WEBVTT
00:00:00.900 --> 00:00:04.100
Let us now conclude with a fun
problem, which is also a
00:00:04.100 --> 00:00:05.970
little bit of a puzzle.
00:00:05.970 --> 00:00:08.210
We are told that the
king comes from a
00:00:08.210 --> 00:00:10.460
family of two children.
00:00:10.460 --> 00:00:13.980
What is the probability that
his sibling is female?
00:00:13.980 --> 00:00:17.220
Well, the problem is too loosely
stated, so we need to
00:00:17.220 --> 00:00:19.370
start by making some
assumptions.
00:00:19.370 --> 00:00:22.800
First, let's assume that we're
dealing with an anachronistic
00:00:22.800 --> 00:00:25.840
kingdom where boys
have precedence.
00:00:29.930 --> 00:00:34.640
In other words, if the royal
family has two children, one
00:00:34.640 --> 00:00:38.960
of which is a boy and one is a
girl, it is always the boy who
00:00:38.960 --> 00:00:43.830
becomes king, even if the
girl was born first.
00:00:43.830 --> 00:00:48.380
Let us also assume that when
a child is born, it has 50%
00:00:48.380 --> 00:00:51.820
probability of being
a boy and 50%
00:00:51.820 --> 00:00:53.320
probability of being a girl.
00:00:58.890 --> 00:01:03.790
And in addition, let's assume
that different children are
00:01:03.790 --> 00:01:07.195
independent as far as their
gender is concerned.
00:01:10.070 --> 00:01:12.850
Given these assumptions,
perhaps we
00:01:12.850 --> 00:01:15.240
can argue as follows.
00:01:15.240 --> 00:01:18.620
The king's sibling is
a child which is
00:01:18.620 --> 00:01:20.600
independent from the king.
00:01:20.600 --> 00:01:24.710
Its gender is independent from
the king's gender, so it's
00:01:24.710 --> 00:01:29.160
going to be a girl with
probability 1/2.
00:01:29.160 --> 00:01:34.750
And so this is one possible
answer to this problem.
00:01:34.750 --> 00:01:36.770
Is this a correct answer?
00:01:36.770 --> 00:01:37.640
Well, let's see.
00:01:37.640 --> 00:01:40.710
We have to make a more precise
model, so let's
00:01:40.710 --> 00:01:42.180
go ahead with it.
00:01:42.180 --> 00:01:47.460
We have two children, so there
are four possible outcomes--
00:01:47.460 --> 00:01:53.240
boy, boy; boy, girl; girl,
boy; and girl, girl.
00:01:55.780 --> 00:02:01.290
Each one of these outcomes has
probability 1/4 according to
00:02:01.290 --> 00:02:02.920
our assumptions.
00:02:02.920 --> 00:02:05.750
For example, the probability of
a boy followed by a boy is
00:02:05.750 --> 00:02:10.960
1/2 times 1/2, where we're
also using independence.
00:02:10.960 --> 00:02:15.150
So each one of these four
outcomes has the same
00:02:15.150 --> 00:02:16.460
probability, 1/4.
00:02:19.660 --> 00:02:23.740
Now, we know that there is a
king, so there must be at
00:02:23.740 --> 00:02:25.720
least one boy.
00:02:25.720 --> 00:02:30.560
Given this information, one
of the outcomes becomes
00:02:30.560 --> 00:02:35.050
impossible, namely the
outcome girl, girl.
00:02:35.050 --> 00:02:39.150
And we're restricted to a
smaller universe with only
00:02:39.150 --> 00:02:41.620
three possible outcomes.
00:02:41.620 --> 00:02:45.470
Our new universe is this green
universe, which includes all
00:02:45.470 --> 00:02:48.100
outcomes that have at least
one boy, so that
00:02:48.100 --> 00:02:50.930
we can get a king.
00:02:50.930 --> 00:02:54.530
We should, therefore, use the
conditional probabilities that
00:02:54.530 --> 00:02:57.590
are appropriate to this
new universe.
00:02:57.590 --> 00:03:01.920
Since these three outcomes
inside the green set have
00:03:01.920 --> 00:03:05.170
equal unconditional
probabilities, they should
00:03:05.170 --> 00:03:08.460
also have equal conditional
probabilities.
00:03:08.460 --> 00:03:12.230
So each one of these three
outcomes should have a
00:03:12.230 --> 00:03:14.460
conditional probability
equal to 1/3.
