WEBVTT

00:00:06.817 --> 00:00:09.617
Imagine we want to build a new space port

00:00:09.617 --> 00:00:13.200
at one of four recently settled 
Martian bases,

00:00:13.200 --> 00:00:16.650
and are holding a vote 
to determine its location.

00:00:16.650 --> 00:00:23.482
Of the hundred colonists on Mars, 
42 live on West Base, 26 on North Base,

00:00:23.482 --> 00:00:28.252
15 on South Base, and 17 on East Base.

00:00:28.252 --> 00:00:32.342
For our purposes, let’s assume 
that everyone prefers the space port

00:00:32.342 --> 00:00:37.155
to be as close to their base as possible, 
and will vote accordingly.

00:00:37.155 --> 00:00:40.445
What is the fairest way 
to conduct that vote?

NOTE Paragraph

00:00:40.445 --> 00:00:44.400
The most straightforward solution 
would be to just let each individual

00:00:44.400 --> 00:00:48.750
cast a single ballot, and choose 
the location with the most votes.

00:00:48.750 --> 00:00:54.119
This is known as plurality voting, 
or "first past the post."

00:00:54.119 --> 00:00:57.179
In this case, West Base wins easily,

00:00:57.179 --> 00:00:59.791
since it has more residents 
than any other.

00:00:59.791 --> 00:01:04.031
And yet, most colonists would consider 
this the worst result,

00:01:04.031 --> 00:01:07.045
given how far it is from everyone else.

00:01:07.045 --> 00:01:12.099
So is plurality vote 
really the fairest method?

NOTE Paragraph

00:01:12.099 --> 00:01:15.939
What if we tried a system 
like instant runoff voting,

00:01:15.939 --> 00:01:19.265
which accounts for the full range 
of people’s preferences

00:01:19.265 --> 00:01:21.591
rather than just their top choices?

00:01:21.591 --> 00:01:23.131
Here’s how it would work.

00:01:23.131 --> 00:01:27.001
First, voters rank 
each of the options from 1 to 4,

00:01:27.001 --> 00:01:29.651
and we compare their top picks.

00:01:29.651 --> 00:01:34.348
South receives the fewest votes 
for first place, so it’s eliminated.

00:01:34.348 --> 00:01:39.716
Its 15 votes get allocated 
to those voters’ second choice—

00:01:39.716 --> 00:01:43.666
East Base— giving it a total of 32.

00:01:43.666 --> 00:01:49.177
We then compare top preferences 
and cut the last place option again.

00:01:49.177 --> 00:01:51.357
This time North Base is eliminated.

00:01:51.357 --> 00:01:54.926
Its residents’ second choice 
would’ve been South Base,

00:01:54.926 --> 00:01:59.190
but since that’s already gone,
the votes go to their third choice.

00:01:59.190 --> 00:02:05.390
That gives East 58 votes over West’s 42, 
making it the winner.

00:02:05.390 --> 00:02:08.090
But this doesn’t seem fair either.

00:02:08.090 --> 00:02:11.806
Not only did East start out 
in second-to-last place,

00:02:11.806 --> 00:02:16.280
but a majority ranked it among 
their two least preferred options.

NOTE Paragraph

00:02:16.280 --> 00:02:20.867
Instead of using rankings, we could try 
voting in multiple rounds,

00:02:20.867 --> 00:02:25.057
with the top two winners 
proceeding to a separate runoff.

00:02:25.057 --> 00:02:29.120
Normally, this would mean West and North 
winning the first round,

00:02:29.120 --> 00:02:30.848
and North winning the second.

00:02:30.848 --> 00:02:33.509
But the residents of East Base realize

00:02:33.509 --> 00:02:36.029
that while they don’t have 
the votes to win,

00:02:36.029 --> 00:02:39.369
they can still skew the results 
in their favor.

00:02:39.369 --> 00:02:43.289
In the first round, they vote 
for South Base instead of their own,

00:02:43.289 --> 00:02:46.299
successfully keeping North 
from advancing.

00:02:46.299 --> 00:02:50.059
Thanks to this "tactical voting"
by East Base residents,

00:02:50.059 --> 00:02:55.177
South wins the second round easily,
despite being the least populated.

00:02:55.177 --> 00:02:59.762
Can a system be called fair and good 
if it incentivizes lying

00:02:59.762 --> 00:03:01.712
about your preferences?

NOTE Paragraph

00:03:01.712 --> 00:03:05.511
Maybe what we need to do 
is let voters express a preference

00:03:05.511 --> 00:03:08.676
in every possible head-to-head matchup.

00:03:08.676 --> 00:03:11.671
This is known as the Condorcet method.

00:03:11.671 --> 00:03:15.203
Consider one matchup: 
West versus North.

00:03:15.203 --> 00:03:18.713
All 100 colonists vote on their preference
between the two.

00:03:18.713 --> 00:03:23.516
So that's West's 42 versus 
the 58 from North, South, and East,

00:03:23.516 --> 00:03:25.731
who would all prefer North.

00:03:25.731 --> 00:03:29.066
Now do the same 
for the other five matchups.

00:03:29.066 --> 00:03:32.661
The victor will be whichever base 
wins the most times.

00:03:32.661 --> 00:03:36.622
Here, North wins three 
and South wins two.

00:03:36.622 --> 00:03:40.082
These are indeed the two 
most central locations,

00:03:40.082 --> 00:03:45.659
and North has the advantage of not being 
anyone’s least preferred choice.

NOTE Paragraph

00:03:45.659 --> 00:03:50.846
So does that make the Condorcet method 
an ideal voting system in general?

00:03:50.846 --> 00:03:53.176
Not necessarily.

00:03:53.176 --> 00:03:55.877
Consider an election 
with three candidates.

00:03:55.877 --> 00:04:01.541
If voters prefer A over B, and B over C, 
but prefer C over A,

00:04:01.541 --> 00:04:04.151
this method fails to select a winner.

NOTE Paragraph

00:04:04.151 --> 00:04:08.027
Over the decades, researchers 
and statisticians have come up with

00:04:08.027 --> 00:04:12.057
dozens of intricate ways 
of conducting and counting votes,

00:04:12.057 --> 00:04:14.840
and some have even been 
put into practice.

00:04:14.840 --> 00:04:16.737
But whichever one you choose,

00:04:16.737 --> 00:04:21.508
it's possible to imagine it delivering 
an unfair result.

NOTE Paragraph

00:04:21.508 --> 00:04:25.128
It turns out that our intuitive concept 
of fairness

00:04:25.128 --> 00:04:29.590
actually contains a number of assumptions 
that may contradict each other.

00:04:29.590 --> 00:04:33.910
It doesn’t seem fair for some voters 
to have more influence than others.

00:04:33.910 --> 00:04:38.253
But nor does it seem fair to simply 
ignore minority preferences,

00:04:38.253 --> 00:04:41.419
or encourage people to game the system.

00:04:41.419 --> 00:04:45.453
In fact, mathematical proofs 
have shown that for any election

00:04:45.453 --> 00:04:47.243
with more than two options,

00:04:47.243 --> 00:04:51.023
it’s impossible to design a voting system 
that doesn’t violate

00:04:51.023 --> 00:04:55.513
at least some theoretically
desirable criteria.

00:04:55.513 --> 00:05:00.030
So while we often think of democracy 
as a simple matter of counting votes,

00:05:00.030 --> 00:05:05.463
it’s also worth considering who benefits
from the different ways of counting them.