How Decision Making is Actually Science: Game Theory Explained

Main ideas and concepts

1) What game theory is (and why it matters)

• Game theory is a field of math/science that analyzes social interactions where outcomes depend on the decisions of multiple people.
• It was pioneered in the 1950s by mathematician John Nash.
• A “game” in game theory is not limited to sports or gambling; it can be any interaction where your payoff/outcome depends on what others do.
• Game theory is used across many disciplines: economics, political science, biology, military strategy, psychology, etc.

2) Two major branches

• Noncooperative (competitive) game theory
  • Focuses on situations where players do not cooperate and where there are typically winners and losers.
• Cooperative game theory
  • Focuses on situations where players work together toward a shared goal.
  • The central fairness question becomes: how much each player should contribute and how much each should receive.

---

Competitive game theory: Nash Equilibrium (Prisoner’s Dilemma)

3) The Prisoner’s Dilemma setup

• Two prisoners: Wanda and Fred.
• Both are arrested and face an initial likely sentence of 2 years based on evidence.
• The prosecutor/DA offers deals:
  • If one confesses and the other does not:
    • The confessor gets immunity / goes free.
    • The silent one gets 10 years.
  • If both confess:
    • Both get 5 years.
  • If neither confesses:
    • Both get 2 years.
• Key feature: Wanda and Fred are separated, don’t know the other’s decision, and have no loyalty/trust.

4) Game-theory logic that leads to the outcome

• The “best overall” combined outcome is: both don’t confess (2 years each).
• But individually:
  • If Wanda expects Fred not to confess:
    • Wanda still benefits from confessing (she could go free).
  • If Wanda expects Fred might confess:
    • Not confessing could be disastrous (10 years).
• Because each player has incentives to betray if they think the other might stay silent, mutual betrayal becomes the stable result.

5) Nash Equilibrium definition (as explained in the subtitles)

• Nash Equilibrium: a choice where a player is better off (or at least not worse) regardless of what the opponent does.
• In this dilemma, the equilibrium is:
  • Both confess, because neither can improve their outcome by unilaterally switching strategies given the other’s likely response.
• The outcome “makes sense” even if it seems unfair because the game structure rewards non-cooperation.

6) General competitive-game takeaway

• In competitive settings, game theory suggests choosing an action that benefits you most regardless of what others do (i.e., equilibrium thinking).

---

Cooperative game theory: Shapley Value (fair division)

7) What the Shapley Value is used for

• The Shapley Value is a method for dividing gains or costs among coalition members based on the value of each person’s contribution.
• It’s designed to formalize “what’s fair.”

8) Detailed instruction-style axioms behind Shapley Value

To qualify as a Shapley-value division, the method satisfies axioms:

1. Marginal contribution

   • A player’s contribution is determined by what is gained or lost by removing them from the coalition/game.
   • (Example: fewer cookies produced on days a member is absent.)

2. Interchangeable players

   • If two players contribute equivalently (same effect on outcomes), they should receive the same reward.
   • (Restaurant orders: same items → same pay share; similar skills → similar wages.)

3. Dummy players

   • If a player contributes nothing, they should receive nothing.
   • The subtitles note this can be controversial because real-world non-contributors might still need support (e.g., disability, maternity leave), even if they don’t “produce.”

4. Decomposition across multiple parts

   • If a game/task has multiple components, the payment/cost should be split by component, not forced to use one uniform rule.
   • Example logic: work done on Monday vs Tuesday should affect rewards differently; ordering pricier items later should increase that share accordingly.

9) Cookies example (how the value is computed conceptually)

• Two cookie producers:
  • You: 10 cookies/hour alone.
  • Friend: 20 cookies/hour alone.
• Together, you make 40 cookies in an hour (synergy/streamlining).
• Total revenue example: $1 per cookie, so $40 total.
• The Shapley Value approach:
  • Consider removing one player and computing how much value the system loses.
  • Compute marginal contributions in different removal orders, then average.

10) Scaling up

• For coalitions with hundreds of players, the method generalizes:
  • Compute each player’s marginal contribution across many combinations/orderings.
  • Then average appropriately.

---

Overall lesson the video emphasizes

• Competitive situations: game theory helps you be smart (find equilibrium-like strategies).
• Cooperative situations: game theory helps you be fair (divide payoffs/costs using contribution-based fairness like the Shapley Value).

---

Speakers / sources featured

• John Nash (source of game theory’s development/pioneering, credited as pioneering in the 1950s)
• Russell Crowe (referenced indirectly as portraying Nash in A Beautiful Mind)
• SciShow (referenced in the closing production/Patreon segments)
• SciShow patrons / Patreon (mentioned as funding contributors)
• No other distinct speaker characters are shown in the subtitles besides the fictional examples:
  • Wanda and Fred (prisoners in the Prisoner’s Dilemma)
  • DA / prosecutor (offers the deals; not named)