Game Theory

What is game theory? Studying predictable behavior in situations in which your choices affect others and their choices affect you.

Why do we play games? "In war the will is directed at an animate object that reacts" - Karl Von Clausewitz in "On War", 1832.

Why Study Game Theory?

"As for the firms that want to get their hands on a sliver of the airwaves, their best bet is to go out first and hire themselves a good game theorist". - The Economist, July 23, 1994 p.70.

How do we benefit?

Skills Required

1. Arrange Problem Set
2. Design Strategy Memo
3. Apply in Real-world
When Making Decisions
  • Take the world as given
  • Make the best decision for yourself depending on:
    • What subjects | opponents do
    • What subjects | opponents think you do

Key Elements

Players: Who is interacting?
Strategies What are their options?
Payoffs: What are their incentives?
Information: What do they know?
Rationality: How do they think?
Remember, controlling the game means keeping the focus on "them".
Gain Advantage Using:
  1. Credibility, threats, and promises.
  2. Reputation and surprise
  3. Exploiting incomplete information: using signalling, selection, and information cascades as tools.
Summary: Game theory helps us learn from experience. It also helps us to predict the outcomes of our strategies. But many caveats. Essentially, game theory teaches how to think strategically.

Hawk-dove theory
If something is in our territory, we think of it as ours i.e. it is our "right" to own it. We will fight to defend it, even if it is relatively useless or even if we can get something better somewhere else. Loss Aversion: We value things we already possess more than things we do not possess. We also value things we make ourselves more than we value other things. How do we get people to fight over things? Put things in their territory as close as possible to them and see who puts up the biggest fight.

"Rights" are what the majority of people do. Sometimes we do not fight for what we regard as ours. Why? Because it is in someone else's territory. "You found it, yes, but in my territory, therefore I expect to keep it" - a strong impulse.

In society, the impulse to collectively punish individuals has been decreasing over time.
"Ultimate Principle" = A general principle we hold to be true for all, such as the right not to suffer collective punishment.
"Proximal Principle" = A situational principle i.e. whatever suits me from one moment to the next.
Collective rights have been increasing over time. People have become more "human" i.e. more likely to stand up for the Ultimate than the Proximal.
Revolutions have always been happening more frequently. But when property and territory are in question, as described previously, the proximal wins, by far. Over time, Ultimate principles benefit Subjects more than Models. Conversely, Proximal principles benefit Models more than Subjects.

Summary: If it is in your territory it belongs to you. Forget ultimate principles, they do not make sense in the animal kingdom. It is humanly impossible to regard both Ultimate Principles AND Proximal Principles as being "right" at the same time. Subjects who are more Proximally-orientated are less of a threat to Models. Meaning, Subjects have less desire to hold Models to account, based on arguments regarding Ultimate Principles. The dilemmas of Subjects i.e. The problems that most concern Subjects, must be as close to Subjects as possible. It is the task of Models to make the centers of power distant from Subjects.

We do good for others because we benefit from doing good for others. This idea is called "reciprocal" altruism but it is also known as "cooperation".
The closer we are to others, the more altruistic we are likely to feel towards them. For example, we are unmoved by statistics about the scale of a problem such as HIV/AIDS. However, if we know someone who is dying of AIDS, we are moved by the problem.

Purpose: In society we want everyone to contribute to the common good.

Process:

  1. We choose whether or not to contribute to general public good at our own expense. However, we do not know all the people to whom we are contributing, so the dominant is not to contribute.
  2. Everyone must have a chance to punish those who have not contributed.
  3. Everyone must have a chance to punish those who did not punish properly.

Expected Outcome: Everyone contributes to the public good. In our society, opinions of subjects about public good are not relevant. Nor are the ideas of Subjects about the best strategies to fulfill the public good. Therefore, Subjects do not experience their own personal worth in contributing to the public good. Naturally, this leads to resentment among Subjects, who may constantly experience the desire to punish others as the solution to their resentment. Being able to experience our own worth is important!

Case Study - Snipers, on both sides, during World War 2 often did not shoot directly at one another. Instead, they showed their skills to one another during battle by cleverly shooting a circle of holes in a wall next to the opposing sniper.

