I’ve known for several months now that I wanted to work on a
project related to climate change during this course. Thwink.org
notes that, for difficult problems, one must take a structured
approach—addressing the challenge intuitively will fail (n.d.). This is
clearly the case when it comes to climate change—untangling the
impacts of this global phenomenon, never mind finding solutions to
adapt to or mitigate them, is the definition of difficult. There are
simply too many factors and non-linear relationships among them
to operate on an intuitive basis, and to try would result in
inappropriately narrowing the solution set to what we think it
should or could be. A colleague of mine who is an expert is
cognitive science has called on coworkers to “resist the urge to
converge” on the solution space, spending more time in the
divergent space, considering the problem from multiple angles.
The CRISP-DM approach to problem solving is particularly relevant
here as the kind of structured approach to use. Meta S. Brown
describes the first phase (developing Business Understanding) tasks
as identifying goals, assessing the situation, define data mining
goals, and develop a plan; each of these tasks has its own
requirements, and this first step is critical (2016). If we have the
wrong business understanding, we have no chance of coming up
with the right solution, because we’ll be solving the wrong problem.
Essentially, Brown calls on us to spend a good amount of time and
energy on problem framing and approach. Another one of our
readings this week, by Paloma Cantero-Gomez, expounds on this
component, and reminds me of a different phrase-go slow to go
fast-that underscores the importance of carefully and thoughtfully
framing the problem. Citing Albert Einstein, she notes that, with an
hour to solve a problem, 55 minutes ought to be spent thinking
about the problem, leaving just 5 for the solution (Cantero-Gomez,
2019). I really appreciate her guidance in the 40-21-10-5 rule, to
consider the universe around the problem, but then to break it
down by narrowing the description from 40 words to 5, getting to
the very essential components, only: the root of the problem
(Cantero-Gomez, 2019).
Going on from there, Cantero-Gomez’s guidance is particularly
useful for data analysts who may not be subject matter experts in
the topic they are working on; we may have a business partner, but
it behooves us to gain a stronger appreciation for the substantive
context, implications, and consequences, to make sure we’re
offering solutions that solve the problem for the business partner
(2019). It’s important then, to be open to reframing or refocusing
the direction of the project; to challenging assumptions and being
clear about what is established fact and what is not; and to engage
in exercises that explicitly force a change in perspective (Cantero-
Gomez, 2019). I use a cognitive science technique called “zoom in,
zoom out” that gets at this idea. Cantero-Gomez tells us to frame
the problem in the form of a question, to encourage open-
mindedness (2019). I find that ensuring the question starts with
who, what, when, where, why, or how is a good way to do this.
One piece of advice that Cantero-Gomez offered was particularly
surprising to me: the importance of using positive language to
engage the brain’s ability to think of the big picture, and to engage
in active listening and problem solving (2019).
Putting all of this advice together, then, I went from my original
problem statement (actually, statements):
1. What is the relationship between climate change and global
health security?
2. Zoom in: How does/will climate change influence the spread
of vector-borne diseases, and what diseases are particularly
influenced by the climate?
3. Why does it matter? because: What populations are most at
risk of increased disease presence and spread?
4. My refined, not-quite-five-word data problem: What factors
govern the spread of vector-borne disease? f
In doing my context gathering, I’ve learned that vector-borne
diseases have unique ideal conditions for spread (CVBD, n.d.).
Therefore, it will be important to keep that in mind in how I design
my model—for example, do I want to consider more than one
disease, or limit my study to the outlook for a particular one, like
malaria?
CRISP-DM will dictate my next steps (Brown, 2016):
• Investigating what suitable data are available. An initial
exploration suggests that reliable sources of information may
include the US Centers for Disease Control (CDC), the World
Health Organization, and the United Nations. Socioeconomic
datasets that would complement data about the diseases may
be found on the World Bank’s website. In addition, I will need
to find a reliable source of projected climate change data
(including temperature, weather events, and humidity/aridity)
and socioeconomic data to make predictions about the future
risk of disease.
• Data preparation, which will include figuring out how to
merge the datasets (i.e. finding the factor that will serve as
the key; in this case, most likely to be a regional descriptor,
like country), and determining a method for evaluating data
fidelity and dealing with missing or bad data.
• Data exploration and modeling. What relationships exist
among the various factors? What can we see by graphing the
data? Which relationships are not obvious, but seem to exist?
• Evaluate model performance—applying the model developed
in training to a test set of data.
• Integrate results: what recommendations can be made based
on the model’s findings? Which locales require more
resources to be invested in disease mitigation/health
adaptations? Where might we be able to reduce these
investments because the risk may decrease with climate
change?
References:
Brown, M. S. (2016, Mar 31). The right process for big data
analytics profit. Forbes.
https://www.forbes.com/sites/metabrown/2016/03/31/open-
standard-process-yields-best-big-data-analytics-
results/?sh=4ca287e43fae
Cantero-Gomez, P. (2019, April 10). How to frame a problem to
find the right solution. Forbes.
https://www.forbes.com/sites/palomacanterogomez/2019/04/10/h
ow-to-frame-a-problem-to-find-the-right-solution/#1282a1125993
CVBD (Companion Vector-Borne Diseases). (n.d.). Occurrence
maps. https://cvbd.elanco.com/cvbd-maps
