Narration: I’m standing in an empty desert-like space made up of nothing but brown dirt
and blue sky that meet at a far off horizon. A green pole appears several yards away and I
walk over to it and step into it. I turn around, and a few seconds later a crowd of people
appear around me. They’re dressed in leather jackets and business suits. We stand for a
moment, all staring straight ahead, and then the people start walking.
[Music]
I’m in the VENLab, Brown’s virtual reality lab, testing out a simulation. I ended up in
the VENLab pretty randomly. I had seen their posters asking for participants, and I wanted
to know what kind of research they were doing there, what virtual reality would be useful
for. That led me to Bill Warren, a professor of Cognitive, Linguistic, and Psychological
Sciences. Bill had me think about some questions that were actually interesting in how tiny
and almost un-interesting they sound to anyone who isn’t looking past the surface. For
example, walking: the question of how we walk places. Every day you walk to class, walk
down sidewalks, move through crowded cafes, and ninety percent of the time you get
where you wanna go and you don’t bump into anything. Walking places without much
trouble is one of our most basic abilities, so most people take it for granted. Bill Warren,
does not.
Bill: There are all these mysteries around, you know, all the time in just your mundane
ordinary behavior: how do I pick up a glass without knocking it over, right how do I walk
down the street? They’re the deepest problems that we’re facing, I think.
[End Music]
Narration: Bill is the head of research at the VENLab. In the lab, people put on virtual
reality goggles, which look like ski-goggles with a screen the size of a small book where the
lenses would be, and they walk through virtual worlds inside of what is in reality a big
empty room that the lab uses for this. While they walk, a grid in the ceiling tracks their
movements down to the millimeter. The researchers have people walk around obstacles,
toward goals, and with other virtual people and by tracking people’s movements they’re
trying to answer Bill’s question: how do we walk down the street? [Music] So why is that
such an important question?
Bill: The problems I’m working on are driven by these sort of philosophical concerns I have
about what is the mind, where does the organization in behavior come from? But I can’t do
experiments on the philosophical question all at once, I gotta pick little problems, and
there’s no free lunch, there has to be an explanation, you can’t just say, oh, the pattern is
there because, you know, Aristotle put it there, right, we can’t just assume we have all of
this knowledge in our heads, that came from nowhere, we gotta account for it.
Narration: So if I asked you how you walk down the street without bumping into people,
you might say that when people get in your way, you stop or go around them. But what’s
going on in your brain to tell you when you need to start slowing down to avoid hitting
someone, or which side you should pass them on? And the question isn’t just what makes
you capable of walking down the street, but also what it looks like when you do, how
people actually act.
[End Music]
The VENLab researchers want to make a computer model that can simulate the way
humans move through complex environments; that can predict the exact path you would
take to walk across a room. The technology they develop will help robots navigate fluidly
and improve simulated tests of evacuation plans for big buildings. [Music] But it will also
add another item to the increasingly long list of ways that computers can mimic us.
In order to create their model, the researchers design different kinds of simulations,
have people walk around in them, and then change tiny aspects of the simulation to see
how the people’s walking patterns change. So in the simulation I did, I walked with a crowd
back and forth on what seemed to me to be the exact same path every time. But according
to the grad student who was helping me, their close measurements pick up tiny changes in
my path.
[End Music]
Greg: If we look at the minute measurements of what you do we see very, very small
differences in behavior, so for example if a couple of the people sort of turn to right, you
will turn to the right, you might not even be perceptually aware that you’re doing it.
Narration: So if a few people in a crowd I was walking with shifted their path, even in such
a small way that I didn’t consciously notice a difference, I would subconsciously adjust my
path to follow.
Bill: We can write a little equation that says if your neighbor changes in this way then the
participant changes in this particular way. And that equation is gonna be our model of how
an individual behaves.
Narration: Putting these observations and equations together, Bill and the researchers have
come up with a set of rules of movement. Rules that can predict human behavior, rules that
you follow while walking without even knowing it. The rules cover all sorts of tiny
behaviors: how we match the speed and direction of people we’re following, how we
calculate our movements to intercept another moving person, how we decide what path to
take to avoid a tree and get to a picnic table on the other side.
Bill: Interestingly we base these models on springs… So if you’re trying to get to a goal it’s
as though your walking direction is attached to the direction of the goal by a spring. If
there’s an obstacle in the way, you’re attached to that direction of the obstacle by another
spring.
[Music]
Narration: So imagine you’re standing in your kitchen and you want to get to your fridge on
the other side of the room, that’s your goal. You are attached to your fridge by a giant
imaginary spring.
Bill: and its as though that goal is kind of pulling me over so I wanna walk straight toward
that goal.
Narration: But there is a table in the way of the straightest path, an obstacle.
Bill: Obstacles are just opposite, right, so its like you don’t want to hit this thing, you know,
its kind of pushing your direction of travel away from it.
Narration: So the obstacle spring, a pushing spring, pushes you to go around the table, and
the goal spring stretches to let you, but as soon as you get around it you’re pulled back
toward the straightest path to the fridge.
Bill: Your own behavior is basically the resultant of all of those spring forces that are acting
on you at any given time.
Narration: The VENLab team has created a computer simulation that can predict the exact
path a person will take to walk through a scene. It can predict human behavior.
Bill: As the world gets more complicated, there are more possible paths, of course. But if I
know the direction you’re initially going, then I can take it from there and predict how
you’re gonna avoid each obstacle. We can predict the vast majority of trials in this way.
[Music]
Narration: It brings up the question of how unique we are, how much of our actions we
really choose, and how we different we are from computers if computers can mimic us.
Bill: Now if you knew I was trying to predict what you did and you didn’t want me to be
able to predict you, you could always do something else. I’m not trying to say that people
are deterministic machines. It’s just that much of the time we’re organizing our movement
in a fairly efficient and effective way.
Narration: But most of the time, put someone in a room, give them a goal to walk to, watch
them take their first step, and you can predict the entire rest of their path.
[Music for another 20 seconds, and fade out]