Discover more from Simplify
Three Lazy Arguments that I Don't Want to Read Again
Absence of evidence, causation, and fallacies
“Absence of Evidence isn't Evidence of the Absence”
Sorry to start off on an awkward note, but did you murder my sister?
Sure, there's no corpse. No one reported the crime to the police. There's no motive. My parents can tell you that they've never had a daughter, and no court records show the birth or death of such an individual.
…But the absence of evidence isn't evidence of the absence, right? Well, sometimes it is.
If we 1) would expect some evidence to exist for X and 2)no such evidence for X manifests, then we have some evidence against X. That's why you don't believe that there’s a leopard in your house. It's why you don't believe that aliens abducted you last night.
Of course, sometimes the absence of evidence isn’t evidence of absence. I recall such instances in the early days of COVID. You'd read something like “no evidence that COVID is airborne.” That wasn't evidence of the absence. We just didn't know yet. As always, it's helpful to analyze these arguments via Phil 101 premise-and-conclusion form.
P1: If there were a leopard in your home, you would see evidence of a leopard in your home
P2: You don’t see evidence of a leopard in your home
C: There is no leopard in your home
If both premises are true (and I can only hope P2 holds for you), then we must accept the conclusion.
Now let's return to the early days of COVID. A similar argument would have looked like this
P1: if COVID were airborne, we’d have evidence of COVID being airborne
P2: We don’t have evidence of COVID being airborne
C: COVID isn't airborne.
At the time, premise two didn’t hold. If COVID were airborne, there would be no evidence of it. If it weren't, there would be no evidence of it. We couldn’t draw any conclusion from a dearth of evidence.
Allow me to cite a wise quote
If an airplane is on your back, it's a turtle. If your back is on a turtle, it's an airplane.
Okay, maybe you don't see the wisdom here. However, the quote boasts symmetry similar to that of the “absence of evidence isn't evidence of absence” one. Just because a phrase sounds nice, doesn't mean it's true. Be skeptical of pithy aphorisms.
“Correlation Isn’t Causation”
Quick caveat: with respect to this quote, I will assume that correlation means “relationship between two variables” rather than the specific Pearson correlation coefficient. Doing so opens up a more interesting discussion.
I scanned some data from the previous EPL season and found something shocking: clubs who scored more goals tended to finish higher in the table. I think that, maybe, scoring more goals causes a team to perform better. Likewise, I also think that conceding fewer goals leads to better performance. One last observation and I'm probably going to get canceled for this: I think winning more games causes a team to finish higher up on the table
…but the correlation isn’t causation, right? Do we really know that goals cause a higher end-of-season ranking? After all, these are only correlations, right?
I once saw the following image posted on LinkedIn:
I get the idea, but the meme doesn’t work. The intention goes something like this: by seeing a cat on top of a collapsed parking structure, we can’t assume that the cat caused the collapse. Instead, we know that cat crawled onto it after the structure collapsed. Wait, though, how do we know that the cat didn’t cause the collapse? We know because of, well, correlations.
We’ve all seen cats before, and we know that they’re pretty light. We’ve also seen parking structures, and we know that something as light as a cat couldn’t knock one down. We also know that cats like to climb stuff, so it wouldn’t be surprising to see a cat crawl onto a broken parking structure. In other words, the correct interpretation of this photo requires us to infer causation from correlations! How do we know light objects don’t knock down heavy ones? Because we’ve observed that relationship in the past. How do we know that cats are light? Because we’ve observed that relationship in the past? Heck, how do we even know that that thing is a cat, to begin with? Well, we’ve seen a lot of things like that look that, and they’ve all been cats.
Yes, sometimes correlations don’t imply causal effects. Consider the relationship between education and income. One can infer that obtaining more education causes increases income later in life. On the contrary, one might conclude that a series of other variables: parental income, conscientiousness, IQ, etc. lead to both higher education attainment and higher incomes. Researchers can try to tease out the direction of causality by looking for natural experiments where one group randomly received more education than another. One can also opt for more technical modeling procedures like instrumental variables. In either case, the researchers must find causation from the relationship between variables.
