect or negative effect in the long run. Applying the easy solution gives the fake impression that the problem is being addressed, thus achieving mental comfort.
Examples:
This is related to means-end inversion. To achieve the root goal (end) we work on a sub-goal (means) that helps root goal. But overly focusing on the sub-goal hurts the root goal.
A similar phenomenon occurs commonly in medicine: treatments usually mainly suppress visible symptoms instead of curing the root cause of illness. This is usually due to technological and budget limitations.
Path dependence means sticking to what worked in the past and avoid changing, even when the paradigm has shifted and the past successful decisions are no longer appropriate.
Law of the instrument: “If the only tool you have is a hammer, it is tempting to treat everything as if it were a nail.”
One easy way to make decisions is to simply follow the people around us. This is beneficial in ancient world: for example, if a tiger comes and some people start fleeing, following them is better than spending time recognizing the tiger.
This is related to social proof heuristic: assuming that surrounding people know the situation better.
Following the crowd is also a great way of reducing responsibility: when everyone is guilty, the law cannot punish everyone. The one that acts independently bears more responsibility.
When many people follow each other, they will confirm each other, creating self-reinforcing momentum. This is also a reason of the momentum in markets. People tend to be overconfident when people around them are confident, and vice versa.
The price can grow not because people believe in the asset, but because the amount of people and money believing in the asset increases. Once that amount stop increasing, price tend to drop despite people believing. Profit can come from the expansion of believers. In this case, the asset is a Veblen good, where higher price induce more demand, until the limit is reached.
Measuring people’s belief by observing the people around you is inaccurate, because the people near you don’t represent all people (representative bias).
Herd mentality is in some sense a kind of trend following strategy. If the trend is some new good technology then following is good regardless of early or late. However, for speculative financial assets, the price grow depends on new people and money entering, so most people will start following too late and cannot profit from it.
Herd mentality is sometimes also called bandwagon effect.
One similar effect, in-group bias: Favoring investments or opinions from people within one’s own group or those who share similar characteristics.
From the cultrual standpoint, herd mentality also facilitates collaboration and society stability.
Mental accounting: Treating different parts of money differently, based on their source or intended use.
For example, one can separate the budgets for entertainment, housing and food. It’s a simple huristic that can avoid excessive spending: if each part doesn’t overspend, then they won’t overspend overall.
Mental accounting is related to sunk cost and loss aversion. If one sub-account is low, people tend to be more saving in that sub-account, making loss aversion more significant, and the previous waste in that sub-account become sunk cost.
In investment, mental accounting can happen on different forms:
Seperate by different time intervals. Setting profit target in each time interval (by month, season or year) can be detrimental in a market with momentum. If the profit in the time interval is meet, stop investing misses large profit from trend. If the profit in the time interval is not meet near the end, then the trader tend to be more nervous and more aggressive, which is dangerous.
However, setting stop-loss in each time interval has be good effects. When the current trading strategy does not fit the market, temporarily stopping could get through the current part of cycle that temporarily doesn’t suit the strategy. The stopping period also help calming and becoming rational.
Separate by different specific assets (e.g. stocks). If the mental accounts are separated based on different stocks, after losing from one stock, one may insist to gain the loss back from the same stock, even if investing in other stocks is better overall.
Separate by different categories of assets. People tend to prefer investing medium risk asset using all money over investing high-risk asset using partial money (barbell strategy), even when the total volatility and expected return are the same, because the invested money is in a different mental account than not-invested money, and risk aversion.
Lipstick effect is related to mental accounting. When the income declines, the mental account of luxury spending still exists, just shrunk, so cheaper lipsticks get more sales.
Mental accounting is one kind of narrow framing bias:
Narrow framing bias: focusing too much on one aspect while neglecting other aspects.
Zero-risk bias: preferring to eliminate one type of risk entirely rather than reducing overall risk (usually at the expense of increasing exposure to other risks).
