We live in a world that’s drowning in information. We have more information at the tip of our button than our ancestors combined. Even though we have more information at our tips than any other generation, we are also prone to a lot more misinformation. And at a much larger scale. And sometimes that can lead to disastrous consequences at an individual as well as cultural level. We might have more information in this age, but the average quality of information has dropped quite drastically.
This might be because, despite the abundance of information, the analytical capacity remains the same. It has become harder to spot the True from the Fake. Or the Real from The Bullshit.
This is an extremely important topic as in my opinion, spotting bullshit is underemphasized and under-taught. And having a rigorous bullshit detector is one of the greatest gifts one can ask for in this data-drowning world.
These are my (non-exhaustive)notes from the book
The difference between lying and bullshit is that lying is a manipulative tactic designed to lead you away from the truth whereas bullshit is written with a gross indifference to it.
The reason bullshit spreads like fire in the internet as the most widely shared posts are one that spark a sense of wonder, shock you or anything that makes extreme ends of your emotions tick. And most extreme claims are too good or too bad to be true.
Successful headlines don’t convey facts, they promise you an emotional experience
Finding true statements in a sea of false ones is like looking for a needle in a haystack
Correlation, Causality, Numbers
• Correlation is not always equal to Causation. There can be an infinite number of variables on the planet and it is highly likely that two totally unrelated metrics can show some correlation for a given period of time and fool you into thinking there is a pattern. A lot of spurious correlations can be created if you try to look for them. Idiots find patterns everywhere, Wiser folks find where there are none.
• The problem with correlation in psychological studies is individual behavior doesn’t map out to the collective behavior often.
• One way to know whether correlation means causation is to view things via inversion and check whether they still holds true. There has to be some creativity to debunk spurious-looking correlations that don’t imply causality.
• It is not enough for numbers to be correct. They need to be placed in an appropriate context so that the reader can interpret them. Often numbers are presented without an appropriate context.
• Advertisers often use numbers to persuade and will use metrics to show them in a better limelight
• ‘Mathiness’ is the tendency to form mathematical equations to impress and persuade an audience. They are many times done with zero respect for accuracy, rigor, and even consistent units.
• There are a lot of zombie statistics out there as numbers by virtue of being quantitative tend to spread quicker.
Goodhart’s Law – When a measure becomes a target, it ceases to be a good measure. This is because people start gaming the system for rewards of the target leaving other parts of the system exposed to unseen risk. This law can explain what caused Banks to blow up in 2007, academia to degrade, etc
Friendship Paradox – Most of your friends are likely to be friends with more people than you are. The explanation surrounding this is an interesting one.
Familiarity increases the stickiness of the myth
Selection Bias
• Selection bias arises when individuals selected for study differ vastly from the individuals eligible for the study.
• “If something is too good or too bad to be true, it most likely is”. Extraordinary claims require extraordinary evidence. Or Assume a fraud unless proven otherwise.
• Data censoring can be done to bolster a study that might not have been true otherwise. Right censoring problem is a type of data censoring problem.
Never assume malice/fraud when incompetence is a likely explanation (in most cases it is). And never assume incompetence when a reasonable mistake can explain things.
In this day and age, data visualization techniques can enhance or mitigate the impact on how data is perceived. Designers have great control over how data is perceived. They can use visually appealing techniques for that purpose in many ways.
• They can use 3D pies and graphs such that the lower pie can look larger and vice versa
• They can use slanted or creative charts/graphs which at first look to the casual reader can reduce/increase the impact of the message
• There were a tonne of such ways of data manipulation via charts and graphs shown in the book which is worth a read. It’s always crucial to ask yourself whether the story data tells aligns with the way it is presented. It can uncover any malice at the behest of designers
Bullshit in Big Data
Big data doesn’t necessarily mean better, it’s just bigger.
Machine learning – a term that gained newfound popularity in recent years is just algorithms combined with more data and higher processing power
GIGO – Garbage In, Garbage Out. An indication that training data is as important if not more important than algorithms. Training Data usually comes from the real world – which is full of human biases and their consequences. Training algorithms on that data perpetuate those biases.
News articles have immensely misinterpreted any advance in artificial intelligence or machine learning to make compelling stories
Machines still fall short when identifying humor, sarcasm, or fake news – for now
Bullshit in science
P values can be hacked by various methods as explained in the book
The P-Value hacking is an example of Goodhart’s law in full flow
News articles tend to mold headlines about science in a way to gets more clicks. Rarely are the mistakes in original articles retracted
Researchers tend to read only positive results in scientific literature and negative results are rarely published. Meta-analysis can be useful: looking at multiple studies simultaneously to get a clearer picture.
Peer review cannot catch every innocent mistake, let alone well-concealed acts of scientific jargon
Refuting Bullshit
Use reductio ad-absurdum , Counter examples, Analogies, and Null Models to get a clear picture.
Spotting Bullshit
Developing a rigorous bullshit detector is a lifelong process and takes continual practice. Thankfully the modern world offers the practice regularly.
The ways to spot bullshit
1. Check out the source
2. Beware of unfair comparisons
3. Be wary of extraordinary claims
4. Think in orders of magnitude – Use fermi estimation for a quick check
5. Avoid confirmation bias
6. Consider multiple hypotheses
Spotting bullshit online is hard and is getting harder day by day. It might be better to reduce the amount of information. Also, the use of fact-checking websites can be extremely useful.
Just because someone has an explanation for a phenomenon doesn’t mean it is the explanation for the phenomenon
Conclusion:
The major issue I have with this book is that it barely touches the surface of the art of bullshit and how it pervades various disciplines. It could have been a lot more elaborative. The rot runs too deep in this modern world and it did feel that authors have squandered a golden opportunity. Despite that, I would recommend this book to almost everybody. I am extremely glad someone wrote a book on this topic and hopefully, many more similar books are written. A good book but somewhat falls short of being a great one.