Full transcript
0:00This video is sponsored by boot.dev. Learn back-end web development through immersive
0:04gamified experiences. Go to boot.dev and use code wonder to get 25% off your first annual plan.
0:12You’re twenty-three years old. You just moved into a modest two-bedroom apartment
0:15with a roommate—a neutral acquaintance you’re splitting costs and responsibilities with. To
0:20ensure everything is fair and cordial, you both establish who does what in the apartment—when and
0:25who takes out the trash, cleans the floors and counters, does the dishes, and so on.
0:30You determine you’ll each do the dishes once a week. You take Sunday. They take Wednesday.
0:35Soon, the first Sunday comes around, and, as planned, you do the dishes. Then the first
0:40Wednesday comes, and your roommate does them. This continues for a couple of weeks. Then,
0:45on one Wednesday, after coming home late from work, you find the dishes piled up in the sink.
0:50You don’t say anything. In your generous nature, you assume it’s a one-off thing,
0:54and your roommate will just do them the next day. When that Sunday comes, however, the pile of
0:59dishes is twice as high, spilling out of the sink and onto the surrounding countertop. They
1:04never did them. You want to tell your roommate that they should do the dishes today, but they
1:08aren’t home—and haven’t been all day. And so, you do them. This keeps things on schedule anyway.
1:15When the following Wednesday comes, your roommate does the dishes. All is well and
1:19back to normal. That is until the following Wednesday, when again, late into the evening,
1:24you find the sink filled with dishes. You go to your roommate and ask them what’s going on. They
1:29assure they will do them. You accept this. The next day, however, the dishes are still there,
1:35pilled even higher. And then the next day. And the next day. You realize there’s a problem.
1:41When that Sunday comes, you wonder what you should do. Do the dishes or leave the already
1:46giant pile to grow bigger? What precedent have you set? What precedent will you set?
1:52Can you reset things? What if you don’t do the dishes, and then your roommate still doesn’t
1:56do them? The kitchen will remain a constant mess. Alternatively, what if you do do them,
2:01and soon you’re always doing them every time? Who are you dealing with here, and how can you
2:07deal with them most effectively? You wonder what decision—based on what strategy—you should make.
2:13*** This is a version of a famous thought experiment
2:17in game theory known as the prisoner’s dilemma—a situation where two people would be better off
2:21cooperating, but each person has some incentive to go against the other. In both not cooperating,
2:27however, both end up worse off. In this case, the individual incentive is not spending time doing
2:32the dishes. The outcome is either a messy kitchen and messy roommate situation or a clean one.
2:40Broadly, game theory is the mathematical study of decision making and strategies in situations
2:45where outcomes depend on others’ choices. More specifically, it examines the nature and effects
2:50of how conflict and cooperation among rational decision-makers can lead to optimal or suboptimal
2:55payoffs. In essence, it is a science of strategy. In social situations, in business, in economics,
3:02and in politics—in every interaction—whether between as little as two people or as many as
3:08nations, decisions are constantly being made that affect everyone involved. As individuals
3:12and collectives, we each possess the power to not only change our own circumstances but also
3:17the circumstances of others—of the world. These decisions and their outcomes can be as benign as
3:23who does the dishes in an apartment to as critical as whether a country and its citizens survive.
3:30Game theory suggests that every decision made with a particular aim can, in principle, be represented
3:36and understood as a mathematical model. In other words, with a clear goal and defined constraints,
3:41a rationally right choice can always be determined. More yet, an optimal strategy
3:46can be determined across multiple choices. And through various computer programs and simulations,
3:52researchers in the field of game theory have actually found a strategy (or an approach and
3:56temperament) that, under many conditions in society and nature, has proven to consistently
4:01be extremely effective. And surprising to many in the field, the strategy is profoundly simple. Even
4:07more compelling, it is hopeful. It is something that each of us can apply to our own lives.
