WEBVTT

00:00:00.000 --> 00:00:05.855
(intro music)

00:00:05.855 --> 00:00:07.370
My name is Laurie Santos.

00:00:07.370 --> 00:00:10.370
I teach psychology at Yale
University, and today

00:00:10.370 --> 00:00:12.549
I want to talk to you about anchoring.

00:00:12.549 --> 00:00:16.119
This lecture is part of a
series on cognitive biases.

00:00:16.119 --> 00:00:19.429
Let's do a math problem.
really quickly, and you've

00:00:19.429 --> 00:00:20.760
gotta do it in your head

00:00:20.760 --> 00:00:21.649
Ready?

00:00:21.649 --> 00:00:27.529
First, multiply the following numbers:
eight times seven times six

00:00:27.529 --> 00:00:32.460
times five times four times three times
two times one.

00:00:32.460 --> 00:00:35.250
OK, that's it.

00:00:35.250 --> 00:00:36.840
What's your guess?

00:00:36.840 --> 00:00:37.840
A thousand?

00:00:37.840 --> 00:00:39.820
Two thousand?

00:00:39.820 --> 00:00:43.030
When the psychologists Danny Kahneman
and Amos Tversky tried this with

00:00:43.030 --> 00:00:45.390
human subjects, subjects on average

00:00:45.390 --> 00:00:47.990
guessed about two thousand
two hundred and fifty.

00:00:47.990 --> 00:00:49.379
Seems like an OK guess.

00:00:49.379 --> 00:00:53.109
But now, let's suppose I gave you
a different math problem.

00:00:53.109 --> 00:00:54.650
What if I gave you this one?

00:00:54.650 --> 00:00:56.050
Ready?

00:00:56.050 --> 00:01:00.030
One times two times three times four

00:01:00.030 --> 00:01:04.770
times five times six times
seven times eight.

00:01:04.770 --> 00:01:06.310
What's your answer?

00:01:06.310 --> 00:01:08.150
If you're like Kahneman and Tversky's

00:01:08.150 --> 00:01:11.040
subjects, your answer might
be a bit different here.

00:01:11.040 --> 00:01:13.880
For this question, their subjects
guessed a lot lower.

00:01:13.880 --> 00:01:17.450
On average they said the answer
was about five hundred and twelve.

00:01:17.450 --> 00:01:19.509
The first amazing thing
about these similar

00:01:19.509 --> 00:01:23.620
mathematical estimates is that people get
the answers really, really wrong.

00:01:23.620 --> 00:01:25.219
In fact, the real answer?

00:01:25.219 --> 00:01:29.339
Well, for both, its forty thousand
three hundred and twenty.

00:01:29.339 --> 00:01:31.939
People are off by an order of magnitude.

00:01:31.939 --> 00:01:35.259
But the second, even more amazing 
thing is that people give

00:01:35.259 --> 00:01:39.600
different answers to the two problems,
even though they're just different ways

00:01:39.600 --> 00:01:42.020
of asking exactly the same question.

00:01:42.020 --> 00:01:44.060
Why do we give completely different answers,

00:01:44.060 --> 00:01:47.079
when the same math problem
is presented differently?

00:01:47.079 --> 00:01:49.500
The answer lies in how we make estimates.

00:01:49.500 --> 00:01:51.590
When you have lots of time to do a math

00:01:51.590 --> 00:01:55.810
problem, like eight times seven times six
times five times four times three times

00:01:55.810 --> 00:01:58.559
two times one, you can multiply all of

00:01:58.559 --> 00:02:01.139
the numbers together and get an exact product.

00:02:01.139 --> 00:02:02.719
But when you have to do the problem

00:02:02.719 --> 00:02:05.279
quickly, you don't really have time to finish.

00:02:05.279 --> 00:02:07.309
So you start with the first numbers.

00:02:07.309 --> 00:02:10.188
You multiply eight times
seven, and get fifty-six.

00:02:10.188 --> 00:02:12.700
And then you've gotta
multiply that by six,

00:02:12.700 --> 00:02:16.670
and, well, you're guessing the final
number's gotta be pretty big, bigger than

00:02:16.670 --> 00:02:19.660
fifty-six, like maybe two thousand or so.

00:02:19.660 --> 00:02:22.450
But when you do the second problem, you start

00:02:22.450 --> 00:02:26.880
with one times two, and, well, that's only
two, and two times three's only six.

