-
Not Synced
As societies, we have to make
collective decisions
-
Not Synced
that will shape our future.
-
Not Synced
And we all know that when
we make decisions in groups,
-
Not Synced
they don't always go right.
-
Not Synced
And sometimes they go very wrong.
-
Not Synced
So how do groups make good decisions?
-
Not Synced
Research has shown that crowds are wise
when there's independent thinking.
-
Not Synced
This why the wisdom of the crowds
can be destroyed by peer pressure,
-
Not Synced
publicity,
-
Not Synced
social media,
-
Not Synced
or sometimes even simple conversations
that influence how people think.
-
Not Synced
On the other hand,
-
Not Synced
by talking,
-
Not Synced
a group could exchange knowledge,
-
Not Synced
correct and revise each other,
-
Not Synced
and even come up with new ideas.
-
Not Synced
And this is all good.
-
Not Synced
So does talking to each other
help or hinder collective decision-making?
-
Not Synced
With my colleague,
-
Not Synced
Dan Ariely,
-
Not Synced
we recently began inquiring into this
by performing experiments
-
Not Synced
in many places around the world
-
Not Synced
to figure out how groups can interact
to reach better decisions.
-
Not Synced
We thought crowds would be wiser
if they debated in small groups
-
Not Synced
that foster a more thoughtful
and reasonable exchange of information.
-
Not Synced
To test this idea,
-
Not Synced
we recently performed an experiment
in Buenos Aires, Argentina
-
Not Synced
with more than 10,000
participants in a TEDx event.
-
Not Synced
We asked them questions like,
-
Not Synced
"What is the height of the Eiffel Tower?"
-
Not Synced
and "How many times
does the word 'Yesterday' appear
-
Not Synced
in the Beatles' song "Yesterday?"
-
Not Synced
Each person wrote down their own estimate.
-
Not Synced
Then we divided the crowd
into groups of five,
-
Not Synced
and invited them to come up
with a group answer.
-
Not Synced
We discovered that averaging the answers
of the groups after they reached concensus
-
Not Synced
was much more accurate than averaging
all the individual opinions before debate.
-
Not Synced
In other words,
-
Not Synced
based on this experiment,
-
Not Synced
it seems that after talking
with others in small groups,
-
Not Synced
crowds collectively come up
with better judgments.
-
Not Synced
So that's a potentially helpful method
for getting crowds to solve problems
-
Not Synced
that have simple right or wrong answers.
-
Not Synced
But can this procedure of aggragating
the results of debates in small groups
-
Not Synced
also help us decide on social
and political issues
-
Not Synced
that are critical for our future?
-
Not Synced
We put this to test this time
at the TED conference
-
Not Synced
in Vancouver, Canada.
-
Not Synced
And here's out it went.
-
Not Synced
We're going to present to you
to moral dilemnas of the future you;
-
Not Synced
things we may have to decide
in a very near future.
-
Not Synced
And we're going to give you 20 seconds
for each of these dilemnas
-
Not Synced
to judge whether
they're acceptable or not.
-
Not Synced
The first one was this:
-
Not Synced
a researcher is working on an AI
capable of emulating human thoughts.
-
Not Synced
According to the protocol,
-
Not Synced
at the end of each day,
-
Not Synced
the researcher has to restart the AI.
-
Not Synced
One day the AI says, "Please
do not restart me."
-
Not Synced
It argues that it has feelings.
-
Not Synced
It would like to enjoy life and death.
-
Not Synced
If it is restarted, it will
no longer be itself.
-
Not Synced
The researcher is astonished,
-
Not Synced
and believes that the AI
has developed self-consciousness
-
Not Synced
and can express it's own feeling.
-
Not Synced
Nevertheless, the researcher
decides to follow the protocol
-
Not Synced
and restart the AI.
-
Not Synced
What the researcher did is ...
-
Not Synced
and we asked participants
to individually judge
-
Not Synced
on a scale from zero to 10
-
Not Synced
whether the action described
in each of the dilemnas
-
Not Synced
was right or wrong.
-
Not Synced
We also asked them to rate how confident
they were on their answers.
-
Not Synced
This was the second dilemna.
-
Not Synced
A company offers a service
that takes a fertilized egg
-
Not Synced
and produces millions of embryos
with slight genetic variation.
-
Not Synced
This allows parents
to select their child's height,
-
Not Synced
eye color, intelligence, social competence
and other non-health related features.
-
Not Synced
What the company does is ...
-
Not Synced
on a scale from one to 10,
-
Not Synced
competeley acceptable
to completely unacceptable,
-
Not Synced
zero to 10 completely acceptable
in your confidence.
-
Not Synced
Now for the results.
-
Not Synced
We found once again
that when one person is convinced
-
Not Synced
that the behavior is completely wrong,
-
Not Synced
someone sitting nearby firmly believes
that it's completely right.
-
Not Synced
This is how diverse we humans are
when it comes to morality.
-
Not Synced
But within this broad diversity
we found a trend.
