The Bullshit Algorithm

If you use Swiffer WetJet, you are a puppy murderer.

But wait, you say. How could I? P&G would never lie to me about the safety of their Swiffer® WetJet™?! But of course. All of those “chemicals.” How could I be so stupid!

Yep. You’re a cold-blooded murderer. Isn’t it lucky that you can tell your story? Now, no one else will need to suffer what your family has suffered. You can warn us. Why don’t you go ahead.

Well, okay. I’ll tell you…

I recently had a neighbor who had to have their 5-year old German Shepherd dog put down due to liver failure. The dog was completely healthy until a few weeks ago, so they had a necropsy done to see what the cause was. The liver levels were unbelievable, as if the dog had ingested poison of some kind. The dog is kept inside, and when he’s outside, someone’s with him, so the idea of him getting into something unknown was hard to believe. My neighbor started going through all the items in the house. When he got to the Swiffer Wetjet, he noticed, in very tiny print, a warning which stated “may be harmful to small children and animals.” He called the company to ask what the contents of the cleaning agent are and was astounded to find out that antifreeze is one of the ingredients. (actually he was told it’s a compound which is one molecule away from anitfreeze). Therefore, just by the dog walking on the floor cleaned with the solution, then licking it’s own paws, and the dog eating from its dishes which were kept on the kitchen floor cleaned with this product, it ingested enough of the solution to destroy its liver.

 Soon after his dog’s death, his housekeepers’ two cats also died of liver failure. They both used the Swiffer Wetjet for quick cleanups on their floors. Necropsies weren’t done on the cats, so they couldn’t file a lawsuit, but he asked that we spread the word to as many people as possible so they don’t lose their animals.

Source: Snopes.com

*Of course, this is a hoax. You may have seen it make the rounds last year…perhaps as recently as a few months ago. But doesn’t it sound convincing? It should. As a professional persuader, I can help tell you why. This story has lots of goodies (17 in fact, but more on that later). Let’s recap the top four:

1.    The helpless and innocent subject: Who is more innocent than the family dog? He doesn’t know better. It’s your job as the owner to protect him from harm, and you failed.

2.    The details: It wasn’t just “a dog”, it was a “5-year old German Shephard”. It wasn’t just that the dog died, it was the sequence of events of walking on the floor, licking his paws, eating from dishes kept on the floor.

3.    Seemingly scientific facts: The writer was brilliant here. If he or she has given the chemical formula, most people would have buzzed right by it. But “one molecule away from antifreeze” … now that’s scary!

4.    Corroborating evidence: The neighbor’s cats also died of similar circumstances (liver failure plus Swiffer WetJet usage). Just in case you thought this might be an isolated incident, your pet is in danger too!

If I were trying to damage the sales of the Swiffer WetJet product line, I could hardly do better. Yes, stories like this one made the rounds before the rise of Facebook, but their impact was much more limited. In the time it took misinformation to spread, the product owner would have the time to craft and spread its own rebuttal. If the situation were serious enough, it could run advertisements. It could update its product packaging. It had options.

But today, stories like this one “go viral” so quickly and with such ferocity that P&G had no time to mount a defense. Yes, Snopes will (eventually) debunk the story, but that can take weeks. By then, sales suffer, and consumer trust erodes.

Isn’t it funny? Wasn’t the promise of data-driven, search engine and social media algorithms that they would amplify the truth and protect us from misinformation by tapping the wisdom of crowds? The fact is that they do not. And cannot. Because that is not what they are designed to do. At the heart of every social media algorithm is a fatal flaw that values persuasion over facts.

Social media platforms (as well as search engines) are not designed for truth. They are designed for popularity. They are bullshit engines.

To understand how we got here, we need to take a step back and understand bullshit.

Best. Academic. Paper. Ever.

