that I actually kind of dreaded.

You have your picture taken that's the way

it's gonna be in the yearbook, forever.

Little did I know that I was actually taking

part in this ritual that I would research.

(uplifting music)

- [Announcer] What if you could write a formula

so advanced that you could wade through millions

of yearbook photos to track social trends.

Researchers at UC Berkeley are helping

historians do just that.

- This project is really about history.

We looked for a data set

that would kind of capture changes

in people and we found that

we could use high school yearbooks

because yearbooks have existed in high schools

in America from about 1902

and throughout this entire time the format

has stayed exactly the same.

You go to school and you dress up nice

and you stand in front of the camera

and you have your picture taken.

So there's millions, literally, of photographs

that are basically all the same.

One of the things we were able to create

was an average image or a composite

of what the average teen looked like

in the 20th century.

We could just measure, literally,

the curvature of the smile.

There's basically one big up curve

until the second world war and then there

are a couple of dips but it keeps going up.

And for every decade we found

a bunch of subgroups.

In the 1960s, we found for girls

all these different hairstyles.

There's the flip, that goes like this.

Then there's the beehive

which is this big bubble thing.

And then in the '70s we get the afros

and the bouffants and all kinds

of stuff like that.

The data kind of speaks to us.

And things emerge from the data.

All kinds of pattern or trend without

needing any human annotation.

And this is a really big deal because

mostly big data is just really a bunch of noise.

If we can find ways to automatically tease

these things out, we, as humans

will have much less work in order to be able

to understand what we see.

(enthusiastic music)