Question: What can we learn from this?
Answer: We do good when we can be seen doing good. We do not do bad when we can be seen not to be doing bad. Conversely, if people are going to do bad they would rather not be seen doing bad.

Summary: Altruism cannot rely on anonymous co-operation without giving in return. In other words, we need to feel good about doing good and also feel good by not doing bad. Therefore, the purpose and outcome of our actions must be seen. This is well understood in the political arena. But policy needs to be understood by Subjects. Otherwise, altruism in society cannot be sustained. Example, policy makers try to get noticed for good policies i.e. policies that benefit Subjects, but they also prefer to hide their policies when they harm Subjects.

Rule of thumb: a policy t hat is foisted upon Subjects without their cooperation, understanding or even knowledge of the policy, is bad for the public. When policies are bad for Subjects, the Subjects question the good intentions of the policy.

Altruism is about actions, cooperation is about thoughts, reflections and search processes. Therefore, we look for signs of cooperation, and we look for evidence of altruism. Why? Because cooperation is always sub-conscious. That is why we value deliberate cooperation less than non-deliberate cooperation e.g. "Can I count on you?" "Let me think about it". When we slow down, we become less cooperative. Why? Because we take the time to consider, think, or reflect on defecting! Models acquire as opposed to earn the cooperation of Subjects:
  1. By making a direct, as opposed to indirect, connection between cooperation and trust.
  2. By making Subjects hurry.
  3. By threatening penalties for slowing down.
Loyalty implies that Subjects expect something in return for their cooperation e.g. "I am deeply committed to those who are committed to me. The people who are close to me know they can count on me". Loyalty works OK when there is a clear and present "them" out there. e.g. Defending someone even when they are wrong is a form of loyalty. Loyalty is a quality that must be proved and tested over time. Therefore, Subject obedience is better for Models.
But there is a problem: "You are obedient, but are you loyal?".
Cooperation and trust form an uneasy relationship e.g. Although you are telling me exactly what I want to hear" (cooperation) "do you mean it? Besides, what do you really mean?" (trust).

Hierarchy

  1. Trust
  2. Cooperation
  3. Loyalty
  4. Obedience
  5. Vascilitation
  6. Defiance
Models must expect cooperation without looking from Subjects. Therefore, Subjects must be made to understand that cooperation with looking has penalties. That is to say, the penalties for defection must be more uppermost in the minds of Subjects than the benefits of cooperating. For example, "You are either with us or you are with the terrorists". In this example, the Subject must be able to avoid looking, while Models, on the other hand, must be able to detect looking. We do not consciously choose when to look and when not to look before we cooperate
To look is to be "strategic", not to look is to be "principled". But no one can consciously choose to avoid or detect looking.
Summary: In free associations we valued "principled" people over "strategic" people. Otherwise we rely on feelings, heuristics and ideology to guide us. In general we value cooperation without looking more the cooperation with looking. For example, we regard giving without looking - disinterested altruism, as highly honorable. The worst taboo in society is a willingness to look at the benefits of defecting, before defecting. The best myth in society is that loyalty, cooperation and trust do not have to be earned, they can be acquired.

What is Common Sense? I know that you know that I know that you know that I know... and so on. Common sense, like cooperation, is not conscious. Common sense has a cost: We regard errors of commission as more offensive than errors of ommission. We do things only when we can be absolutely sure that society will interpret our actions as being well-intentioned. In other words, there needs to be evidence that everyone will perceive an event in the same way. An apology, although it may not be heartfelt at all, is evidence of being sorry. I believe that you believe that I believe that you believe you are sorry. When someone "apolgises" we all feel like we can, or at least ought to, move on... What else can we do?
Symbolic gestures, such as handshakes, although not having any intrinsic value, have common knowledge value. Symbolic gestures are controversial; flags are particularly controversial. A handshake, or no handshake, between two politicians can make the world headlines as a controversy. The battle for the common sense ground can get out of control. Symbolic gestures can come in the form of bombs. Phrases such as "sending a strong signal", "our message is clear" and so on are symbolic gestures with the aim of occupying the common sense of the moment. The battle for the common sense is commonly understood to be a game but it is certainly not a joke! Why? Because symbolic gestures, although they are literally symbolic nonsense, condition future behaviour.

