Bullshit Ahead

The_fin_de_siècle_newspaper_proprietor_(cropped)

* Article originally published on Medium as Bullshit Ahead

There’s always been bullshit, everywhere. Call it humbug, drivel, balderdash, it’s all the same. Much has been written about the fact that we seem to be swimming in bullshit (e.g., in this great article by Nathaniel Barr), much less on defining what exactly is bullshit. Harry Frankfurt’s essay is very good and informative, but it falls short of giving a definition that can work as a good metric.

Harry talks of bullshit as a disregard for the truth. That’s different from a lie, in the sense that a liar knows what is the truth, and deliberately presents an untruth with the intention of deceiving the receiving party, whereas a bullshitter works with carelessness, providing little information and a lot of hot air, where the intention is to cause a feeling on the receiving party, be it admiration or confusion.

The problem is: given an arbitrary sentence, can you derive the likelihood that it is bullshit?

I propose initially the use of four metrics:

  • Explanation
  • Ambiguity
  • Plural use
  • Bad grammar

I’ll go through each one to attempt to convince you that these are good starting points to detect whether or not a sentence is bullshit. Then, in subsequent articles, I’ll discuss how we might be able to measure each of these metrics.

Explanation

Tim has promised a week ago to send you the technical proposal for review by the end of the day yesterday, but didn’t. When you ask him “Did you finish the proposal?” Tim replies:

I started working on it, but the team meeting yesterday took too long, so I will have to finish it today.

For some time I have considered explanations to be a very clear indication of bullshit. When people fail, instead of falling on their sword, apologising, and showing how they can improve or make up for their failure, they provide an explanation, an excuse. It comes from the wish to not be seen as irresponsible, or forgetful, that is, the excuse is an attempt to maintain one’s status by shifting the blame to someone or something else.

The mature thing to do, of course, is to admit one’s failure, fix it, and see your failure as an opportunity to improve yourself. It might have been that the reason Tim didn’t manage to complete the proposal in time was due to a lack of time, but that should have been communicated earlier, or he should have planned his priorities better. In any case, the causes for him failing to meet the deadline do not alleviate the fact that the proposal was not delivered on time, and is therefore unnecessary to be stated. In other words, the information of the cause for his failure does not add any value to the mission of delivering the proposal. The only value it adds is for a process of continuous improvement, as a retrospective, so that in the future he might be able to decide what is more important: to attend a meeting or to fulfil his promise.

It is this lack of value that makes me think that excuses are big indicators of bullshit. An excuse it’s not necessarily a lie, although in many cases it is. It is an attempt to shift the direction of the conversation, so that the other party focuses more on the excuse than on the failure.

Ambiguity

This is again an attempt to confuse, by uttering words with little informational value. Politicians are well known for their prowess at crafting ambiguous statements. Who’s better to illustrate ambiguous political statements than the master himself, Donald J. Trump. When asked in May 2016 whether he’d be open to raising the federal minimum wage of $7.25 an hour. He replied he’d be:

[…] open to doing something with it, because I don’t like that.

This statement gets double points due to the fact that it also contains an explanation, but the key point here is the vagueness. what does “doing something with it” mean? Does that mean raising it, lowering it, investigating the cost-benefit and then deciding? There is a non-commitment that is typical of political public speech.

Plural use

This is also very popular with politicians, but it happens everywhere. In certain circumstances it is correct to talk on behalf of a group, when talking explicitly of things related to that group. Most of the time, however, the usage of plurals is an attempt to dissipate blame and shift focus, muddling the responsibility boundaries, or to bring the other party into the situation, so as to influence their emotion rather than their reason:

We must try to work together to overcome the difficulties we are facing.*

I admit this is not a strong metric, and there might be many reasons why someone might say something in the plural form, but together with other indicators it could help decide whether a statement is bullshit or not.

Bad grammar

This might be controversial. The theory here is that bullshit statements are lazy statements, unpolished and not well crafted. As mentioned by Harry Frankfurt, when someone concocts a lie there was a lot of effort put into it. Lying is not easy. Bullshitting on the other hand requires low cognitive effort, and my theory is that people who have a higher tendency to bullshit are people who will be paying little attention to their grammatical correctness. I am willing to be convinced otherwise though.


BSORNOT: a robotic assistant

How to use these metrics to detect bullshit? And are they enough? I think there’s still a lot of work to be done, but I decided to create a little Twitter robot to start experimenting anyway.

BSORNOT replies to mentions with an analysis of whether the text is likely to be bullshit or not, and which metric does it fall foul of. In the future I want to be able to also give the confidence level and an analysis of why the robot has reached its conclusion. I also want to make the tool respond to REST requests, and to be able to accept labels for statements from authorised users so that it can learn how better to detect bullshit.

If you want to collaborate to the improvement of the robot, please feel free to fork the project and submit your pull request. This is a public project on github, and my ambition is that it will be useful in extending our cognitive ability to separate facts from fiction, valuable information from bullshit.

 


*Side note: There is a hidden ambiguity here: when talking about “We”, does that include the other party or not? For example, a newly wed couple might be talking about their honeymoon: “We had a wonderful time”, or a group of friends might be talking about the time they went out camping: “We had a wonderful time”. The first clearly only includes the couple, whereas the second statement might include the listening party or not. In this the Vietnamese language is superior, when it makes the distinction between “We, including you” (Chúng ta) and “We, not including you” (Chúng tôi). In linguistics, this is known as clusivity.