Discover more from Course Notes: Continuous Business Learning
Calling Bullshit 5/5
Carl T. Bergstrom, Jevin D. West
9/ The Susceptibility of Science
Scientists seek both truth and recognition (usually – for being the first to make a discovery, the priority rule). Reputations are built with publications, lectures and guest appearances. But not all science is good science, and many “discoveries” can’t be replicated and thus need to be put to rest. Social psychology is my pet peeve, but even serious topics like cancer biology are greatly affected.
Scientists prefer positive hypotheses (A leads to B) over negative ones (A doesn’t lead to B) due to the fact that there are more negative hypotheses than positive, negative results are not as sexy as positive and may even look like a poorly run experiment.
p-hacking is when scientists try to find strong correlations and are able to find the matching data sets. However, for the experiment to be honest, the choice of data sets and variables must be made before the data is crunched (a hypothesis instead of hacking the outcome). Otherwise, it’s possible to see statistically significant results without any real patterns.
The papers that are published are a biased sample of all experiments conducted. Significant results are strongly overrepresented in the literature and non-significant results are underrepresented (kept in the researchers’ drawers). There are always false positives, and scientific papers are not an exception.
In science between 5% and 15% of published results are negative. Are scientists less likely to publish negative results or are they choosing the experiments that are likely to generate positive results?
News sources are happy to publish preliminary results and seldom report when the studies covered previously don’t pan out. Clickbait works, but negative results aren’t exciting.
A single study doesn’t prove anything; it just slightly shifts the consensus towards accepting a hypothesis. Some science writers do the “cafeteria science”, picking a subset of studies that tells a consistent and compelling story.
The most surprising studies are the ones that make the most exciting articles. Hm…
When a measure becomes a target, it cases to be a good measure: when the number of publications is a measure, there are always unscrupulous publications that will accept these publications for a fee with minimum level of peer review. Articles like this pollute the internet and mud the waters making telling facts from fiction hard.
These predatory journals can also be used to spread misinformation and lend credibility to articles as if they were peer-reviewed.
Any scientific paper can be wrong, regardless of where it was published. Peer review doesn’t guarantee that published papers are correct. They check for the logic, reasoning, and the conclusions. They can’t recreate the lab experiments; they won’t debug the program and even delve deep in the data itself.
A legitimate paper is written in good faith, carried out using appropriate methodologies and taken seriously by the scientific community. The journal has to be prestigious, and if one sees an extraordinary claim in a low-tier journal, this usually means that the paper is not legitimate. The more important the discovery – the more prestigious the journal will be. The topic must closely match the profile of the journal.
It’s a good idea to check if the article was retracted. MK: I usually google the name of the article followed by “criticism” or “debunked”. This works surprisingly well for many articles.
Question the source of information.
Who is telling me this?
How do they know it? (Fakes usually have one source and no cross-references.)
What is this person trying to sell me? (People are selling all kinds of stuff from used cars to ideas and beliefs.)
Beware of unfair comparisons. (There are more germs on airport security trays than on toilet seats. Yes, if this about respiratory viruses, no, if this about e-coli.) Triggering emotions is a powerful technique. Are we comparing apples and apples?
If it seems too good or too bad to be true. This one is quite obvious. Dig to the source; a tweet is a starting point, not the final proof. If something is substantially down – what is up then? Context matters. Also, shocking facts on social media exist because they get spread faster and wider.
Think in orders of magnitude. Lies are designed to lead away from the truth; bullshit is produced with gross indifference to the truth. Thus, lies are usually made plausible, while bullshit can smell bad from the beginning. It’s not hard to use mental math (Fermi estimation) and if the mental result is off the advertised number by a factor of 10 or more – maybe we’re dealing with bullshit.
Avoid confirmation bias. It’s a tendency to notice, believe and share information that’s consistent with our pre-existing beliefs. Why double check information before sharing in social media if we know it’s true?
Consider multiple hypotheses. Having an explanation doesn’t mean having the explanation; it’s completely possible to make wrong conclusions from the right data. It’s tempting to look at pieces of data obtained for the same period of time and imply causation, but it’s a shortcut that can turn wrong.
Triangulate. If an extraordinary claim was made online, find its source and see if other media picked up the story independently. Sources can be fake or non-existent. When possible – get the original report and compare. Use a reverse image search on Google to check image sources. Beware of deepfakes. The age of the media matters: the longer it’s been around – the easier it is to make a judgement about their reputation and reliability. Repetition of a story doesn’t make it true. Read less news. Think more, share less.
11/ Refuting Bullshit
Calling bullshit should be done responsibly, appropriately, and respectfully. Sometimes we produce bullshit and expect others to point us in the right direction, hence respect. Calling bullshit is a public activity, as the society will be better off without this piece of bullshit. There are cultural norms about it, of course. Many criticisms are better delivered in private and are simply not worth five minutes of fame and a lifetime of regret. After all, malice is more rare than incompetence or an honest mistake.
It’s really bad to be proven wrong while claiming something to be bullshit. It’s much easier to come up with bullshit than to refute it. So preparation is a must.
Reductio ad absurdum – a technique of extrapolating the assumptions to ridiculous conclusions. MK: There’s no better list of examples than here. The models chosen can be too simplistic for the task or not take into account physical world limitations.
Be memorable – debunking bullshit with good humour leaves a lasting impression on everyone.
Find counterexamples – if A implies B, then find the case where A is true, but B is not. It’s not easy to do on the spot, but the effect is spectacular.
Provide analogies – that’s quite obvious, but beware of false analogies.
Redraw figures – (that’s what I’m guilty of in at least one instance) the choice of charts can be misleading: cumulative data (which can’t go down by definition) over several periods of time can disguise the drop in absolute (i.e., non-cumulative) increases. In my case it was the number of flight bookings, but it works equally well with car sales.
Create a null model (if possible) – if it’s assumed that X is a necessary condition for Y, a null model will demonstrate that Y can occur without X present.
The issues of identity in people cloud their judgements: take away a wrong belief – and the person’s identity may fall apart. Hence, they will fight tooth and nail to keep their beliefs. Again, saying something many times doesn’t make it true. The hard trick is replacing the missing part of one’s belief system with new information.
We can’t have ready-made arguments for most of the bullshit we see, so sometimes it’s just better to avoid the confrontation altogether.
Being smart doesn’t need to be at the expense of others. There’s this “well, actually” type of sentence showing everyone how knowledgeable to person is. Such types of arguments don’t move the conversation forward, they focus on the person speaking rather than the topic being discussed. “Well, actually” people in many ways are no better than bullshitters: they don’t make others smarter.
/ the end
Thanks for reading Course Notes: Continuous Business Learning! Subscribe for free to receive new posts and support my work.