In Defense of Unanswerable Questions (They Are Not Pointless or Unscientific)
Short of time? Read the key takeaways.
❓ Some dismiss questions that evidence can't settle. A common view holds that if no empirical evidence could favor one answer over the others, a question is pointless. Taken seriously, this view would discard many important scientific questions.
🔭 Evidence often can't pick between competing theories. In quantum physics, scientists largely agree on the math but disagree on what it means. In historical sciences, past events can't be replicated, and decisive new evidence may never appear.
⚖️ Use inference to the best explanation. When evidence can't settle a question, compare rival explanations by their theoretical virtues, such as simplicity, explanatory power, fertility, and falsifiability, then favor the most virtuous one.
🧪 This reasoning is common in science. Darwin described his method as seeing how many classes of facts his theory could explain, and Poincaré saw simplicity as essential. Theoretical virtues don't guarantee truth, but they support careful best guesses.
💭 Unanswerable questions can carry moral weight. We may never be able to empirically test whether a machine is conscious. Redefining consciousness as behavior risks misclassifying who matters morally, either ignoring real suffering or attributing suffering where none exists.
Roughly two decades ago, one of my closest friends invited me on a trip with some of his extended family. On a walk through the countryside, I got talking to his cousin about science and philosophy. These days, that cousin is a somewhat famous public intellectual, but before he was introducing his YouTube viewers to computer science concepts, he was introducing me to the concept of the “singularity” and we spoke at length about whether machines could ever be conscious. During that conversation, this soon-to-be-semi-famous person said something that I have encountered many times since, along the lines of:
“Since we cannot ever really know for certain whether any given machine is conscious, the question is kind of pointless.”
He went on to argue that we ought to redefine “conscious” to mean something like “capable of certain behaviors” (e.g., passing the Turing Test), so that the question becomes empirically answerable. But that’s a whole other can of worms.
A similar sentiment is brought up when people resist thought experiments with statements like, ”Why bother asking about this if it could never happen in real life?” You may feel this way yourself: if something can’t be tested empirically, why bother talking about it? How could a question matter if there’s no way of checking the world to see whether proposed answers are true?
This view is common. I have encountered it from philosophers, scientists, and even when chatting to readers of this newsletter. In this article, I’m going to briefly outline a couple of arguments against this view. In the process, I hope to enrich your view of how science works and offer you a framework for deciding between answers to unanswerable questions.
What’s the view being attacked?
Let’s give the view we’ll be arguing against a name and a simple definition, to make it easier to talk about. The view has lots of philosophical cousins (e.g., verificationism, scientism, pragmatism) but its closest relative seems to be in the provocative and eminently quotable principle known as “Hume’s Fork”, which states:
“Does it contain any abstract reasoning concerning quantity or number? No. Does it contain any experimental reasoning concerning matter of fact and existence? No. Commit it then to the flames: for it can contain nothing but sophistry and illusion.” — David Hume, An Enquiry Concerning Human Understanding
People who knew me in my early 20s might associate Hume’s Fork with my habit of saying dramatically (about anything I disliked) “Commit it then to the flames!” But Hume’s Fork is more generally associated with the philosophical tradition known as “empiricism”. However, the view we’re talking about is not quite as strong as this. Hume’s Fork claims that empirically unanswerable questions contain nothing but sophistry and illusion (unless they are matters of pure mathematics), whereas the view we’re discussing has a more practical focus: it says that such questions have no practical value and so are pointless or useless (or similar). So, let’s call the view “Practical Empiricism”.
Practical Empiricism: If no empirical evidence could favor one potential answer to a question over all the others, then it is a pointless (or useless, or uninteresting, or unscientific) question.
Note: Arguing against Practical Empiricism doesn’t mean defending all empirically unanswerable questions. Some really are a useless waste of time (the poster child for such questions is “How many angels can dance on the head of a pin?”, which has become a parody of medieval theology), but I hope to show you that some are not.
Argument 1: Practical Empiricism misunderstands science
I don’t know about you, but my interest in science is driven, in large part, by:
A desire to understand the universe around me
The belief that the methods of science give us some of the best odds for increasing that understanding
If you want to demonstrate that science is useful, you might point to inventions it has given us, like bridges and GPS and non-stick frying pans. But you don’t have to point to inventions to demonstrate the usefulness of science: something else it gives us is understanding through explanation. Science explains things. Science is useful because it explains things. The explanation is a use, even when the explanation doesn’t lead to invention.