00:03:17.079 --> 00:03:21.340
In two of these outcomes the
sibling is a girl and
00:03:21.340 --> 00:03:25.275
therefore, the conditional
probability given that there
00:03:25.275 --> 00:03:28.930
is a king and therefore given
that there is a boy, the
00:03:28.930 --> 00:03:34.570
conditional probability
is going to be 2/3.
00:03:34.570 --> 00:03:37.680
So this is actually the official
answer to this
00:03:37.680 --> 00:03:40.710
problem, and this answer
is incorrect.
00:03:43.610 --> 00:03:45.960
Are we satisfied with
this answer?
00:03:45.960 --> 00:03:48.760
Maybe yes, maybe no.
00:03:48.760 --> 00:03:52.579
Actually, some more assumptions
are needed in
00:03:52.579 --> 00:03:56.290
order to say that 2/3 is
the correct answer.
00:03:56.290 --> 00:04:00.160
Let me state what these
assumptions are.
00:04:00.160 --> 00:04:04.960
We assume that the royal family
decided to have exactly
00:04:04.960 --> 00:04:06.830
two children.
00:04:06.830 --> 00:04:11.900
So the number two that we
have here is not random.
00:04:11.900 --> 00:04:14.760
It was something that
was predetermined.
00:04:14.760 --> 00:04:20.140
Once they decided to have the
two children, they had them.
00:04:20.140 --> 00:04:23.720
At least one turned out
to be a boy and that
00:04:23.720 --> 00:04:26.460
boy became a king.
00:04:26.460 --> 00:04:30.780
Under this situation, indeed,
the probability that the
00:04:30.780 --> 00:04:36.540
sibling of the king
is female is 2/3.
00:04:36.540 --> 00:04:39.760
But these assumptions that I
just stated are not the only
00:04:39.760 --> 00:04:41.840
possible ones.
00:04:41.840 --> 00:04:46.040
Let's consider some alternative
assumptions.
00:04:46.040 --> 00:04:48.190
For example, suppose
that the royal
00:04:48.190 --> 00:04:51.110
family operated as follows.
00:04:51.110 --> 00:04:56.930
They decided to have children
until they get one boy.
00:05:00.820 --> 00:05:02.610
What does this tell us?
00:05:02.610 --> 00:05:06.080
Well, since they had two
children, this tells us
00:05:06.080 --> 00:05:06.660
something--
00:05:06.660 --> 00:05:10.300
that the first child
was a girl.
00:05:13.170 --> 00:05:16.110
So in this case, the probability
that the king's
00:05:16.110 --> 00:05:20.340
sibling is a girl
is equal to 1.
00:05:20.340 --> 00:05:23.350
The only reason why they had two
children was because the
00:05:23.350 --> 00:05:27.570
first was a girl and then
the second was a boy.
00:05:27.570 --> 00:05:30.890
Suppose that the royal family
made some different choices.
00:05:30.890 --> 00:05:36.670
They decided to have children
until they would get two boys,
00:05:36.670 --> 00:05:41.190
just to be sure that the line
of succession was secured.
00:05:41.190 --> 00:05:45.520
In this case, if we are told
that there are only two
00:05:45.520 --> 00:05:51.460
children, this means that there
were exactly two boys,
00:05:51.460 --> 00:05:54.220
because if one of the two
children was a girl, the royal
00:05:54.220 --> 00:05:56.170
family would have continued.
00:05:56.170 --> 00:05:59.720
So in this particular case,
the probability that the
00:05:59.720 --> 00:06:02.970
sibling is a girl is
equal to zero.
00:06:02.970 --> 00:06:05.970
And you can think of other
scenarios, as well, that might
00:06:05.970 --> 00:06:09.610
give you different answers.
00:06:09.610 --> 00:06:15.330
So 2/3 is the official answer,
as long as we make the precise
00:06:15.330 --> 00:06:19.120
assumptions that the number of
children, the number two, was
00:06:19.120 --> 00:06:24.280
predetermined before anything
else happened.
00:06:24.280 --> 00:06:27.070
The general moral from this
story is that when we deal
00:06:27.070 --> 00:06:30.440
with situations that are
described in words somewhat
00:06:30.440 --> 00:06:34.010
vaguely, we must be very careful
to state whatever
00:06:34.010 --> 00:06:36.090
assumptions are being made.
00:06:36.090 --> 00:06:39.909
And that needs to be done before
we are able to fix a
00:06:39.909 --> 00:06:43.050
particular probabilistic
model.
00:06:43.050 --> 00:06:47.710
This process of modeling will
always be something of an art
00:06:47.710 --> 00:06:50.590
in which judgment calls
will have to be made.