9:59:59.000,9:59:59.000
As societies, we have to make[br]collective decisions

9:59:59.000,9:59:59.000
that will shape our future.

9:59:59.000,9:59:59.000
And we all know that when[br]we make decisions in groups,

9:59:59.000,9:59:59.000
they don't always go right.

9:59:59.000,9:59:59.000
And sometimes they go very wrong.

9:59:59.000,9:59:59.000
So how do groups make good decisions?

9:59:59.000,9:59:59.000
Research has shown that crowds are wise[br]when there's independent thinking.

9:59:59.000,9:59:59.000
This why the wisdom of the crowds[br]can be destroyed by peer pressure,

9:59:59.000,9:59:59.000
publicity,

9:59:59.000,9:59:59.000
social media,

9:59:59.000,9:59:59.000
or sometimes even simple conversations[br]that influence how people think.

9:59:59.000,9:59:59.000
On the other hand,

9:59:59.000,9:59:59.000
by talking,

9:59:59.000,9:59:59.000
a group could exchange knowledge,

9:59:59.000,9:59:59.000
correct and revise each other,

9:59:59.000,9:59:59.000
and even come up with new ideas.

9:59:59.000,9:59:59.000
And this is all good.

9:59:59.000,9:59:59.000
So does talking to each other[br]help or hinder collective decision-making?

9:59:59.000,9:59:59.000
With my colleague,

9:59:59.000,9:59:59.000
Dan Ariely,

9:59:59.000,9:59:59.000
we recently began inquiring into this[br]by performing experiments

9:59:59.000,9:59:59.000
in many places around the world

9:59:59.000,9:59:59.000
to figure out how groups can interact[br]to reach better decisions.

9:59:59.000,9:59:59.000
We thought crowds would be wiser[br]if they debated in small groups

9:59:59.000,9:59:59.000
that foster a more thoughtful[br]and reasonable exchange of information.

9:59:59.000,9:59:59.000
To test this idea,

9:59:59.000,9:59:59.000
we recently performed an experiment[br]in Buenos Aires, Argentina

9:59:59.000,9:59:59.000
with more than 10,000[br]participants in a TEDx event.

9:59:59.000,9:59:59.000
We asked them questions like,

9:59:59.000,9:59:59.000
"What is the height of the Eiffel Tower?"

9:59:59.000,9:59:59.000
and "How many times[br]does the word 'Yesterday' appear

9:59:59.000,9:59:59.000
in the Beatles' song "Yesterday?"

9:59:59.000,9:59:59.000
Each person wrote down their own estimate.

9:59:59.000,9:59:59.000
Then we divided the crowd[br]into groups of five,

9:59:59.000,9:59:59.000
and invited them to come up[br]with a group answer.

9:59:59.000,9:59:59.000
We discovered that averaging the answers[br]of the groups after they reached concensus

9:59:59.000,9:59:59.000
was much more accurate than averaging[br]all the individual opinions before debate.

9:59:59.000,9:59:59.000
In other words,

9:59:59.000,9:59:59.000
based on this experiment,

9:59:59.000,9:59:59.000
it seems that after talking[br]with others in small groups,

9:59:59.000,9:59:59.000
crowds collectively come up[br]with better judgments.

9:59:59.000,9:59:59.000
So that's a potentially helpful method[br]for getting crowds to solve problems

9:59:59.000,9:59:59.000
that have simple right or wrong answers.

9:59:59.000,9:59:59.000
But can this procedure of aggragating[br]the results of debates in small groups

9:59:59.000,9:59:59.000
also help us decide on social[br]and political issues

9:59:59.000,9:59:59.000
that are critical for our future?

9:59:59.000,9:59:59.000
We put this to test this time[br]at the TED conference

9:59:59.000,9:59:59.000
in Vancouver, Canada.

9:59:59.000,9:59:59.000
And here's out it went.

9:59:59.000,9:59:59.000
We're going to present to you[br]to moral dilemnas of the future you;

9:59:59.000,9:59:59.000
things we may have to decide[br]in a very near future.

9:59:59.000,9:59:59.000
And we're going to give you 20 seconds[br]for each of these dilemnas

9:59:59.000,9:59:59.000
to judge whether[br]they're acceptable or not.

9:59:59.000,9:59:59.000
The first one was this:

9:59:59.000,9:59:59.000
a researcher is working on an AI[br]capable of emulating human thoughts.

9:59:59.000,9:59:59.000
According to the protocol,

9:59:59.000,9:59:59.000
at the end of each day,

9:59:59.000,9:59:59.000
the researcher has to restart the AI.

9:59:59.000,9:59:59.000
One day the AI says, "Please[br]do not restart me."

9:59:59.000,9:59:59.000
It argues that it has feelings.