Spreading rumours is an attempt to occupy the common sense ground, but it is hardly ever successful. Why? Because I cannot be sure that you believe the rumour. Innuendo is much more powerful, due to the insidiousness of it. Innuendo bypasses our critical faculties and settles in our subconscious. We we use innuendo we get to have our cake and eat it too! But neither rumour-spreading nor innuendo can create common sense i.e. a solid foundation for future action; only an error of commission can create solid common sense. Why not an error of omission e.g. telling an untruth? Because we cannot be sure that everyone else has seen the error of omission. "Who threw the first stone?" - powerful argument in the game of common sense. Judges warn and advise a jury about such matters. But how effective can the judge be if the game of common sense is not conscious?

Summary: Our common sense is what other people see happening. We are prepared to voluntarily condition our behaviour, but only if our behaviour creates common knowledge. In the arena of common sense, what we believe other people see happening matters more than what we see happening.

What is beautiful? Beauty is in the eye of the beholder. But beauty is not conscious. Examples: Common theme in these examples! Spending on visible beauty e.g. clothes, as opposed to inconspicuous beauty e.g. chocolate increases in proportion to the wealth of the people with whom we are associating. What about fashion? Fashion is what everyone knows is beautiful - a form of socialization. Ironically, fashion happens when beauty becomes common sense.

Looking beautiful is supposed to indicate other things about us too e.g. wealth, status, intelligence, "goodness" and so on. What is fashionable? Fashionable is whatever people without $$$$ cannot afford to do or wear. If something becomes affordable for everyone, it is no longer fashionable. We are not trying to signal wealth, we are trying to look beautiful. Looking beautiful is easier when we have $$$$ and harder for those with less $$$$. Looking beautiful when we cannot afford to look beautiful is also known as "Costly Signalling". It is "costly" in terms of effort.

But anyone can wear white for example. Yes, but white what? Linen, Polyester, Nylon? Social psychologists measured award-winning topless models of the year between 1960-2000. When economic conditions were hard,the models were heavier. It used to be hard for people without $$$$ to have body fat, now it is hard for them to be slim. In places where it is still hard for people without $$$$ to be fat, heavier wives find a better mate.

Fair skin used to be fashionable for caucasians in the U.S. Why? Because fair skin indicated indoor work while tanned skin signalled outdoor work. Then, with the explosion of air travel, tanned skin became fashionable. Tanned skin indicates that someone can affort to travel to hot climates. Tanning salons help us look like we can afford to travel to hot climates all year round. Prediction: Tanned skin is likely to go out of fashion again.

Summary: The "beauty" of Fashion - can you afford to look good? Beauty is in the eye of the beholder, fashion is in the bank account of the trend-setter; those of us without $$$$ ought to try and keep up!

Subjects imitate and repeat things that have good outcomes - for themselves. But Subjects also imitate Models who they regard as "successful", regardless of the outcomes for Subjects. This phenomenon is known as "forgetting about function". It is based on superstition - "If it feels right, do it decisively and then hope for the best". This phenomenon is not conscious and its a natural tendency in everyone. Research is based mainly on experiments with children, chimpanzees and pigeons.

Subjects believe strongly that positive association with prestigious Models will lead to better outcomes for themselves. How do we know this? Policies that harm the public i.e. policies that benefit the few at the expense of the many, can be implemented if the Model is regarded by the public as having "prestige". Subjects are less focused on the policy itself and more focused on the prestige of the Model.

For example, a political leader is trying to design and implement a policy that is not in the public interest. The "forgetting about function" thesis predicts that Subjects will argue with one another about how the political leader is approaching the problem of gaining the consent of Subjects. However, Subjects tend not to consider how the policy will actually affect them, even though the policy will harm them when it finally becomes legislation or law. Subjects also tend to suggest better ideas and offer handy advice for the Model, without considering the consequences of the policy for themselves, as Subjects.

