Facebook

Friday, February 22, 2019

Jussie Smollett, Bad Inferences, and Narrative

Introduction
I've been seeing what I take to be a lot of bad inferences by smart people concerning the Jussie Smollet hoax. There is a long-running narrative on parts of the right (particularly online) that we should be skeptical of the authenticity of many hate crimes. The Jussie Smollet hoax is pouring gasoline on this narrative and spreading it outside its usual domain on the right.

While both the hoax and the narrative are ugly, this is a beautiful opportunity to talk about some of my favorite critical thinking concepts....

Key Concepts
Fallacy of Confirming Evidence: Sister of confirmation bias, the fallacy of confirming evidence is when we count only confirming evidence and ignore disconfirming evidence when forming our conclusions. For example, suppose I hold the belief that vaccines cause autism. I go out into the world and I see an autistic child and I find out that child was also vaccinated. Hypothesis confirmed! I see another child with autism that was vaccinated. Yet more evidence. Ah ha! Vaccines cause autism. I could do this all day long: Find autistic children, discover their vaccination status, and if its positive count it as confirmation for my hypothesis.

The obvious error is that I'm not taking into account all the children who have been vaccinated but aren't autistic. In general terms, I'm only taking into account positive evidence and ignoring disconfirming evidence as I form my view.

The fallacy of confirming evidence often works together with motivated reasoning. Rather than examine a data set then come to a conclusion, I begin with the conclusion, "vaccines cause autism", then go out into the world and carefully select only the evidence that supports this view.

Good reasoning requires that we take into account both confirming and disconfirming evidence. Which leads to our next concept...

Framing: Absolute Numbers vs Rates: It's very easy to mislead people with absolute numbers since they provide no context. For example, if you hear that 20 people got A's in my class last semester you might think my class is easy. But not so fast. To make the correct evaluation you need to know how many people were in my class total. If there were only 20 students in my class then 20 A's is a decent indication that either my class is easy or I'm the world's greatest teacher. However, if it turns out that I had 500 students in my class, then you might draw different conclusions.

The lesson here is that we cannot evaluate absolute numbers without context and using rates is an excellent way of giving context. Partisan media and groups often use absolute numbers as a way of creating a narrative.

There are a bunch more, but this should be enough to get the party started. I've listed some other ones at the end of this post for the keeners.

The Jussie Smollet Hoax and Hate Crime Hoaxes
With these critical thinking concepts in our back pocket, let's take a look at hate crime hoaxes. Several right-wing media outlets have helpfully compiled lists of all the hate crime hoaxes during the Trump presidency going back to 2016. These lists are graciously prepared in order to save us from the epidemic of liberal hate crime hoaxes aimed to delegitimize the moral bonafides of Trump and his supporters.

I counted about 20 on the list. Let's triple that for fun. That's 60 hoax hate crimes since 2016. That makes 20/year!!!! OMG we're over-run with hate crime hoaxes. All hate crimes must be hoaxes. #DontBelieveThem

Oh, wait. We need to know the total number of reported hate crimes/year. The FBI puts it at around 7000/year. Let's do some math: Let's see...7000/20....that's 0.286%. So, less that one percent of reported hate crimes are hoaxes (if we triple the actual number). Clearly this is an epidemic. Our immediate reaction to someone claiming to be the victim of a hate crime should be to disbelieve them because there's a .286% chance it's a hoax:


Of course, there's a 99% chance that it isn't but let's not let statistics interfere with the narrative folks! Let's also keep in mind that the FBI and other reporting agencies estimate that the number of actual hate crimes is much higher than the number that actually get reported. This means that the percentage of hate crimes that are hoaxes is probably even lower than 2/10th of a percent.

As a final note, suppose absolutely everyone who was subject to a hate crime is included in the FBI statistics (which is very unlikely since the groups who are typically subject to hate crimes have good reasons to fear the police). Suppose we also multiply the actual incidence of confirmed hoaxes by TEN. That would be 20x10=200 hate crime hoaxes since 2016. Which means ~67 hoaxes per year. 7000/67= ~1%. So, even in the most charitable interpretation of the hate crime hoax epidemic, the incidence rate doesn't rise above 1%.

Don't fall for the right-wing narrative. Remember, facts not feelings!

Bonus Round:
Availability Bias: This is the tendency to think that the examples that most easily come to mind are also the most representative examples of a phenomena. The availability bias explains why many people are afraid of flying. When there's an airplane accident it's all over the news. We don't hear major news reports of all the airplanes that didn't crash. So, when some people think of airplane safety the first thing that comes to mind is the crashes, not the same flights. Because these are the examples that most readily come to mind, the mind takes them to be the most representative cases of airplane safety.

In the case of hoaxes, we are inundated with stories if there is a hoax (especially if you are in a right wing media ecosystem). The 7000 legitimate cases rarely get the media coverage the hoaxes do. Since the hoaxes are the most available cases, the mind takes them as the most representative cases, and extrapolates from them general conclusions about hate crimes.

Selection Bias: A selection bias will operate in conjunction with the availability bias. Which sorts of cases are the most likely to make the news? The ones that are outliers for a variety of reasons. They often involve high profile people or are anomalous for various reasons. There are 7000 hate crimes per year. Why don't we see all of them reported? Why doesn't right wing media report all the actual cases? There's selection bias going on. That media will only pick up the ones that serve to fulfill a narrative.

Another selection bias is that those who commit hate crime hoaxes are most likely to do it for attention. They want to get noticed. Hence, these types of cases will disproportionately enter the media cycle.

Base Rate Neglect/Base Rate Fallacy: This one's a bit tricky to explain so I'll hand over the details to the wikipedia article. Suppose the incidence rate of a phenomena is low. For example, 1% of all hate crimes are hoaxes . That means that for every case, all things being equal, we should assume that there's a 1% chance that it's a hoax. However, people fixate on the particulars of each case ignoring the base rate. It's not that particulars don't matter, it's that people place too much weight on the particulars in their reasoning while putting too little on the base rate.


No comments:

Post a Comment