It’s related to risk aversion and availability bias: risk aversion usually focuses the risk that comes to mind, and ignores other kinds of risks. It’s also related to binary thinking: thinking that an risk is either completely eliminated or not being taken any action on. The rational decision is to do a tradeoff to reduce overall risk.
Examples:
It should not be simplified to “avoiding risk is bad”. The point is to not do extreme tradeoffs to eliminte one kind of risk but increase exposure to other kinds of risks.
People tend to avoid regret. Regret aversion has two aspects:
The world is full of randomness. There is no decision that guarantees to be optimal. We should accept that we cannot always make perfect decisions. Validating the strategy in the long run is more important than result in of individual decisions.
We tend to regret doing something in short term, but regret not doing something in the long term. Reference.
‘I have led a toothless life’, he thought. ‘A toothless life. I have never bitten into anything. I was waiting. I was reserving myself for later on-and I have just noticed that my teeth have gone. …’
— Jean-Paul Sartre
Having safety measures make people feel safer and take more risks.
For example, drivers may drive faster when safety belt is on, and cyclists may ride faster when wearing helmet.
People tend to be overconfident in familiar situations, but that’s where accidents likely to occur:
Most accidents (69%) occurred on slopes that were very familiar to the victims. Fewer accidents occurred on slopes that were somewhat familiar (13%) and unfamiliar (18%) to the victim.
— Evidence of heuristic traps in recreational avalanche accidents
“Fight or flight” are the two options for dealing with physical threat (e.g. a tiger) in the ancient world.
But in the modern world, there are non-physical threats and modern risks (e.g. exam failure, losing job). These modern threats can be dealt with neither concrete fight or flight. So they may cause depression, anxiety and immobilization.
The rational activities (System 2 activities) require mental energy (willpower):
If there is no enough mental energy, one is less likely to resist impulse behaviors or think about hard problems, and possibly have difficulty in social interactions.
These factors affect mental energy:
Mental resting is different to body resting. Intense thinking when lying on the bed even consumes mental energy. Mental resting involves focusing on simple things with low cognitive demand. Different people may have different preferences. These mental resting activities may include listening to musics, watching simple videos, reading simple books, casual drawing and writing, playing games, meditation, etc.
Before you try to increase your willpower, try to decrease the friction in your environment.
— James Clear, Link
In the process of self-justification, people’s memory may be distorted. Human memory is actually very unreliable. People usually cannot notice that their memory has been distorted, and insist that their memory is correct.
People tend to simplify their memory and fill the gaps using their own beliefs. This is also an information compression process, at the same time producing wrong memory and biases.
Each recall can reshape the memory according to the existing beliefs. (This is similar to quantum effects: if you observe something, you change it.)
Information is costly to store, and even more costly to index and query. Sometimes forgetting is just not being able to query the specific memory that is stored in brain (and may be recalled if some cue were found that enables querying it). The “querying capacity” of brain is limited and can be occupied by distracting things.
(Side note: the similar principle also applies to computer databases. Just writing information into a log is easy and fast. But indexing the information to make them queryable is harder.)
Taking notes is one way to mitigate the unreliable memory issue.
Every time a messsage is relayed through a person, some of its information gets lost, and some noise gets added. The person relaying the message will add their own understanding (which can be misleading), and omit the information that they think is not important (but can be actually important). This issue is very common in big corporation and governments. Good communication requires reducing middlemen.
People usually remember the “special” things well. This is an information compression mechanism that filters out the unimportant details.
Peak-end rule: People judge an experience largely based on how they felt at its peak (its most intense point) and at its end. The most efficient way to improve user experience is to improve the experience in the peak and in the end (and at the start).
Serial position effect: people tend to recall the first and last items in a series best, and the middle items worst. Interestingly, the same effect also applies to LLMs, called “lost in the middle”.
Cryptomnesia: Treating other peoples’ idea as own original idea, after forgetting the source of the idea.
People seeks information that they are interested in. The seeking of interesting of information drives both curiosity and information addiction.
As with food, we spent most of our history deprived of information and craving it; now we have way too much of it to function and manage its entropy and toxicity.