Harry G. Frankfurt, professor of philosophy at Princeton University asked the obvious question in 2005:

“One of the most salient features of our culture is that there is so much bullshit. Everyone knows this. Each of us contributes his share. But we tend to take the situation for granted. Most people are rather confident of their ability to recognize bullshit and to avoid being taken in by it. So the phenomenon has not aroused much deliberate concern, or attracted much sustained inquiry. In consequence, we have no clear understanding of what bullshit is, why there is so much of it, or what functions it serves.”

One of the oddest things about this paper, and I highly recommend you read the entire 20 pages, is the thorough disassembly of a topic everybody knows exists, but no one seems to understand.

Frankfurt made bullshit a technical term.

Here’s the crux of it: Most of us tend to think of the world in terms of facts and fictions, truths and lies. As we become more sophisticated, we understand people can have different perceptions (read: opinions) about the value truth brings or harm lies cause. However, those opinions exist on a different level than the “objective foundation” of fact and fiction.

Professional persuaders know this is not the way the world works.

The purpose of much of the communication we see – between people in our private lives, our consumer relationships, and the political sphere – is not to illuminate the truth, but rather to persuade. In fact, a mix of truths, half-truths, and outright lies is a great way to do it. Real facts are messy, incomplete, and often contradict each other. Outright lies can be fact-checked and objectively disproven. On the other hand, a skilled bullshitter can weave a tidy and convincing story based on a mix of facts and fictions. Facts are indeed objective facts to the bullshitter, but their value is not their factual basis, but rather their ability to persuade. A half-truth or lie might do just as well. The entire spectrum is at the bullshitter’s disposal, where his non-bullshitting competitor only has the facts. It’s not a fair fight.

Frankfurt makes the case that bullshit has a place in everyday life. Without it, we would be paralyzed with uncertainty and unable to make the simplest decisions and tend to the most basic relationship tasks. (Are you really going to tell your husband his haircut looks stupid?) Bullshit is as natural as…well…bullshit.

So, if bullshit is natural, and perhaps even necessary, where’s the problem? We’ve been dealing with bullshit since the instant we developed culture and language. What’s different now?

The Search Engine, Social Media, Data-Driven (Bull)shit Storm

The internet generally, and social media specifically, is not a truth platform, it is a popularity platform. That might come as a major surprise to many of you, or as blindingly obvious, but it’s important to unpack how these algorithms work so that we can understand the depth of the bullshit problem.

The bullshitty foundation of the internet as we know it: Search and Social

At a high level, how does a search engine algorithm work? The basic concept is authority. In short, that means how credible one source of truth is than another. In some cases, that’s obvious: Your state’s department of motor vehicles website is probably a more authoritative source for driver licensing procedures than your cousin’s floral arrangement blog. But it’s not humans that make those judgments. Algorithms need to do that work for obvious reasons of scope and scale.

Those non-human algorithms need clear rules for how to determine credibility. One of those important rules is simple: How many other websites link back to that one website for a particular search term or function? Link backs are an important proxy for credibility. Yes, it’s more complicated than that (Google, Bing and others strip out obvious gaming of that system), but at its heart, “authority” equals “popularity”, not truth, and not facts.

In other words, your cousin’s floral blog could become a leading authority on driver licensing with enough time and effort … and others agreeing that it is an authoritative source by linking to it in the context of that search term. This is the “wisdom of crowds” idea in a nutshell – the ultimate authority rests in shared agreement of “truth,” not actual truth based on objective facts.

Let’s translate: Sometimes search engines are right. Sometimes they’re wrong. But they always represent persuasion and popularity. Search engines are bullshit engines.

Let’s translate again: That little search window on your computer that you rely on to find facts is feeding you bullshit. Remember, true bullshit has some fact and some fiction, but it’s all persuasion. So yes, you’re getting some facts, some of the time. But just as often you’re getting hoodwinked.

If a search engine is a bullshit engine, social media is a bullshit rocket.

Social media algorithms completely dispense with the idea of truth. They are designed to enhance social connections. What drives a social media algorithm is something more than authority in a search engine (although that still matters). The most important driver of the algorithm is engagement, aka social proof. That takes the form of likes, clicks, shares, comments, reposts, etc.