The thing is, many explanations (including in science) are in competition with alternative explanations, and the question “Which explanation is correct?” is often not something that can be settled empirically. Either because the empirical evidence is roughly equally likely according to competing theories, or because the empirical evidence offers inconclusive support for one theory over others. This general phenomenon is known as the “underdetermination of theory by evidence”. We can see Einstein explicitly acknowledging it in the following passage:
“Physical concepts [...] are not, however it may seem, uniquely determined by the external world. In our endeavour to understand reality we are somewhat like a man trying to understand the mechanism of a closed watch. He sees the face and the moving hands, even hears its ticking, but he has no way of opening the case. If he is ingenious he may form some picture of a mechanism which could be responsible for all the things he observes, but he may never be quite sure his picture is the only one which could explain his observations.”
The prevalence of underdetermination in science is debated (some say it affects literally every theory, others say there are some theories it doesn’t affect), but its presence causes a problem for Practical Empiricists. Here are just three examples.
Example 1: Quantum Physics
You probably know that quantum physics is weird, but did you know that scientists have offered lots of different, competing interpretations of that weirdness? Scientists largely agree on the underlying mathematics and the predictions they generate, but there is still a lot of disagreement on how to interpret what that mathematics tells us about the universe. This matters because it dramatically affects our views about what’s out there in the universe – from whether every possible outcome occurs in a branching multiverse, to whether quantum physics describes reality itself or only our expectations about it.
Here’s a fascinating decision tree, showing the various questions (in the green boxes) that the different interpretations of quantum physics (in the blue and red boxes) disagree about:
Some of these questions look to be, in principle, not the sort of thing that an empirical test can answer. For example: “Quantum realm?” asks: Does the math of quantum mechanics describe a physical reality that underpins our experimental observations?
Example 2: Historical Sciences
Many sciences are concerned with questions about historical events that are inaccessible to us now except through traces that have survived. For example:
Astronomers might ask, “How did the early universe develop?”
Evolutionary biologists might ask, “Why did bipedalism evolve?”
Archaeologists might ask, “What was this ancient site used for?”
In each case, the event scientists want to understand is not one that can be replicated again under controlled conditions. The best that scientists can do is develop theories that accommodate all the evidence available and argue about which theory is better (more on this below) until new evidence shows up. If it shows up at all.
When it comes to such questions, things are often as they appeared to paleontologists in the 1980s, trying to figure out why the dinosaurs went extinct. As one historian of science put it:
“None of the evidence available at the time provided strong support for any one of these hypotheses over the others, and most paleontologists suspected that we would never know which is correct.”
Fortunately for paleontologists, the surprising discovery of anomalous amounts of iridium at the geological boundary between the Cretaceous and the Paleogene periods set off a new wave of investigation that gradually led to the dominance of the meteorite impact theory. But nobody could have known that such evidence would show up, and for many scientific questions about the past, no such smoking gun ever does.
Example 3: Conspiracy theories and superstitions
A feature of some conspiracy theories and superstitions is that they are unfalsifiable, which means they can incorporate any piece of evidence you throw at them. If someone claims that a shadowy organization controls all the worlds’ governments and you point to some investigation that failed to find evidence of this fact, the believer might reply, “Of course! But that’s what you’d expect to see, because the organization is powerful enough to cover up its tracks!” When a conspiracy theory or superstition is unfalsifiable like this, there is no empirical evidence that can settle the question of whether the theory is false. No empirical evidence can show that a more scientific explanation is to be preferred.
What Practical Empiricism gets wrong
The examples above raise problems for Practical Empiricists. It seems that they are committed to saying that the questions raised in those examples are useless, uninteresting, and so on. Such questions should be committed to the flames. If the Einstein quote is apt, then the problem goes far beyond these examples and affects all of science! But, are Practical Empiricists actually willing to pay those costs? Are they actually willing to dismiss all those scientific endeavors and more? If so, then it seems that by throwing out the bathwater (i.e., truly useless questions like “how many angels can dance on the head of a pin?”) they’re committed to throwing out the baby too.
Answers to the questions raised by the first two examples (quantum physics and historical sciences) offer to deepen and enrich the explanations offered by those sciences and thereby deepen and enrich the usefulness of science (since explanation is one of the uses of science).
Answers to the question raised by the third example (conspiracy theories) are of obvious practical importance: being able to discern truth from false conspiracy or superstition is an important aspect of navigating the world and can have profound effects on behavior.
And yet none of them can be settled empirically.
The key fact here is that sometimes (perhaps always) the empirical evidence is compatible with multiple different theories, but that doesn’t mean there’s nothing rational left to say about those theories, to help us decide between them.