00:02:26.880 --> 00:02:28.530
Your answer's gonna be pretty small,

00:02:28.530 --> 00:02:31.310
maybe only like five hundred or so.

00:02:31.310 --> 00:02:33.780
This process of guessing based on the first

00:02:33.780 --> 00:02:36.110
number you see is what's
known as "anchoring."

00:02:36.110 --> 00:02:37.680
The first number we think of

00:02:37.680 --> 00:02:39.730
when we do our estimate is the anchor.

00:02:39.730 --> 00:02:41.830
And once we have an anchor in our head,

00:02:41.830 --> 00:02:44.730
well, we sort of adjust
as needed from there.

00:02:44.730 --> 00:02:48.750
The problem is that our minds are biased
not to adjust as much as we need to.

00:02:48.750 --> 00:02:51.580
The anchors are cognitively really strong.

00:02:51.580 --> 00:02:54.530
In the first, problem you probably
started with fifty-six, and

00:02:54.530 --> 00:02:57.720
then adjusted to an even
bigger number from there.

00:02:57.720 --> 00:03:00.999
And in the second problem, you started
with six, and then adjusted from there.

00:03:00.999 --> 00:03:05.849
The problem is that starting at different
points leads to different final guesses.

00:03:05.849 --> 00:03:10.970
Like real anchors, our estimated anchors
kinda get us stuck in one spot.

00:03:10.970 --> 00:03:14.709
We often fail to drag the anchor far
enough to get to a correct answer.

00:03:14.709 --> 00:03:17.709
Kahneman and Tversky discovered that this

00:03:17.709 --> 00:03:19.889
sort of anchoring bias happens all the time,

00:03:19.889 --> 00:03:22.480
even for anchors that are totally arbitrary.

00:03:22.480 --> 00:03:25.480
For example, they asked
people to spin a wheel with

00:03:25.480 --> 00:03:28.320
numbers from one to a hundred,
and then asked them to estimate

00:03:28.320 --> 00:03:31.739
what percentage of countries in
the United Nations are African.

00:03:31.739 --> 00:03:34.739
People who spun a ten on
the wheel estimated that

00:03:34.739 --> 00:03:36.769
the number was about twenty-five percent.

00:03:36.769 --> 00:03:39.769
But people who spun a
sixty-five estimated that

00:03:39.769 --> 00:03:41.650
the number was forty-five percent.

00:03:41.650 --> 00:03:46.340
In another experiment, Dan Ariely
and his colleagues had people

00:03:46.340 --> 00:03:49.300
write down the last two digits
of their social security number.

00:03:49.300 --> 00:03:50.770
They were then asked whether they would

00:03:50.770 --> 00:03:54.119
pay that amount in dollars
for a nice bottle of wine.

00:03:54.119 --> 00:03:58.250
Ariely and colleagues found that people
in the highest quintile of social security

00:03:58.250 --> 00:04:02.799
numbers would pay three to four times
as much for the exact same good.

00:04:02.799 --> 00:04:04.939
Just setting up a larger anchor can make a

00:04:04.939 --> 00:04:07.099
person who would pay eight
dollars for the bottle

00:04:07.099 --> 00:04:10.649
of wine be willing to spend
twenty-seven dollars instead.

00:04:10.649 --> 00:04:14.619
Sadly for us, sales people use
anchors against us all the time.

00:04:14.619 --> 00:04:18.399
How many times have you noticed
a salesperson or an advertisement

00:04:18.399 --> 00:04:21.100
anchoring you to a particular price, or

00:04:21.100 --> 00:04:23.890
even to how much of a particular
product you should buy?

00:04:23.890 --> 00:04:26.430
Whether it's buying a car, or a sweater,

00:04:26.430 --> 00:04:30.270
or even renting a hotel room, our
intuitions about what prices

00:04:30.270 --> 00:04:34.510
are reasonable to pay often come
from some arbitrary anchor.

00:04:34.510 --> 00:04:38.490
So, the next time you're given an
anchor, take a minute to think.

00:04:38.490 --> 00:04:40.139
Remember what happens when you

00:04:40.139 --> 00:04:42.300
drop your anger too high, and then

00:04:42.300 --> 00:04:45.330
consider thinking of a
very different number.

00:04:45.330 --> 00:04:49.129
It might affect your final estimate
more than you expect.