Public rituals that emphasize the prestige of Models are necessary for gaining the cooperation of Subjects. Public rituals with this purpose are also known as "Reinforcement Training".

Subjects also believe that having a negative association with prestigious Models will mean negative outcomes for themselves. For example, if a Subject, who has no prestige, is completely against the policy and starts to protest, other Subjects will feel like it is sour grapes, begrudging and so on.

Experiment

  1. Show children a video of "prestigious" Models getting attention, and another Model getting ignored.
  2. Then, show both Models expressing preferences over food, art, beverage etc.
  3. Now, elicit Subjects' preferences over these categories.
The findings are overwhelmingly conclusive.

What is the difference between "learning" and "training"?
Answer: The presence of a reward | punishment system. Known among Models as the "Logistic Regression" thesis.

Summary: Subjects are more likely to take orders, advice and so on, from someone who is "successful". Subjects do not need to be aware of the "usefulness" of what they are doing. Subjects can be trained to want what they want, it is not necessary for Subject's preferences to evolve. Reinforcement rituals and training must be done often enough for training to be effective.

Overall purpose is to make Subjects "feel" better-off than they evidently are.

Models must control the superstructure of the game i.e. both the game and the rules of the game. Subjects will play automatically. Why? Because Subjects automatically find themselves within a sub-structure game - a game with one another: act alone for a high individual payoff or act together for less payoff - also known as the "Prisoner's Dilemma". For Models, the game of commitment is a combination of two games - "chicken" and "negotiation". Models, if not in charge of the superstructure of the game ought to do one of three things:
  1. Attempt to change the game.
  2. Declare to Subjects that the game is bogus.
  3. Get out!
Threats and promises also known as Carrot and Stick. Both threats and promises go against the incentives of Models. How? Because threatening Subjects may mean losing their cooperation, while making promises means rewarding Subjects. However, Models can turn threats into promises by taking something away from Subjects, and then promising it back in return for cooperation aka "compromise". But Models still risk losing Subjects' cooperation. Therefore, we need an heuristic, and here it is: To be effective, both threats AND promises must appear to Subjects to go against the interestes of Models.

Therefore, make threats clear and promises vague, or at least tenuous. Here is how: Include a potential promise within each clear threat, and include a potential threat within each vague promise. The promise from the Model is turned into a compromise from the Subject. The threat from the Model is turned into a commitment from the Subject.

Example:
Threat: I cannot afford to pay you in your job. If you do not work for less, then I will have to fire you (threat) and I cannot afford to replace you (against own interest).
Promise: However, if you do work for less, then you will keep your job (promise) at some expense to myself (against own interest), given that I cannot even afford to pay you.

(THREAT + potential promise) + (PROMISE + potential threat) = Subject Commitment.

Talk is cheap. For Models, just stating what they will do without strategically seeking commitment from the Subject is self-defeating. For example: "I will fire you!" Full-stop. No commitment sought, no compromise from Subject. Subject is forewarned and therefore prepared i.e. considering their own options, which means Models lose control.

To be effective, warnings or assurances must appear to Subjects as being consistent with Model's own incentives. Why? Because their purpose is to remind Subjects that Models have options. Example: "I can always move my company to another country".

Summary: Threats, promises, warnings and assurances involve risks for Models, So it is crucial for Models to frame them correctly. If Models do not follow through on threats, promises, warnings or assurances, they risk losing the respect of Subjects. Also known as a U-Turn. If Models are not in control of the framework, they should postpone the game till the situation improves.