9:59:59.000,9:59:59.000
It would like to enjoy life and death.

9:59:59.000,9:59:59.000
If it is restarted, it will[br]no longer be itself.

9:59:59.000,9:59:59.000
The researcher is astonished,

9:59:59.000,9:59:59.000
and believes that the AI[br]has developed self-consciousness

9:59:59.000,9:59:59.000
and can express it's own feeling.

9:59:59.000,9:59:59.000
Nevertheless, the researcher[br]decides to follow the protocol

9:59:59.000,9:59:59.000
and restart the AI.

9:59:59.000,9:59:59.000
What the researcher did is ...

9:59:59.000,9:59:59.000
and we asked participants[br]to individually judge

9:59:59.000,9:59:59.000
on a scale from zero to 10

9:59:59.000,9:59:59.000
whether the action described[br]in each of the dilemnas

9:59:59.000,9:59:59.000
was right or wrong.

9:59:59.000,9:59:59.000
We also asked them to rate how confident[br]they were on their answers.

9:59:59.000,9:59:59.000
This was the second dilemna.

9:59:59.000,9:59:59.000
A company offers a service[br]that takes a fertilized egg

9:59:59.000,9:59:59.000
and produces millions of embryos[br]with slight genetic variation.

9:59:59.000,9:59:59.000
This allows parents[br]to select their child's height,

9:59:59.000,9:59:59.000
eye color, intelligence, social competence[br]and other non-health related features.

9:59:59.000,9:59:59.000
What the company does is ...

9:59:59.000,9:59:59.000
on a scale from one to 10,

9:59:59.000,9:59:59.000
competeley acceptable[br]to completely unacceptable,

9:59:59.000,9:59:59.000
zero to 10 completely acceptable[br]in your confidence.

9:59:59.000,9:59:59.000
Now for the results.

9:59:59.000,9:59:59.000
We found once again[br]that when one person is convinced

9:59:59.000,9:59:59.000
that the behavior is completely wrong,

9:59:59.000,9:59:59.000
someone sitting nearby firmly believes[br]that it's completely right.

9:59:59.000,9:59:59.000
This is how diverse we humans are[br]when it comes to morality.

9:59:59.000,9:59:59.000
But within this broad diversity[br]we found a trend.

9:59:59.000,9:59:59.000
A majority of the people at TED[br]thought that it was acceptable

9:59:59.000,9:59:59.000
to ignore the feelings of the AI[br]and shut it down,

9:59:59.000,9:59:59.000
and that it is wrong[br]to play with our genes

9:59:59.000,9:59:59.000
to select for cosmetic changes[br]that aren't related to health.

9:59:59.000,9:59:59.000
Then we asked everyone[br]to gather into groups of three.

9:59:59.000,9:59:59.000
And they were given two minutes to debate

9:59:59.000,9:59:59.000
and try to come up[br]with a concencus.

9:59:59.000,9:59:59.000
Two minutes to debate.

9:59:59.000,9:59:59.000
I'll tell you when it's time with a gong.

9:59:59.000,9:59:59.000
(Gong)

9:59:59.000,9:59:59.000
DA: OK.

9:59:59.000,9:59:59.000
MS: It's time to stop.

9:59:59.000,9:59:59.000
People, people --

9:59:59.000,9:59:59.000
and we found that many groups[br]reached a concencus

9:59:59.000,9:59:59.000
even when they were composed of people[br]with completely opposite views.

9:59:59.000,9:59:59.000
What distinguised the groups[br]that reached a concensus

9:59:59.000,9:59:59.000
from those that didn't?

9:59:59.000,9:59:59.000
Typically, people that have[br]extreme opinions

9:59:59.000,9:59:59.000
are more confident in their answers.

9:59:59.000,9:59:59.000
Instead, those who respond[br]closer to the middle

9:59:59.000,9:59:59.000
are often unsure of whether[br]something is right or wrong,

9:59:59.000,9:59:59.000
so their confidence level is lower.

9:59:59.000,9:59:59.000
However, there is another set of people

9:59:59.000,9:59:59.000
who are very confident in answering[br]somewhere in the middle.

9:59:59.000,9:59:59.000
We think these high-confident grays[br]are folks who understand

9:59:59.000,9:59:59.000
that both arguments have merit.

9:59:59.000,9:59:59.000
They're gray not because they're unsure,

9:59:59.000,9:59:59.000
but because they believe

9:59:59.000,9:59:59.000
that the moral dilemna[br]faces two valid opposing arguments.

9:59:59.000,9:59:59.000
And we discvoered that groups[br]that include highly confident grays

9:59:59.000,9:59:59.000
are much more likely to reach consensus.

