ADVANCES
16 Scientific American, November 2019 Illustration by Thomas Fuchsappropriate for the problems these deltas
are facing,” he says.
Unlike rocky continental coasts, deltas
are made of soft river sediments that are
deposited over thousands of years and can
easily compact and subside. Subsidence can
grow worse when upstream dams block
the incoming flow of new sediments in riv-
ers or when groundwater or natural gas is
pumped up from below, removing underly-
ing support for the land. Urban infrastruc-
ture can also prevent water from seeping
into the earth and refilling aquifers. All these
forces are at play in the Mekong, which is
subsiding in some areas at rates approach-
ing five centimeters a year—and the rate at
which the entire delta is subsiding is among
the fastest in the world. According to Nguy-
en Hong Quan, a hydrogeologist at Viet-
nam National University, flooding has
grown more common all across the delta.
Numerous international assessments
of deltas are based on topography informa-
tion gathered in February 2000 by the
space shuttle Endeavour. Known as the Shut -
tle Radar Topography Mission, this global
survey was sponsored in part by the U.S.
Department of Defense, and data from the
project are now publicly available. Elevation
assessments use other space-based mea-
surements as well, but in general they are
prone to vertical errors ranging up to 10
meters or more. “Not so bad if you’re mod-
eling the Himalayas,” Törnqvist says. “But
for a low-lying delta, that’s a whole different
story.” Organizations such as the World
Bank rely on these assessments when mak-
ing policy decisions, including where to allo-
cate flood-preparedness resources.
The gold-standard remote-sensing sys-
tem used for measuring delta heights—
lidar, which is often mounted on aircraft—
is accurate to within a few centimeters. But
it is expensive and generally unavailable
in developing countries.
Space shuttle data had put the Mekong’s
average elevation at 2.6 meters. But Min-
derhoud, who was on-site with a Dutch
research team studying the delta, was
skeptical. He found that those measure-
ments had strange elevation patterns that
were inconsistent with the local terrain.
Minderhoud says his Vietnamese col-
leagues knew their government had been
collecting ground-based survey data and
even some lidar measurements. Vietnam-
ese academics, however, had not published
the data in international journals, according
to Minderhoud.
Robert Nicholls, a coastal engineer at
the University of Southampton in England,
says it is not unusual for governments to
withhold topography measurements for
national security reasons. Because thosedata can be used to support strategic mili-
tary operations, “they are not in the public
domain,” Nicholls says. And governments
may simply not want to stir drama among
local populations, Törnqvist notes.
To gain access to the Vietnamese data,
Minderhoud first had to build trust with
government institutions and identify oppor-
tunities for cooperation. “I tried to find out
how my own research might contribute to
their goals,” he says. “The key was to make
this a combined effort.” In time, he wound
up with almost 20,000 elevation points
measured throughout the delta.
Minderhoud’s team also performed a
crucial step that is frequently neglected in
regional assessments: the researchers cali-
brated the data to a local benchmark for
zero elevation at an island town called Hon
Dau. This was necessary because oceanHUMAN BEHAVIORProcrastination
Tech Support
“Cognitive prosthesis” motivates
people to finish tasksChoosing between instant gratification
and future benefit can easily lead to short-
sighted decisions: streaming TV instead of
going to the gym, for example, or scrolling
through social media rather than working
on a challenging project. “Because of this
misalignment between immediate reward
and long-term value, people often struggle
to do what’s best for them in the long run,”
says Falk Lieder, a cognitive scientist at the
Max Planck Institute for Intelligent Sys-
tems in Tübingen, Germany.
To guide individuals toward optimalchoices, Lieder and his colleagues de -
signed a digital tool they call a “cognitive
prosthesis.” It helps to match a decision’s
immediate reward with its long-term
worth—using artificial intelligence to aug-
ment human decision-making through a
to-do list. The researchers de veloped a set
of models and algorithms that consider
various elements such as a list of tasks,an individual’s subjective aversion to each
and the amount of time available. The sys-
tem then assigns reward points to each
task in a way that is customized to encour-
age that person to complete them all.
“The idea was to turn the challenging
projects that people pursue in the real
world into a gamelike environment,” Lieder
says. “The point system [gives] people
proximal, attainable goals that signal that
they’re making progress.”
The team tested the setup in a series
of experiments with human subjects.
The results, published online in August in
Nature Human Behaviour, revealed that the
AI support system helped people make
better, faster decisions and procrastinate
less—and it made them more likely to
complete all the assigned tasks. In one
experiment, in which the researchers pre-
sented 120 participants with a list of sever-
al writing assignments, they found thatSouth China SeaGulf of
ThailandCAMBODIA VIETNAMHo Chi Minh CityMeko
ng
i^ R
ev
rMekongRiver^ Delta0 50
KilometersMap by Mapping Specialists© 2019 Scientific American