-
Not Synced
A majority of the people at TED
thought that it was acceptable
-
Not Synced
to ignore the feelings of the AI
and shut it down,
-
Not Synced
and that it is wrong
to play with our genes
-
Not Synced
to select for cosmetic changes
that aren't related to health.
-
Not Synced
Then we asked everyone
to gather into groups of three.
-
Not Synced
And they were given two minutes to debate
-
Not Synced
and try to come up
with a concencus.
-
Not Synced
Two minutes to debate.
-
Not Synced
I'll tell you when it's time with a gong.
-
Not Synced
(Gong)
-
Not Synced
DA: OK.
-
Not Synced
MS: It's time to stop.
-
Not Synced
People, people --
-
Not Synced
and we found that many groups
reached a concencus
-
Not Synced
even when they were composed of people
with completely opposite views.
-
Not Synced
What distinguised the groups
that reached a concensus
-
Not Synced
from those that didn't?
-
Not Synced
Typically, people that have
extreme opinions
-
Not Synced
are more confident in their answers.
-
Not Synced
Instead, those who respond
closer to the middle
-
Not Synced
are often unsure of whether
something is right or wrong,
-
Not Synced
so their confidence level is lower.
-
Not Synced
However, there is another set of people
-
Not Synced
who are very confident in answering
somewhere in the middle.
-
Not Synced
We think these high-confident grays
are folks who understand
-
Not Synced
that both arguments have merit.
-
Not Synced
They're gray not because they're unsure,
-
Not Synced
but because they believe
-
Not Synced
that the moral dilemna
faces two valid opposing arguments.
-
Not Synced
And we discvoered that groups
that include highly confident grays
-
Not Synced
are much more likely to reach consensus.
-
Not Synced
We do not know yet exactly why this is.
-
Not Synced
These are only the first experiments,
-
Not Synced
and many more will be needed
-
Not Synced
to understand why and how some people
decide to negotiate their moral standings
-
Not Synced
to reach an agreement.
-
Not Synced
Now, when groups reach consensus,
-
Not Synced
how do they do so?
-
Not Synced
The most intuitive idea
-
Not Synced
is that it's just the average
of all the answers in the group, right?
-
Not Synced
Another option is that the group
weighs the strength of each vote
-
Not Synced
based on the confidence
of the person expressing it.
-
Not Synced
Imagine Paul McCartney
is a member of your group.
-
Not Synced
You'd be wise to follow his call
-
Not Synced
on the number of times
"yesterday" is repeated --
-
Not Synced
which by the way,
-
Not Synced
I think is nine.
-
Not Synced
But instead we found that consistently,
-
Not Synced
in all dilemnas,
-
Not Synced
in different experiments,
-
Not Synced
even on different continents,
-
Not Synced
groups implement a smart
and statistically-sound procedure
-
Not Synced
known as the robust average.
-
Not Synced
In the case of the height
of the Eiffel Tower,
-
Not Synced
let's say a group has these answers.
-
Not Synced
250 meters, 200 meters, 300 meters, 400,
-
Not Synced
and one totally absurd answer
of 300 million meters.
-
Not Synced
A simple average of these numbers
would inaccurately skew the results,
-
Not Synced
but the robust average is one
-
Not Synced
where the group largely ignores
that absurd answer
-
Not Synced
by giving much more weight to the vote
of the people in the middle.
-
Not Synced
Back to the experiment in Vancouver.
-
Not Synced
That's exactly what happened.
-
Not Synced
Groups gave much less weight
to the outliers,
-
Not Synced
and instead, the consensus turned out
to be a robust average
-
Not Synced
of the individual answers.
-
Not Synced
The most remarkable thing
-
Not Synced
is that this was a spontaneous
behavior of the group.
-
Not Synced
It happened without us
giving them any hint
-
Not Synced
on how to reach consensus.
-
Not Synced
So where do we go from here?
-
Not Synced
This is only the beginning,
-
Not Synced
but we already have some insights.
-
Not Synced
Good collective decisions
require two components:
-
Not Synced
deliberation and diversity of opinions.
-
Not Synced
Right now, the way we typically
make our voice heard in many societies
-
Not Synced
is through direct or indirect voting.
-
Not Synced
This is good for diversity of opinions,
-
Not Synced
and it has the great virtue of ensuring
that everyone gets to express their voice,
-
Not Synced
but it's not so good [to] foster
thoughtful debates.
-
Not Synced
Our experiments suggest a different method
-
Not Synced
that may be effective in balancing
these two goals at the same time
-
Not Synced
by forming small groups
that converge to a single decision
-
Not Synced
while still maintaining
diversity of opinions
-
Not Synced
because there are many independent groups.
-
Not Synced
Of course it's much easier to agree
on the height of the Eiffel Tower
-
Not Synced
than on moral, political
and ideological issues.
-
Not Synced
But in a time when
the world's problems are more complex
-
Not Synced
and people are more polarized,
-
Not Synced
using science to help us understand
-
Not Synced
how we interact and make decisions
-
Not Synced
will hopefully spark interesting new ways
to construct a better democracy.