[♪INTRO]
Let's be honest: No one really enjoys airport security.
It's annoying, it can take forever,
and it seems like the rules are always changing.
People tell pollsters that they'd fly more if security wasn't such a hassle,
but most people also say they're fine sacrificing some privacy for more security.
But those screening areas where no one knows what to do with their shoes
are only one small part of it all.
Every airport is an onion of security measures that all work together,
and those layers are constantly shifting and evolving.
Over the last 10 years or so, at least in the United States,
some of that evolution has been inspired by game theorists.
These researchers have been helping decide where security should be,
how much there should be, and how often it should change.
And one common response to all those questions has been: be random.
Game theory is a broad subject that sits in between math and economics.
It originally focused on how people play games, like the name suggests.
So a game theorist might work out when to bluff in poker,
or why your friend Camille always hoards sheep in Catan.
But while some researchers still do study actual games,
the field has expanded far beyond those early roots.
Plenty of today's game theorists are looking at other situations
where intelligent actors (like groups of people)
interact with each other and make decisions.
So things like soldiers making choices on the battlefield,
businesses competing against each other, or even election results.
Or Braess's Paradox from our video about reducing traffic.
That's straight out of game theory.
Adding more roads can make traffic worse, not better.
Because we all try to make our own trip faster,
which ends up making everyone's drive slower, including our own.
Governments and businesses also bring in game theorists
to help with crucial decisions or to build new systems.
Like, imagine that you run an international airport in a large city,
and you've stopped a few people
who were planning to attack the airport in the past.
So you want police watching for anything suspicious,
you want dogs sniffing for anything suspicious,
and you want security checking people's belongings…
for anything suspicious.
But you only have a certain amount of money,
a certain number of officers and dogs,
and you don't want everyone to completely hate your airport
because security takes five hours.
So what do you do?
You call a game theorist.
At least, that's what they did in LA back in 2007.
Now, to be totally clear, whether security screenings and the TSA
are the most effective way to find or stop threats
is a completely different question
from the one these researchers are trying to answer.
So the goal, both for these researchers and in this video,
isn't to dig into the science or sociology of security protocols in general.
We're just going to explain where game theory has been implemented
at airports so far, and what the reported results and criticisms have been.
And this all really started at LAX, the Los Angeles International Airport.
The Los Angeles World Airport Police
wanted checkpoints along the roads to LAX,
but there weren't enough police to cover every entrance every day.
So LAX contacted a professor named Milind Tambe
and asked what they should do.
Together, they developed a computer program called ARMOR,
short for Assistant for Randomized Monitoring Over Routes.
It generates recommended security schedules for the airport.
ARMOR is based on a kind of game theory called Stackelberg game theory.
In Stackelberg games, one person makes a move while the second watches,
and then the second has to respond to their move.
The model is usually applied
to businesses competing against each other to sell something.
One business goes first and charges what they want for, say, a T-shirt.
Then other businesses have to respond and choose what they want to charge
for their own cool, nerdy T-shirts
so that people still want to buy them,
but they're not so cheap that the business loses money.
But since the first business knows
that others are going to respond to their initial decision,
that changes what the first business does at the beginning.
So before anyone does anything,
you end up with a sort of
"they know that we know that they know that we know…" kind of situation.
Tambe and other researchers have realized
that Stackelberg games can also apply to security situations.
The first player is the airport, setting up their security in a certain way.
The second player is someone trying to attack the airport
or even just bring something dangerous through security.
Each player gets to see what the other does
attackers can go to the airport and look at the security,
and security can monitor attackers whenever they catch something.
So based on the optimal strategy in certain kinds of Stackelberg games,
Tambe and the rest of the researchers who were asked for input
said that the solution to LAX's problem was more randomness.
If they couldn't cover every entrance with police every single day,
they also shouldn't have a regular schedule for which ones are guarded.
Like, if the airport had 6 entrances,
you shouldn't have police checkpoints at entrances 1, 2, and 3 on Day 1,
then 2, 3, and 4 on Day 2, then 3, 4, and 5 on Day 3.
Because it's not too hard to figure out
that on Day 107, entrances 5, 6, and 1 will be covered,
and entrances 2, 3, and 4 won't be.
The same goes for paths taken by police dogs patrolling the airport.
If they regularly followed the same route,
you could predict where they'd be at a certain time of day.
In these non-random cases, with the right information, someone could
sneak through an unguarded door or know they won't run into a dog.
Even if the schedules seem complex
to someone who's not paying close attention,
if they have a structure, they can probably be exploited.
A key element of ARMOR is also that
humans shouldn't try to put randomness into the schedule ourselves.
That's not enough.
Because no matter how "LOL SO RANDOM" we think we are,
humans just aren't good at producing or identifying random sequences.
