REASONING BEYOND

THE BACKWARD INDUCTION

HAFSA JAHAN
 REASONING BEYOND
THE BACKWARD INDUCTION

 WRITTEN BY
HAFSA JAHAN

 Backward induction in game

theory is an iterative process of
reasoning backward in time, from
the end of a problem or situation,
to solve finite extensive form and
sequential games, and infer a
sequence of optimal actions. This
process of reasoning allows a
player to think of the end of a
problem and then apply the
sequence of happenings to decide
what to do at a particular stage.

Backward induction is the process of reasoning

backwards in time, from the end of a problem
or situation, to determine a sequence of
optimal actions. It proceeds by examining the
last point at which a decision is to be made and
then identifying what action would be most
optimal at that moment.

 We all know already that rational

behavior refers to a decision-making
process that is based on making choices
that result in the optimal level of benefit or
utility for an individual. The assumption of
rational behavior implies that people
would rather take actions that benefit
them versus actions that are neutral or
harm them. According to rational
behavior highlighted in this theory
sometimes reflect in real life but the
theory does not precisely predict humans.

 In this theory, a player that makes the last

move in a game uses an optimal strategy
determined by this theory. Using this
information, one can then determine what to
do at the second-to-last time of decision. This
process continues backwards until one has
determined the best action for every possible
situation (i.e. for every possible information
set) at every point in time.

 This process continues backward until the

best action for every point in time has been
determined. Effectively, one is determining
the Nash equilibrium of each subgame of the
original game. For example: Let's say Rohan is
the player who will be making the last move
in a game. So, determined by this theory, he
will be using an optimal strategy in order to
benefit the most. On the other hand, let's say
Nikita is another player who will be making
the action before the last one (made by
Rohan).

This action is determined by the last player's

action which is Rohan. As a result, we can tell
that the gaming process moves backward
until the best optimal action for each
subgame is determined. Backward induction
was first used in 1875 by Arthur Cayley, who
uncovered the method while trying to solve
the infamous Secretary problem.

Imagine a maze: Getting through one requires

taking detours, retracing your steps, and
taking alternative routes. It is essentially a
convoluted network of pathways or corridors
between walls or hedges that is intended to
befuddle anyone trying to navigate it,
frequently for fun. The process of creating a
maze involves planning how its walls and
corridors will be organized.

 There are numerous ways to create mazes,

and various labyrinth creation algorithms
can be used to create them automatically or
manually by a computer. Forward induction
is the idea that participants in a game assume
their opponents made rational decisions in
the past and will make reasonable decisions
in the future, even when confronted with an
unexpected event. When you begin at the
beginning of the maze and attempt to
navigate your way to the conclusion, this is
known as forward induction.

Backwards induction is the process of starting

at one end of the maze and attempting to
navigate your way to the other. Backward
induction is the process of figuring out the
best course of action by thinking backwards
in time from the conclusion of a scenario or
problem.

 You should eliminate possibilities

that, like dead ends, do not lead to
the maze's exit while using forward
induction. When using backwards
induction, you should exclude
alternatives that are dead ends or
that cannot be reached effectively,
such as a longer path that overlaps a
shorter path; in this case, you would
eliminate the longer path.
 Knowing the state you wish to reach
and then formulating every action that
could lead there are necessary for
backward induction. The way players
actually play may not be captured by
backward induction. By making plays
that deviate from the equilibrium
course of play, we may be trying to
make our opponent think we're crazy. If
this is the case, backward induction
won't be of any use to us.

 It is a relatively limited class of games that it

can analyze as an algorithm, but other
solution concepts for games with partial or
imperfect information also make use of its
logic, albeit informally. How can a backward
induction be used to solve a sequential game?
Let's attempt to define a sequential game
first. When playing a sequential game, one
person takes the first action, then the other
player takes action after. There are two
players in successive games: player 1 and
player 2.

Consider Player 1 having two strategies. Left

and right are the two approaches. Let's
assume that player 2 employs two strategies.
Player 2 has two strategies: up and down.
Now, if Player 1 plays to the left, Player 2 has
two options. On the other hand, Player 2 has
two options if Player 1 plays Right. Now,
before the game even starts, player 1 can
anticipate what move player 2 will make and
prepare accordingly.

A simultaneous game, often known as a static

game, is one in which each player makes
their decision independently of the decisions
made by the other players. In contrast to
sequential games, which require players to
take turns, simultaneous games are played
simultaneously (moves alternate between
players).

 In a sequential game, the player 1 can select

the best option given the options available to
the player 2. Now, let's say that Player 2's
payoffs are given. When Player 1 selects left,
Player 2's Up matrix is (1, 3), while Player 2's
Down matrix is when Player 1 selects right (3,
2). In addition, if Player 1 selects Right, Player
2 will receive the following payoff matrix: (4,
2). And if Player 1 selects Right, Player 2 will
receive the following payoff matrix if Down
is selected: (3, 1).

 The numbers that are listed at the start of

each payout matrix are now 1, 3, 4, and 3.
These are Player 1's payoff. However, the
numbers that are listed second in each
payout matrix are 3, 2, 2, and 1. These
represent Player 2's payoff. Using backward
induction, we can now determine the player's
preference given the payout matrix. As an
illustration, Player 1 choose the Right. After
then, Player 2 has two choices. Therefore,
Player 2 must select either the matrix (4, 2) or
(3, 1).

The reward for Player 2 is represented by the

second digit in the matrix, as we have already
learnt. Let's focus on Player 2's matrix now if
Player 1 makes the Right choice. Player 2's
payout matrix is 2 and 1. It is obvious that 2 is
greater than 1. Player 2 will always select the
option Up because the bigger number is
always chosen. As a result, we may infer that
Player 2 will always select Up if Player 1
makes the Right choice. Let's now focus on
the other component.

 What will player 2 select if player 1 makes a

left decision? Player 2's payout matrix has the
values 3 and 2. Since 3 is bigger than 2, we
can state with certainty that Player 2 will
select 3—basically the choice Up. Now that 3
is greater than 2, we can predict that if Player
1 choose Left, Player 2 will select Up. In both
scenarios, Player 2 will select Up.

 Now, if we compare the matrices (1,3) and

(4,2), we see that Player 1's and Player 2's
positions are 1 and 4, respectively, and 3 and
2. Player 1 will select 4 because it is bigger
than 1 and therefore 4. Since Player 2 will
always take the option Up because Player 2 is
more than Player 1 and Player 1 can also
have higher value, which is 4, Player 1 will
anticipate this payout matrix and will always
play Right. As a result, we may deduce that
the Nash Equilibrium is (4, 2) using the
backward induction method (R, U). As a
result, this is how backward induction is used
to solve a sequential game.

References:

Rust, John (9 September 2016). Dynamic

Programming. The New Palgrave Dictionary
of Economics: Palgrave Macmillan. ISBN 978-
1-349-95121-5.

What Is Backward Induction? Definition,

How It Works, and Example, By WILL
KENTON, Reviewed by CHARLES POTTERS,
Investopedia