4:14Before going any further, it’s important to note that in the context of game theory, a “game” is
4:19not how we conventionally think of one—though it can also include traditional games. Rather,
4:24“game” simply refers to any interaction that occurs between multiple decision-makers,
4:28where the outcome and payoff of the interaction for each individual depends on the choices made
4:32by the others. This can include games like chess and poker, but it also includes nearly everything
4:38else. Of course, not literally everything—but all direct interactions that occur between
4:43individuals or groups that involve competition or cooperation, where there is a mutually affected
4:48outcome. And this is almost everything. Importantly, however, game theory does
4:53delineate two types of interactions (or games): cooperative and non-cooperative. Cooperative
4:59game theory includes dynamics like players on the same sports team, roommates (in theory),
5:04business partnerships, as well as international alliances and trade agreements. In these cases,
5:10goals are shared, resources and information are often freely exchanged, and fairness and mutual
5:15benefit is both implied and actively pursued. Non-cooperative game theory, however,
5:20is far more prevalent in the world—and arguably much more interesting. In non-cooperative games,
5:26there are typically winners and losers, as players act independently in their own interests, making
5:31choices intentionally to benefit themselves, potentially at the expense of their opponents.
5:36This sort of non-cooperative dynamic and tension is often used and simplistically
5:40reproduced in game shows. For example, in the late-2000s British game show Golden Balls,
5:45two strangers would sit across from each other and decide whether they wanted to split or steal
5:49a large sum of money with the other person. Each persons’ choice directly affected whether either
5:54individual got any money and how much, but neither would know the other’s final choice
5:59until it was revealed and locked in. If both chose to split, they shared the money equally. If one
6:04chose to split, and the other chose to steal, the one who chose to steal got all the money,
6:09and the other person got nothing. If both chose to steal, neither person got anything.
6:14In these sorts of one-off situations, where one can either split or steal–cooperate or defect—game
6:20theory shows us that there is a clear rational choice. What is known as the dominant strategy
6:26refers to a choice that provides a player with the best results no matter what the other player does.
6:31And this is always the most rational choice to make. The choice is not about what could
6:35lead to the best possible outcome, but about choosing for the best outcome no matter what
6:39the other person decides to do—since you have no control over that. And so, in Golden Balls,
6:44the most rational thing to do would be to always steal. This is because if one person splits,
6:50then the other person does better by stealing. If one person steals, again, the other person,
6:55in a sense, does better by stealing, because they get the same amount as splitting (zero),
7:00but are not manipulated or exploited by the other person. Technically, this is what game
7:04theorists would call a weakly dominant strategy, since the literal payoff in this later situation
7:09would be equal to splitting, rather than better. Of course, however, life is not a gameshow.
7:16Interactions are almost never one-offs without lingering, continued effects. Decisions are rarely
7:21as simple as splitting or stealing, and outcomes are rarely as simple as half, all, or zero. In
7:27real life, there is almost always a much greater interplay with time, repeated interactions,
7:32uncertainty, leverage, and resources. If someone does or doesn’t do the dishes once,
7:37that game is not over. The relationship and space are and can be either benefited or strained,
7:42moving forward. When a business smears or partners with another business, that game is not over.
7:48Retaliation or a growth in resources can and will likely follow. When a country attacks, retaliates,
7:54or allies with another, that game is not over. Wars can begin or end. Nations can begin or end.
8:02With all this in mind, what is the most effective decision-making strategy (or approach and
8:07temperament) in life in general. Is there one? In 1980, political scientist Robert Axelrod set
8:14out to test and answer this very question. Using computer programs to model different
8:19decision-making strategies, Axelrod orchestrated an experiment. He had leading theorists from
8:24various disciplines and places around the world create and submit programs that would
8:28compete in a tournament of an iterated version of the prisoner’s dilemma. The
8:32goal was to submit the best strategy and win. The rules of the tournament were simple. Each
8:37player (or program) played a single game against every other player—as well as a copy of itself.