Causation isn’t some glowing green rock we can discover in a laboratory. No one can observe causality. We can only infer causation from some theoretical understanding of the world. Sometimes, that understanding stems from math, logic, or definitions. In the soccer example, we believe that goal scoring causes winning because such a relationship is built into the definition of the sport. The winner of the match is, by definition, the one who scores more. Other times, we rely on past observations. We infer that the collapse preceded the cat because of our folk understanding of physics (light stuff doesn’t break heavy stuff) and our previous observations of cats. In a world where we routinely saw cats take down heavy structures, we’d change our interpretation of the photo.
Another common argument tactic involves the invocation of philosophical party fouls. Understanding logical fallacies can help, but often times they seem to substitute for actual wisdom.
One problem is the lack of distinction between “always wrong” and “sometimes wrong” fallacies. The first camp would include circular reasoning or incorrect logic (e.g., using A implies B to argue that B implies A). These errors always invalidate an argument. No valid conclusion can be derived from an argument that commits either mistake. On the other hand, some fallacies are only wrong sometimes. Consider the “slippery slope” fallacy. Some people get pressured into smoking a single cigarette before gradually becoming daily smokers. Others put the thing in their mouth, nearly throw up, and never touch the stuff again. Sometimes, democracies slide into dictatorships. Other times, democratic governments gain temporary powers and then lose them. We can’t craft make a general claim about these topics. Some slopes are slippery, and others aren’t.
Imagine a friend saying, “I won’t have a beer with my meal tonight, because that could lead to me becoming an alcoholic.” That might represent a slippery slope, but the friend might hold that belief for a good reason. Maybe multiple family members have struggled with alcoholism. Maybe the individual has seen addicted tendencies in himself with gambling or drugs. Yet, the person to his left could make the same claim without any reasoning to back it up. One slope looks slippery while the other doesn’t.
Another example involves the tu quoque fallacy. These arguments usually take the form of “yeah, but you do X as well.” Again, though, this is only sometimes wrong.
Let’s say that a businessman makes the following claim:
I’m sick of all this welfare. The government shouldn’t be handing people money!
An activist responds by arguing “yeah, well you took government subsidies to start your business!” Is this a fallacy? Well… maybe.
People rarely spell out their arguments in premise-and-conclusion form, but one interpretation of the businessman’s argument may look like this:
P1: Adults should not receive from the government
P2: Welfare recipients are adults
C: Welfare recipients should not receive money from the government.
In this case, the activist has a point. If the businessman took government money, then he probably believes it’s okay to do so. If he holds that belief, then P1 fails, and his argument doesn’t work. The activist infers this belief from the businessman’s actions, which is a reasonable thing to do. If someone owns multiple cats, wears cat-themed clothing, and has cat wallpaper on their cell phone, you assume that they like cats. That’s not a fallacy.
Sure, the businessman could respond with the following:
Yes, I get government money, but they shouldn’t give me money either! No one should get free cash. However, given that these programs exist, I’m going to take advantage of them, regardless of my principles.
Now, there’s no longer a contradiction in the businessman’s argument. He’s just acting in a way that’s incongruent with his beliefs, so the tu quoque argument no longer holds. Unlike circular reasoning, we can’t generalize this fallacy. We have to consider the argument placed in front of us.
A Better Way Forward
Instead of resorting to canned one-liners, just show why the thing is wrong. Find the flaw in the argument and explain it.
Let me return to the “correlation isn’t causation” example for a second. Consider the following claim:
Cities with more crime have more police. I think that police cause crime!
I could say “correlation isn’t causation.” I could also say “cities with more crime hire more police.” That’s no more complicated than the “correlation isn’t causation” mantra, and it provides a reason for rejecting the claim. The non-canned version opens the possibility for a sophisticated discussion about the relationship between police and crime. We can’t, on the other hand, do much with “correlation isn’t causation.”
Here’s another example:
Countries that spend more on education have better health outcomes. I think education spending improves health.
One possible refutation: higher GDP increases both education spending and health. From there, we can debate how to find the true direction of causality. We can only engage in that debate, however, if we abandon lazy one-liners in favor of real arguments.