9:59:59.000,9:59:59.000
We do not know yet exactly why this is.

9:59:59.000,9:59:59.000
These are only the first experiments,

9:59:59.000,9:59:59.000
and many more will be needed

9:59:59.000,9:59:59.000
to understand why and how some people[br]decide to negotiate their moral standings

9:59:59.000,9:59:59.000
to reach an agreement.

9:59:59.000,9:59:59.000
Now, when groups reach consensus,

9:59:59.000,9:59:59.000
how do they do so?

9:59:59.000,9:59:59.000
The most intuitive idea

9:59:59.000,9:59:59.000
is that it's just the average[br]of all the answers in the group, right?

9:59:59.000,9:59:59.000
Another option is that the group[br]weighs the strength of each vote

9:59:59.000,9:59:59.000
based on the confidence[br]of the person expressing it.

9:59:59.000,9:59:59.000
Imagine Paul McCartney[br]is a member of your group.

9:59:59.000,9:59:59.000
You'd be wise to follow his call

9:59:59.000,9:59:59.000
on the number of times[br]"yesterday" is repeated --

9:59:59.000,9:59:59.000
which by the way,

9:59:59.000,9:59:59.000
I think is nine.

9:59:59.000,9:59:59.000
But instead we found that consistently,

9:59:59.000,9:59:59.000
in all dilemnas,

9:59:59.000,9:59:59.000
in different experiments,

9:59:59.000,9:59:59.000
even on different continents,

9:59:59.000,9:59:59.000
groups implement a smart[br]and statistically-sound procedure

9:59:59.000,9:59:59.000
known as the robust average.

9:59:59.000,9:59:59.000
In the case of the height[br]of the Eiffel Tower,

9:59:59.000,9:59:59.000
let's say a group has these answers.

9:59:59.000,9:59:59.000
250 meters, 200 meters, 300 meters, 400,

9:59:59.000,9:59:59.000
and one totally absurd answer[br]of 300 million meters.

9:59:59.000,9:59:59.000
A simple average of these numbers[br]would inaccurately skew the results,

9:59:59.000,9:59:59.000
but the robust average is one

9:59:59.000,9:59:59.000
where the group largely ignores[br]that absurd answer

9:59:59.000,9:59:59.000
by giving much more weight to the vote[br]of the people in the middle.

9:59:59.000,9:59:59.000
Back to the experiment in Vancouver.

9:59:59.000,9:59:59.000
That's exactly what happened.

9:59:59.000,9:59:59.000
Groups gave much less weight[br]to the outliers,

9:59:59.000,9:59:59.000
and instead, the consensus turned out[br]to be a robust average

9:59:59.000,9:59:59.000
of the individual answers.

9:59:59.000,9:59:59.000
The most remarkable thing

9:59:59.000,9:59:59.000
is that this was a spontaneous[br]behavior of the group.

9:59:59.000,9:59:59.000
It happened without us[br]giving them any hint

9:59:59.000,9:59:59.000
on how to reach consensus.

9:59:59.000,9:59:59.000
So where do we go from here?

9:59:59.000,9:59:59.000
This is only the beginning,

9:59:59.000,9:59:59.000
but we already have some insights.

9:59:59.000,9:59:59.000
Good collective decisions[br]require two components:

9:59:59.000,9:59:59.000
deliberation and diversity of opinions.

9:59:59.000,9:59:59.000
Right now, the way we typically[br]make our voice heard in many societies

9:59:59.000,9:59:59.000
is through direct or indirect voting.

9:59:59.000,9:59:59.000
This is good for diversity of opinions,

9:59:59.000,9:59:59.000
and it has the great virtue of ensuring[br]that everyone gets to express their voice,

9:59:59.000,9:59:59.000
but it's not so good [to] foster[br]thoughtful debates.

9:59:59.000,9:59:59.000
Our experiments suggest a different method

9:59:59.000,9:59:59.000
that may be effective in balancing[br]these two goals at the same time

9:59:59.000,9:59:59.000
by forming small groups[br]that converge to a single decision

9:59:59.000,9:59:59.000
while still maintaining[br]diversity of opinions

9:59:59.000,9:59:59.000
because there are many independent groups.

9:59:59.000,9:59:59.000
Of course it's much easier to agree[br]on the height of the Eiffel Tower

9:59:59.000,9:59:59.000
than on moral, political[br]and ideological issues.

9:59:59.000,9:59:59.000
But in a time when[br]the world's problems are more complex

9:59:59.000,9:59:59.000
and people are more polarized,

9:59:59.000,9:59:59.000
using science to help us understand

9:59:59.000,9:59:59.000
how we interact and make decisions

9:59:59.000,9:59:59.000
will hopefully spark interesting new ways[br]to construct a better democracy.