...the game of "angry negotiation". Examples: Negotiations with organized labour, Price wars, Market battles. Hidden information is always a crucial factor; we are never fully sure how fare the other side will go to win or not to lose.
Game Plan:
  1. Retaliate
  2. Escalate
  3. Disarm
Strategy: Important Questions:
(a) Can either side win?
OR (b) Will both sides lose no matter who strikes first?
If (a), how far will your opponent go to avoid losing? Models need lots of knowledge to decide; too much knowledge to ask for realistically! However, you can probe your opponents limits by gradually escalating the risk - both yours and theirs.
If (b), convey mistaken beliefs, like you just don't "get it". This softens up Subjects, who are more likely to come onboard. AKA "fooling-proof" aka "Deception". All about deception from there on.
Rule of thumb: Convince the opponent that your strategy is more simple than it actually is. Better still, convince your opponent that you have no strategy, just tactics. Why? Your opponent will imagine that you do not care i.e. that you are "crazy". This is hard to do when you have to maintain the support of Subjects. Therefore, remind your Subjects from the beginning to the end of the conflict how "crazy" your opponent is and how "sane" you are. At the same time, remind your opponent that you may be "crazy". Who is "craziest"? - no one will ever know; who is "sanest"? - the one who wants to avoid being hurt most.

To escape from a game of brinkmanship of mutual harm, there are two options: Retaliate harshly i.e. "hurt after being hurt yourself", but only if you have both the ability and incentive for escalation, OR remove strategies that limit your opponent's ability or incentive to retaliate. If your opponent retaliates, decrease the severity of harm to yourself and your opponent while also decreasing the severity of retaliation by your opponent. This can take time.

Always remember: being known by your opponent to know the game theory inside out is not necessarily an advantage.

Summary: "All warfare is based on deception" - Sun Tzu, "The Art of War", 500 BC.

Effort is also known as "hidden action".
Problem: Probability of success depends on Subjects' effort.
Models cannot see effort. Lack of effort on the Subjects' part ought to be regarded as "moral hazard". Why? Because lack of effort can be a slippery slope to defiance, or worse, Defection!

One solution: Connect Subjects' hidden action with visible outcomes. In other words, incentivise Subjects to make as much effort as possible. Incentives cost! So High Hurdle = high incentive; Low Hurdle=low incentive. Routine effort=routine incentive; Low effort= "Moral Hazard!". Costs must balance out over time. The highest incentive ought to be as low as possible.

Best Solution: Subjects must prefer to put in high effort. Conversely, Subjects must prefer not to put in routine effort or less. Subjects must regard expected reward as worthwhile i.e. enough to make a routine effort. However, no base reward is necessary when using the incentives scheme. Remember, Models want to induce high effort without giving away the farm to do it. Negative rewarding i.e. punishment for low effort - Moral Hazard.
Negative rewarding might mean too much risk for Subjects. Remember that Subjects are risk-averse.

Other Solutions: Fixed incentive schemes - no incentives for high effort. Therefore, Models cannot apply effort-based scheme. Subject Monitoring - but routine effort becomes the norm.

Ideal Solution:

Tough vs Soft aka Lean and hungry look vs puppy dog ploy.

Game Plan

  1. Decrease risk
  2. Create more options
  3. Commit strategically
Remember, if you take actions that are in favour of your own incentives, you lose credibility with Subjects. The opposite is true with peers e.g. opponents - you cannot be credible if you propose to take actions that go against your own incentives. How to be credible with peers?

Playing Soft aka Puppy Dog Ploy

"A surrounded enemy must be given a way out" - Sun Tzu in "Art Of War" - 500 BC. Why? Because Opponents' power can come from not being able to retreat. How? Agree to share, while commiting to no opportunistic behaviour. But, reducing your own potential payoffs in an irreversible way, while being credible. This is hard to do. Therefore, use Third-parties, such as the government, to commit to opponents on your behalf or to subsidise your project directly.

When contracting with Subjects, the game almost always leads to the "prisoner's dilemma" for those who are potentially willing to make strategic commitments. However, third-party contracts, via the government, can resolve this.

The Flip Side

You must, simultaneously, make it difficult for your Opponent to act in ways that will hurt you. You must also help your opponent commit to actions that are in your favour. How?

Play Tough - Lean and Hungry Look

Use this strategy only when winning is more important than competing. "At the critical moment, the leader of an army acts like one who has climbed up a height, and then kicks away the ladder behind him" - Sun Tzu, "The Art Of War", 500 BC.