The higher the engagement, the more authority the post (and author) have, especially when certain posts “go viral.” All that means is that the engagement rate gains enough attention fast enough to feed on itself, bending the exponential curve.

Most of the time, what goes viral are puppy videos, prom dances, pratfalls, and pornography. Mostly harmless, but let’s ignore those for now.

Every social media algorithm – every one of them – uses some proprietary combination of those factors (along with advertising dollars) to determine what becomes “popular” consistently. It’s not hard to spot. With a little training, you can do it too.

Here’s your first lesson: What story seems more likely to go viral?

  1. Sustained wellness comes from eating a balanced diet of healthy food, lowering stress, and exercising regularly.
  2. Drinking bleach is the most effective way to stay hydrated during the summer months.
  3. You can lose up to five pounds in the first two days using a clove and pomegranate enema.

The first is obvious, but boring, truth. No chance for virality there. The second is just as obviously a lie. (Please don’t try that at home. You’ll die.) The third is pure bullshit, and you can see immediately why it’s so compelling. It seems like it could have some truth to it. That one has potential!

Let this sink in: The two most common ways you learn about your world, the search engine and the social media timeline, are designed from the ground up to feed you bullshit.

It gets worse. You aren’t as good as you think you are at detecting bullshit.

Sure. A clove and pomegranate enema seems like bullshit (although I can think of stranger things). If you try one, I think you deserve what you get. But for most people, when we see examples like that one, we feel pretty confident we can pick bullshit out of our social media feed and safely ignore it.

We’re wrong.

To paraphrase a more famous phrase: You may be able to catch
all
of the bullshit
some
of the time, and you will catch
some
of the bullshit
all
of the time, but you will never catch
all
of the bullshit
all
of the time.

Your social media feed scrolls by too quickly. There are too many stories. There is not enough time. No one has the energy to fact check every story that floats by or every search result that finds its way to page one. What’s worse, until today, many of you believed search engines and social media platforms somehow prioritized the truth over bullshit. They do not. They prioritize authority and popularity – a bullshitter’s two favorite foods.

The average person sees thousands of search engine results and social media posts each day. You physically cannot fact check them all. No one can. It is a virtual certainty you have been bullshitted today. And the worst part? You don’t know which ones they were.

If we’re going to be continually drenched in a bullshit storm, we could use an umbrella.

I think it’s only fair we built our own bullshit algorithm.

To the uninitiated, an algorithm seems like some bizarre technical concept that only engineers and programmers can understand – that you need to learn special language skills or grow a thick beard. You don’t. An algorithm is super easy: It’s a set of rules. Let’s write a simple one right now, shall we?

IF the weather outside EQUALS “raining”,

THEN pack an umbrella.

Yep. That’s it. That’s all there is to it. In fact, algorithms are all around you. All recipes are algorithms. So is (essentially) all of mathematics. You are so familiar with algorithms that you write, perform, and revise them every day without thinking about them. And yes, software algorithms (like those designed to drive an autonomous car) are super complicated. But that doesn’t mean we should be scared of the basic premise.

Anyone can do it.

Remember the game “20 Questions”? That game was a sort of algorithm. Here’s my adaptation for detecting bullshit.

Step 1: Open your social media feed and pick out a story. It can be any story.

Step 2: Read the story and answer the following 20 questions.

Step 3: The more questions you answer “yes” to, the higher the likelihood that story is bullshit.

Does the story…

1. …feature a powerless, helpless, or disadvantaged victim?

2. …push a political or identity hot button?

3. …result in the most dramatic outcome possible (death versus injury)?

4. …include irrelevant details (details not directly relevant to the crux of the situation)?

5. …suggest a simplistic next step or action (get rid of X, stop eating Y)?

6. …include a “twist” in the story, a surprise, or a big reveal?