— N. N. Taleb, Link
Most information in the world is junk.
The best way to think about it is it’s like with food. There was a time, like centuries ago in many countries, where food was scarce, so people ate whatever they could get, especially if it was full of fat and sugar. And they thought that more food is always good. …
Then we reach a time of abundance in food. We have all these industrialized processed food, which is artificially full of fat and sugar and salt and whatever. It was always been for us that more food is always good. No, definitely not all these junk food.
And the same thing has happend with information. Information was once scarce. So if you could get your hands on a book you would read it, because there was nothing else.
And now information is abundant. We are flooded with information, and much of it is junk information, which is artificially full of greed, anger and fear, because of this battle for attention.
It’s not good for us. We basically need to go on an information diet. Again the first step is to realize that it’s not the case that more information is always good for us. We need a limited amount. And we actually need more time to digest the information. And then we have to be of course also careful about the quality of what we take in, because of the abundance of junk information.
The basic misconception I think is this link between information and truth. The people think “ok if I get a lot of information, this is the raw material of truth, and more information will mean more knowledge”. That’s not the case. Even in nature more information is not about the truth.
The basic function of information in history, and also in biology, is to connect. Information is connection. And when you look at history you see that, very often, the easiest way to connect the people is not with the truth. Because truth is a costly and rare kind of information. It’s usually easier to connect people with fantasy, with fiction. Why? Because the truth tends to be not just costly, truth tends to be complicated, and it tends to be uncomfortable and sometimes painful.
In politics, a politician who would tell people the whole truth about their nation is unlikely to win the elections. Every nation has these skeleton in the cupboard, all these dark sides and dark episodes that people don’t want to be confronted with.
If you want to connect nations, religions, political parties, you often do it with fiction and fantasies.
— Yuval Noah Harari, Link
Information bias: Seeking out more information even when more information is no longer useful.
With confirmation bias, more information lead to higher confidence, but not better accuracy. This is contraty to statistics, where more samples lead to more accurate result (but still suffer from systematic sampling bias).
Having no information is better than having wrong information. Wrong information reinforced by confirmation bias can make you stuck in a wrong path.
Popularity of false information increase the value of true information. The “carrot problem”: during WWII, the British claimed their pilots’ night-vision success came from eating carrots, hiding their new radar technology. The best way of hiding something is to override it with another thing.
Browsing social media makes people learn biased distribution of world. Such as:
What’s more, social media may make people:
Social medias are doing “natural selection” to memes. The recommendation algorithm makes the posts that induce more interactions (likes and arguing) more popular. It selects the memes that are good at letting human to spread them.
What memes have higher ability to spread?
In the ancient world, when there was no algorithmic recommendataion, there was still the “natural selection” of memes (stories, cultures).
Memes facilitate being spreaded. On the contrary, antimemes resist being spreaded.
People wants attention from others. Some people try to gain attention by posting things on the internet.
Attention is a psychological commodity which people value inherently.
…
Producers who go viral produce 183% more posts per day for the subsequent month.
— Paying attention by Karthik Srinivasan
Giving randomized feedback (variable-ratio reinforcement) make people more addicted to the behavior. Random outcome is usually more exciting than known outcome.
Examples:
This is related to information addiction. Randomized things give more information than deterministic things.
Unfortunately, just knowing the cognitive biases is not enough to avoid and overcome them. A lot of cognitive biases originate from the biological basis of human’s cognitive function, which cannot change from just knowledge.
Reference: G.I. Joe Phenomena: Understanding the Limits of Metacognitive Awareness on Debiasing
Consider two financial trading strategies:
In the long term, strategy 2 can greatly outperforms strategy 1, but people prefer strategy 1, because of many reasons:
It’s a common misconception that a you need a win rate more than 50% to be profitable. With a favorable odds, profit is possible despite a low win rate. The skew of profit distribution is important.
Disposition effect:
Disposition effect works well in oscillating markets. However, markets can also exhibit momentum, where disposition effect is detrimental.