Summary: With strategy, credibility is important. Tough or Soft? That depends on what your opponent wants to do e.g. play less aggressively i.e. compete or play more aggressively i.e. win. And, then it depends on what you want your opponent to do next!

What is it?
For Subjects, Reputation is transparent actions that are known and understood to be "sane".
For Models, Reputation is opaque actions that are known and understood to be "crazy". Models' Reputation aka "mystique".
Reputation can help establish credibility. Losing reputation is costly to credibility. Losing credibility is costly materially. Getting Crazy aka "Strategic Irrationality"
Subjects' view: "Its our lives, but its just a game to them".
Playing Strategic Irrationality
Purpose: To make a group of Subjects want to enter the arena while deterring all others.
Process: Provoke Subjects using one of two techniques:
Sequence of decisions:
  1. Do Subjects enter arena?
  2. If so, do Models fight?
  3. If so, do Subjects stay in arena?
  4. If so, do Models fight?
Remember, if groups of Subjects keep entering the arena then: However, each group of Subjects must worry about their own reputation i.e. sane people acting Crazy. They also must worry about the crazy Model. Outcome: Crazy opponent and Sane Subjects acting crazy are BOTH opponents of Sane Subjects.

Summary: Reputation must be preserved and enhanced. In order to both preserve and enhance reputation games of "strategic irrationality" must be played with Subjects repeatedly. Either side can win if they "get crazy". However, "crazy" for Subjects means loss of reputation.

What is it? Not all it's cracked-up to be.
"Only the paranoid survive" - Andy Grove, Intel Co-founder.

Most Subjects assume "rationality" of decision-makers or "Models". That is, Subjects believe that Models form beliefs about the world by choosing an action that maximizes the welfare of Subjects, given those beliefs. Rationality is something Subjects assume about Models, and is, therefore "common knowledge" among Subjects. It follows, then, that Subject also regard Models as having "correct beliefs".

Problem #1: Rationality requires that both Subjects and Models understand each other. That is, in strategic situations, rationality requires that both Subject and Models know what the other believes. But, in principle, rationality requires enormous powers of imagination and computation. Too much!
Solution #1:However, "conformity" is rational for Subjects, even if they do not know, or do not care what Models are doing. Conformity is a tendency to ignore your own information, non-rationally. But when Subjects ignore their own information it requires an "Information Cascade" from Models, meaning that Subjects look to Models for information about what Models are doing.

So, Subjects conform! Why? For one of two reasons:

We can predict, therefore, that Subjects' behaviour is relatively stable over time.

Problem #2: Models typically reveal relatively little information about their decisions. Subjects with more or better information may swing the behaviour of an entire group of Subjects. When this happens we call it "paradigm shift". However, while some Subjects may be rational, most Subjects may not be. Therefore, Subjects rely strongly on the assumption that Models are "rational", and Subjects then look at information from Models in a rational way.
Solution #2: Models should play their dominant strategy. Period. There is no need to think about irrationality of Subjects. Why? Because Assumed Rationality + Dominant Strategy = No need for common knowledge or correct beliefs. Sometimes Models make mistakes, or, for other reasons, may fail to appear "rational" to Subjects. Yet, dominant strategies give Subjects a clear prescription of what to do, regardless.

Summary: Subjects, assuming Models are rational, look to Models for prescriptions of what to do. But Models reveal little or nothing about their decisions. This creates an information vacuum among Subjects. However, Models, having social influence over Subjects, can fill the information vacuum among Subjects with an information cascade in the form of "fads".

"A cynic knows the price of Everything and the value of Nothing" - Oscar Wilde, Lady Windemere's Fan, 1892.
The word "auction" comes from the latin "auctus", which is the past participle of "augere", to increase. Auction aka "Angry Madness Tournament". The Auctioneer decides the rules but takes bidders i.e. the envionment as given.

Types of Bidder

Private Value: What others know does not effect your willingness to pay e.g. where you eat-out.
Common Value: What others estimate might effect your willingness to pay e.g. unproven oil Field. There are many ways that people might fail to play in equilibrium. One is 'animal spirits' i.e. you want to win more if someone else is trying to win. Bidders sometimes collude. Since winning means you have the highest signal, always bid as if you have the highest signal.