8:42In each game, each player had the option to either cooperate with or defect against their opponent.
8:47If both players cooperated, they both received three points. If one cooperated and the other
8:52defected, the player who defected received five points, and the player who cooperated
8:56received zero. If both defected, both received one point. Each individual game contained 200
9:03rounds. The player with the most points by the end of all the matchups in the tournament won.
9:08In total, fourteen programs were submitted—and then Axelrod added one that defected or cooperated
9:13at random, with a 50% probability each round. Most of the submitted strategies began with
9:18cooperation, while others began with early defections. Some players were complex and
9:23calculating, probing for weakness and then exploiting it—like a program called Graaskamp.
9:28Some mixed in random moves to utilize confusion and surprise–like a program called Joss. Others
9:33were far more straightforward. Together, the programs spanned from what Axelrod referred
9:37to as simple and nice to cunning and nasty. After the tournament concluded, Axelrod, along
9:44with many other game theorists, found the results profoundly surprising. He ran the whole tournament
9:50again, five times over, to ensure the results were dependable and repeatable. Each time, the results
9:56were consistent, and the same winner emerged: a program called tit-for-tat, which was one of the
10:02simplest and most cooperative programs of all. To further elevate the complexity and better
10:07mirror real-world circumstances, Axelrod conducted a second tournament. This time, there was no
10:12defined number of total rounds per game. With it now being a random, unknown number, players could
10:18no longer track and calibrate their decisions against a defined endgame. Just like reality.
10:23This time, sixty-two program strategies were submitted—and again, Axelrod added one that
10:28was random. The results were very consistent to the first tournament. Again, tit-for-tat won.
10:36Axelrod and many other game theorists found this extremely surprising because the expectation was
10:40that the winning strategy would be either highly complex, highly competitive, or both
10:45(i.e. cunning and nasty). And yet, tit-for-tat was generally simple, nice, and forgiving.
10:52In terms of specific gameplay, tit-for-tat always starts with cooperation. From there,
10:57it always copies its opponent’s last move. And so, it continues to cooperate unless or until
11:02its opponent defects. At that point, tit-for-tat immediately defects back and continues to do so
11:08unless or until its opponent cooperates again. As soon as its opponent cooperates again, tit-for-tat
11:13then forgives (or no longer accounts for previous moves) and returns to cooperating unless or until
11:19its opponent defects. So on and so forth. Interestingly, the results of this strategy
11:24equated to tit-for-tat never winning any individual games, since one-on-one,
11:29it can only lose or draw. But across all match ups and games, it cooperated with
11:33enough other players to consistently end up with the highest score overall and win the tournament.
11:39Axelrod writes in The Evolution of Cooperation: What accounts for TIT FOR TAT’s robust success
11:44is its combination of being nice, retaliatory, forgiving, and clear. Its niceness prevents it
11:50from getting into unnecessary trouble. Its retaliation discourages the other side from
11:55persisting whenever defection is tried. Its forgiveness helps restore mutual cooperation.
12:00And its clarity makes it intelligible to the other player, thereby eliciting long-term cooperation.
12:06Moreover, almost all top performing players in the tournaments shared in these similar
12:10qualities. And in later simulations with even more realistic, chaotic conditions,
12:14a more generous version of tit-for-tat, that occasionally forgave defections instead of
12:19reciprocating them, proved to be even more effective. Nasty players, on the other hand,
12:24seemed to often find themselves in defecting wars that lead to mutual destruction. Axelrod says:
12:29“What makes it possible for cooperation to emerge is the fact that the players might meet again.”
12:35The takeaway from this is reasonable clear. In continued non-cooperative, competitive
12:40interactions—like these tournaments—it is often beneficial to, at the very least, try to be nice.