7. …feature “scientism” (little evidence with big conclusions)?

8. …include hard to verify evidence (no links to reputable source, or only links to other non-authoritative sources)?

9. …use anecdotal versus statistical corroborating evidence?

10. …make grammatical or spelling errors, or use clumsy language?

11. …use over the top emotional appeals incongruent with the situation?

12. …use scientific jargon (e.g. “dihydrogen oxide” instead of the more common “water”)?

13. …attempt to be relatable using the experience of people “like you”?

14. …make spurious correlations (seeing patterns of related items that could have other causes)?

15. …dangle dread (chemicals!) without explaining the context of risks?

16. …push for urgent, immediate action?

17. …include charts, graphs, images, or videos that don’t have anything to do with the core features of the story?

18. …hint at a conspiracy, that someone is hiding something (ideally, a “big corporation” or “big government”)?

19. …publish first in a “bullshit attractor” (TED Talk, Facebook, etc.)?

20. …include statistics touting its popularity (e.g. how many people are talking about this)?

Let’s apply our new Bullshit Detection Algorithm to our Swiffer story from earlier. How’d it score? Pretty well, actually! It received a 17 out of 20 by my count. How could we have made it even bullshittier? (Remember, you don’t have to stick with the facts.)

Item 2: Add a detail about the owners of the dog as “Trump supporters.”

Item 18: Hint that the author knew some who worked at P&G who “had information” about these pet deaths, but she would be fired if she said anything.

Item 20: Include the number of “likes” or “shares” in the article, showing its popularity.

Easy, isn’t it?

Where do we go from here?

It’s not realistic to take every story you read through your new Bullshit Detection Algorithm. It’s also not realistic to stop using search engines and social media. They are too ingrained in the fabric of our daily lives. Maybe we should crowdsource a Chrome plugin to help automate the process of Bullshit Detection…to fight fire with fire? Let me know if you’ll throw in 20 bucks.

But at the very least, you can rest easy that you didn’t kill your dog by cleaning your floors with a Swiffer WetJet. And if you’re considering losing weight using a clove and pomegranate enema, you might want to try your new Bullshit Detection Algorithm first.

Source notes for this article:

Swiffer WetJet Pet Danger

https://www.snopes.com/fact-check/swiffer-wetjet-pet-danger/

No bullshit here. You can read the story (and the fact checking) for yourself.

 

Swiffer WetJet Hardwood Floor Spray Mop Starter Kit

(Ingredients list)

http://smartlabel.pg.com/00037000928119.html#top

If you’re curious, you can see the ingredients list for yourself. I’ll warn you, P&G doesn’t give you a chemistry lesson. For example, you’ll need to find a different (authoritative) site for more information on the ingredients.

Here is a link for more on PROPYLENE GLYCOL n-BUTYL ETHER, for example:

https://www.cdc.gov/niosh/ipcsneng/neng1614.html

Again, unless you have a background in chemistry, more information can get even more confusing. This is the one “chemically close” to Antifreeze, or ETHYLENE GLYCOL…that’s why I picked it. It’s clearly not antifreeze, but it sure sounds like it, doesn’t it? If you look at the CDC entry for this one, however, it’s basically screaming at you to run to the hospital if you ingest too much of it. Lots of chemical names sound the same, but are very different. I sort of wish I paid more attention in chemistry in college…

https://www.cdc.gov/niosh/ershdb/emergencyresponsecard_29750031.html

 

On Bullshit, Harry Frankfurt, Princeton University

http://www2.csudh.edu/ccauthen/576f12/frankfurt__harry_-_on_bullshit.pdf

This is what I hoped every academic paper would be like in graduate school, and while some were quite good, and informative, and interesting, nothing was as satisfying to read as Frankfurt’s 20 pages. Thanks, Jeremy Rose, for having us read it in grad school!

Editor’s Note: This article is republished with permission of the author with first publication on LinkedIn.

Posted in: Internet Resources, KM, Search Engines, Social Media