Hallmarks of All Auctions

There is a Prize that has a Value that is never less if others value it more, for which each party makes a bid and where the highest bidder gets the prize by paying the cost of the bid.
Your Bid determines that you win, the cost determines what you pay. Sounds hyper-rational. But remember you are in a hypo-rational tournament
In theory, it helps to have a highest bid aka "optimal bid". How? It trades off risk or not winning vs extra gain from winning with a lower bid. In other words, an optimal bid will keep your 'animal spirits' in check. This is significant in an "angry madness tournament" where Winner pays all.
Animal Spirits: Wanting to win more when someone else is trying to win. Who benefits in a "call out bids" auction - auctioneer or bidders? Who benefits in a sealed-bid auction - auctioneer or bidders?
Collusion: Agreeing on strategies with other bidders beforehand. Who benefits, in the main - auctioneer or bidders? Who benefits in a "call out bids" auction - auctioneer or bidders? Who benefits in a sealed-bid auction - auctioneer or bidders?

Dilemma: What would I be willing to pay, given what I know before submitting my bid? Versus Given what I know before submitting my bid AND someone else is willing to bid more? AKA "Winner's Curse".

There are a variety of real-world phenomena that can be interpreted as Auctions aka "auctions in disguise".

Summary: It's Tough to be Tough.

The aim is to master the unsophistication of Subjects with regard to decision-making. Game theory postulates that everyone is homogeneous but in reality everybody is heterogenous. However, if there are enough irrational Subjects, rational Subjects will start imitating them. How to create enough irrational Subjects? With a sense of immediacy, because emotional systems are activated by a sense of immediacy.

Question: What about the dangers of too many irrational Subjects?
Answer: When emotional autopilot is on, the tendency is to suppress it when it gets out of control.

Heuristics or "Guidelines"

We make decisions using heuristics - informal rules that generate an approximate answer to a problem. Subjects tend to overuse heuristics for decision-making. Why? Because it is "adaptive". But all adaptive processes have side-effects, that is to say, sometimes heuristics lead to incorrect answers, and yet Subjects still base their decisions on them. These are known as Subject Biases.
Subjects expect that a sequence of events generated by a random process will represent the essential characteristics of that process, even when the sequence is short. So, for example, if a coin is fa ir, Subjects expect HHH to be followed by a T - aka "Gambler's Fallacy". The reverse is also true - Subjects expect that the essential characteristics of a process will be represented, not only globally in the entire sequence, but also locally in each of its parts. Broad example: "Running an economy is the same as balancing the budget for a small household."

Anchoring as Necessity

Subjects assess the frequency of a class or the probability of an event by the ease with which instances or occurances can be brought to mind. Therefore, Subjects need Anchoring e.g. a default "option", a lowest "option", a highest "option" and so on. Anchoring effects are strongest when anchors have implicit information value e.g. the "highest" relative to similar other known situations. But anchoring effects are also strong when Subjects do not have much time to think about the problem.

Anticipation in Decision Making

Sophisticated Subjects anticipate; naive Subjects do not anticipate. Anticipate what? "Shrouded information" i.e. long-term consequences of decisions e.g. buying a printer while forgetting the cost of repeat need to buy cartridges.
Question: If 50% of advertising budget is wasted, can you tell which 50%?
Answer: The 50% that makes Subjects anticipate the aftermarket. Subjects, although they are not rational are "internally consistent". Internal consistency aka "maintained assumption" e.g. an "irrational" cocaine addict will still "consistently" prefer cocaine.
Peak-end rule: Subjects remember the end of the experience far more than the beginning or the middle of the experience. Therefore, it is the end of the experience that most influences whether or not Subjects repeat the experience e.g. giveaways as political election approaches. Naive Subjects decide under the false assumption that later selves will do what the earlier self wants. Sophisticated Subjects make decisions based on correct beliefs about the choices of later selves.

Summary: Subjects are irrational, they are creatures of habit i.e. "internally consistent".