12:46To lead with niceness and cooperation. This is not a weakness, but a strength. Conversely,
12:52an individual who often leads with or insights defection is more likely to weaken itself
12:57and lose over time—even if it initially appears as if they are winning. Moreover,
13:02holding grudges is a weakness; forgiveness is a strength. Of course, however, weakness itself is
13:08a weakness. That is, letting someone do you wrong without consequence will lead to being
13:13taken advantage of and losing. One’s approach to exacting consequence, however, is also important:
13:19it must be relatively equal, consistent, and clear, not opaque and manipulative.
13:25From a moral and historical perspective, the winning tit-for-tat strategy essentially mirrors
13:29an eye-for-an-eye ethos. That is, justice and punishment should be proportional to the harm
13:34caused by an offense, but after proportional consequence, balance and cooperation can and
13:40should be restored. On a more individual level, it essentially equates to being kind, forthright,
13:45and understanding, but never a push-over. Of course, there are problems and limitations
13:50both with Axelrod’s experiment as well as game theory in general. Programs, simulations,
13:55and theories can arguably never truly reproduce, model, or assess the true scale and complexity of
14:01real-world interactions. Real interactions often involve many people and many issues;
14:06many perspectives, many goals; shifting ideas and opportunities; asymmetric leverage and resources;
14:12known and unknown information; vast errors and chaos; and, perhaps most importantly, they involve
14:19the very emotional, sentimental, spiteful, and irrational nature of the human mind. As humans, we
14:25feel and hope and believe at least as much as if not more than we assess, calculate, and execute.
14:33Ultimately, however, game theory teaches us many important things. Perhaps one of the most
14:37important being is that every interaction and game is not always about winning.
14:42A strategy always focused on winning can actually be the least effective at winning overall—whereas
14:47one less focused on always winning can, over the long term, win. If we want to truly be
14:52successful across various areas of life, there are going to be—need to be—many instances of draws
14:58and losses. But so long as we continue forward with each new moment and each new interaction,
15:03open and willing to try again, ensuring we stand up for ourselves and hold true to our values,
15:08while striving to meet and unite with the world around us, we can steadily and surely
15:13move toward bigger, more important wins—wins of cooperation, kindness, and mutual benefit.
15:20We can never truly know or control if people will cooperate or defect with us, but we can know and
15:25control if we will and why. And we can know that each decision we make will likely influence the
15:31nature and outcome of all the games in which we participate—present and future—potentially making
15:36or breaking relationships, goals, systems, or even society and the planet. And so,
15:42at least for starters, for own sake, when our day comes, let’s be sure we do the dishes.
15:55[Music] Games are a fundamental aspect of all of life and so naturally our brains are wired
16:01to often treat and make sense of things through gamification. It is how we effectively build,
16:06understand, and improve. That's why this video sponsor boot.dev is such an amazing fit for this
16:12video and such an incredible platform for learning programming, particularly back-end
16:16web development. Boot.dev Dev functions as a programming RPG game using modern game design
16:22to not only immerse you into coding, but help you effectively achieve your goals in real life.
16:27Whether it's passing technical interviews, building a successful programming career,
16:31or creating your own web applications and services. Go on quests, fight bosses, earn XP,
16:37levels, and achievements, all while building and understanding real programming projects in Python,
16:42SQL, and Go. If you're ever stuck, Boots, a bear wizard, is always around to help keep you moving
16:48without letting you off the hook. Boot.dev also recently launched the training grounds,
16:53which is a place where you can continually practice coding through infinite challenges before
16:57venturing off into quests. As you move through the games and courses without even realizing it,
17:02your programming abilities will grow. And of course, this is an extremely valuable skill. The
17:08median salary for back-end developers in the US in 2024 was well over $100,000. And programmers often
17:14have the option to work remotely, which can be a great benefit. Start your journey today and level
17:19up in the game of life by going to boot.dev and use code wonder to get 25% off your first annual
17:25plan. And of course, as always, thank you so much for watching in general and see you next video.
17:37[Music]