Author: Carl T. Bergstrom and Jevin West
Publication date: 4 August 2020
Publisher: Random House
Number of pages: 336 pages
“Where is the data?”
That is the most common question we ask every time we want to hinder ourselves from misinformation. We lay our trust on numbers. We believe that data never lie and numbers offer precision. However, it is crucial to keep in mind that data presentation is frequently mislead while numbers can be miscounted, easily manipulated, and don’t tell the whole truth.
Calling Bullshit: The Art of Skepticism in a Data-driven World presents guidance on how to view and think critically about the information that circulated in news, magazines, and internet. This book is written by Carl Bergstrom, an evolutionary biologist, and Jevin D. West, a data scientist. Both of them are experts of misinformation in science.
The capability to spot bullshit on data is an important skill when we are surrounded by people who are trying to sell us something through information. Whilst it is hard to dismantle the bullshit because sometimes they are explained in a fancy terms that barely understood by common people. In this book, Bergstrom and West give the readers sets of tips on how to improve critical thinking. The most impressive part of this book is they offer enlightenments on how to detect the data-based misinformation only by looking at inputs and outputs without a deep knowledge of statistics or the related field. This book deserves to be included in a compulsory reading list at school and universities across educational stage.
Summary
The trait of bullshit
Who doesn’t familiar with a falsehood that vaccine cause autism? It has been more than 20 years since this misinformation were spread out and there is no single evidence that vaccine cause autism; even enormous research said that they do not. Yet, misinformation about vaccine persists.
Falsehood flies, and truth comes limping after it.
Jonathan Swift
The bullshiters only care about persuasion and impression rather than the truth. Creating bullshit is much easier than cleaning it up and the ones who clean it up are at loss position to those bullshiters.
Bullshit about causation
Every time we read an article, we have a high expectation about what we ought to do based on research. How to lose weight, what to eat to keep us fit, what is the effect of doing something in long term, or any other practical tips from a newly published journal. Unfortunately, oftentimes we forget that correlation does not imply causation.
It is not rare for ordinary people or even professional researchers to refer connection as causations. Proposing a cause-and-effect relationship based on single correlation is the most frequent misuse of data. It is not a wise move to build a perspective claims without evidence of causation from somethings that only correlated to each other.
Bullshit in science
We have to understand the nature of science: nothing is above questioning; the outcomes of an experiment are not absolute fact about nature; every claim is open to objection and every fact can be reversed in the face of evidence.
Most publishers publish findings that show significant results and nonsignificant results are rarely denoted. It calls file drawer effect. Just like an iceberg above the water’s surface, there is a high chance that what we understand from a scientific literature are only the positive results. We hardly acknowledge how many unpublished negative results are lying beneath the whole studies.
Another dark side of bullshit in science comes from how we rated scientists based on their number of publications. So-called predatory publishers up to hacking statistical assumptions lead to deliberate misdirection in the form of scientific fraud.
Bullshit with data visualisation
The objective of providing data visualisation is to let the readers with deeper and more nuanced perspectives in a glance effectively. The issues lies on not many of us have a knowledge in how to define data graphics. To make it worse, while data presentation seems to be unbiased, the designer has a prominent rule over the message a graphic delivers.
The most frequent abuses of data visualisation are exaggerating the shades of certain region in a chart that do not represent the numerical value and creating a graphic for the sake of eye-catching decoration instead of about the data.
Bullshit with big data
Policy makers [are] earnestly having meeting to discuss the right of robots when they should be talking about discrimination in algorithmic decision making.
Zachary Lipton, AI Researcher
Our aspiration in technology is to let the machine works in more objective manner than human. Sad to say that machines build their own rules to take decisions. They are not independent of human biases because they depend on the data they are fed. Algorithmic transparency and accountability are required to build a health environment in tech. Yet, many algorithms in big data are traded as confidential.