Our brains just like patterns too much.
One of the simplest examples is that
if researchers have people imagine a coin flip,
about 80% will say "heads" came up.
Because when we talk about coin flips, we usually say "heads" first.
Seriously.
So instead, Tambe and the other researchers designed ARMOR to produce
a mathematically random schedule, without a discernible pattern in it.
The program was also flexible enough to adjust based on new information
or pressing security concerns.
Like, if a certain entrance will be way busier than usual one day,
you'd probably want more security there than usual.
After six months of ARMOR,
the airport police were very positive about the program.
A few potential threats, like cars carrying weapons,
were arrested by officers scheduled by ARMOR.
The new random schedule also reportedly made police seem more present
at the airport,
and the program reduced the scheduling work of the people in charge
letting them do other work, instead.
ARMOR was successful enough at LAX that soon,
Tambe was one of a number of researchers hired by the TSA
to apply game theory to airports across the US.
There are hundreds of airports nationwide, and they're all different.
Each one has its own baggage area and ticket counters
and terminals and security desks
so you can't have the exact same security strategy at each one.
For one thing, that would be super expensive.
The TSA's budget might be large, but it's not that large.
Plus, it's just not necessary.
Some airports are more threatened than others.
That doesn't mean the whole budget should just be thrown at large airports
like LAX, either, though.
Because leaving the rest without security is a… bad idea.
The team's first step was a program called IRIS,
which stands for Intelligent Randomization In Scheduling.
It adapted the kinds of Stackelberg games used for ARMOR to schedule
where Federal Air Marshals,
a kind of in-flight security and police force,
should be flying.
Then, they developed GUARDS:
Game-theoretic Unpredictable and Randomly Deployed Security.
Clearly, they really like a good acronym.
With GUARDS, they didn't just look at handling one thing at a time,
like which roads to secure or which Air Marshals to put onto which flights.
GUARDS is focused on the upper layer of organization:
Where and how the TSA should monitor each airport
so that they're all as safe as possible, without any predictable patterns.
This means that GUARDS also had to take different security measures
and types of attackers into account.
Someone aiming for the ticket window likely has different goals
and strategies than someone targeting the planes.
Since these game theory-based approaches seem to have been effective
so far, we'll probably keep seeing more of them.
But that doesn't mean they're perfect, either.
GUARDS isn't designed to handle all the nuances of the real world,
and it can't dictate how every nook and cranny of every airport
in the country should be run.
It's still just a computer program running with a set of variables,
and can't include every possible threat in its scheduling calculations.
And to schedule well, each threat that it does handle has to be rated
like, essentially on a scale of 0-10.
Which obviously has its problems and subjectivity.
But in 2011, GUARDS was submitted to the TSA to be tested
at an undisclosed airport.
And if it was successful,
the TSA planned to incorporate the program into their scheduling elsewhere.
Although since then, they've been pretty quiet about what happened.
Which is… understandable.
That's a security measure too.
Game-theoretic approaches to security also don't stop at the airport.
Tambe and other researchers have branched out
and developed a program called PROTECT,
or Port Resilience Operational/Tactical Enforcement to Combat Terrorism.
And it helps protect the port of Boston.
Ports are huge, open spaces; they're dealing with boats
and they're not contained in the same way that airports are.
Which means that the techniques that worked in an airport
might not work in a port, and PROTECT accounts for that.
According to the researchers, it uses a more accurate model than ARMOR
or GUARDS for how people make decisions, and it includes more information
about which points are more important to watch than others.
After recent incidents, researchers have also proposed ways
to use game theory to protect public spaces during events.
But that's an even harder problem!
You have another big space, but it's crowded with people
and there are a near-infinite number of things that could happen.
They've even proposed using game theory to address something
that might seem completely unrelated: protecting animals from poaching.
Illegal poaching, fishing, and logging are huge environmental problems,
but they keep happening because it's impossible to watch everything at once.
By treating poachers like opponents in certain kinds of games,
and looking at where they've gone so far,
game theorists can try and predict where they'll strike next.
Plus, poaching, fishing, and logging are all much more common
than any other type of attack we've talked about so far.
That's not a good thing, but it does mean that after game theorists
have made predictions, it's easier to check what actually happened.
That lets them see whether their models do a better job of protecting
lives than what we're doing now, and update them to do better in the future.
So whether it's airports or wildlife preserves,
game theorists are out there trying to make the world better for us
and all of our sheep-hoarding, Catan-playing friends.
Thanks for watching this episode of SciShow!
If you want to learn more about game theory
and the science of decision making,
we have a whole video explaining some of the basics.
And if you want to keep up with all sorts of deep dives into science topics,
you can click subscribe.
[♪OUTRO]
No comments:
Post a Comment