How to answer empirically unanswerable questions
If you’re a Practical Empiricist, you might be wondering how on earth we’re supposed to decide between competing explanations without empirical testing. Fortunately, there is a well-established rational method of doing so that is widely used in science and philosophy and beyond. It’s called “inference to the best explanation” (sometimes also referred to as “abductive” reasoning, which is an alternative to deductive and inductive reasoning).
Inference to the best explanation involves comparing competing explanations by comparing their “theoretical virtues” and selecting the explanation that is the most theoretically virtuous. The theoretical virtues include things like:
Empirical Adequacy: Does it fit the available empirical evidence? But this article is specifically about cases where this can’t settle the matter, so we need to look to other criteria.
Simplicity/Parsimony: Other things being equal, is the theory simpler than its rivals? (This is sometimes known as the principle of “Occam's Razor”.) Simplicity comes in a variety of forms, such as:
Having fewer assumptions
Positing fewer kinds of entities
Being able to be described using fewer parameters or with a lower "complexity" model
Requiring fewer logical or mathematical tools to express the theory.
However, it's debated precisely how to use Occam's razor properly (and why it works, if it does).
Explanatory Power: A good theory should provide deeper explanations and unify diverse phenomena under a coherent framework rather than just describing individual observations.
Fertility: A fertile theory should open up new areas of research and suggest novel predictions or experiments.
Falsifiability: Does the theory make testable predictions that could lead to observations that prove the theory false?
Notice that two of those virtues explicitly address empirical evidence (empirical adequacy and falsifiability). Inference to the best explanation doesn’t ignore empirical evidence; instead, it understands something that Practical Empiricists do not: that empirical evidence doesn’t exhaust the considerations we can use to rationally decide between theories.
Of course, having more theoretical virtues than alternatives doesn’t guarantee that a theory is true. Theoretical virtues are features of theories that scientists and philosophers believe tend to be possessed more by true theories than false theories, so they help us to make informed, careful, best guesses about the answers to empirically unanswerable questions, even if they don’t give us absolute certainty. (But then again, nothing can guarantee a theory is true and all beliefs should come in degrees anyway.)
This approach really is ubiquitous in science. For example, when Darwin encountered the objection that his theory of evolution couldn’t be proved, he responded by saying his reasoning was guided by the virtue of explanatory power:
“my general line of argument [was] inventing a theory, and seeing how many classes of facts the theory would explain”
And Poincaré famously championed the virtue of simplicity as essential to science:
“those who do not believe that natural laws have to be simple, are still often obliged to act as if they did. They could not entirely avoid this necessity without making impossible all generalization, and consequently all science.”
Yet, Practical Empiricists seem unable to countenance using any virtues other than empirical adequacy and falsifiability to decide between theories (since they require that the question of which theory is true can be settled empirically). That makes their view a narrow and impoverished view of the methods of science and reasoning.
If you’re skeptical of inference to the best explanation, you might want to check out our article that walks you through using theoretical virtues to decide between a scientific explanation and a conspiracy theory (you can see that article here). Or read about Einstein’s work on simplicity.
Competition between theories is not always settled by empirical investigation – as in the case of quantum mechanics, historical sciences, conspiracy theories, superstitions, and many more. Yet we are still able to make rational judgments about them; we can rationally have a preference between competing scientific theories, or between science and conspiracy, even when empirical tests can’t settle the matter. We do this by using the method of inference to the best explanation. This kind of reasoning is ubiquitous in science. Thus, empirically unanswerable questions are not inherently unscientific.
And since such theories provide us with explanations of the world around us, they are also useful and interesting.
Argument 2: Empirically unanswerable questions can be morally important
Now, here’s a second (and much shorter) argument. Consider the question discussed near the start of this article: “Could machines ever be conscious?” My friend’s cousin suggested that we can’t ever really empirically settle the answer to that question; therefore, we would be better off redefining ‘conscious’ to mean something we can actually determine empirically (such as passing the Turing Test). Here’s why that’s a bad idea.
The way we ordinarily think of consciousness importantly involves things like:
having experiences,
having a first-person perspective, and
there being something it is like to be you.
Unfortunately, we can’t access those facts about anybody but ourselves. We can’t currently (perhaps ever) empirically test whether any other being is actually having experiences that are rendered in its own first-person perspective, or whether it merely acts like it does. That’s a version of “the problem of other minds”. That’s why it’s probably right to say that we cannot ever really empirically test for certain whether any given machine is conscious; we each can’t do so for anyone other than ourselves!
But this kind of empirically inaccessible consciousness is deeply important. Consider: an important reason we think it is fine to kick a rock but not to kick a puppy is that rocks don’t experience you kicking them, but we think puppies do. Kicking a puppy causes the puppy to experience suffering, and that experience of suffering requires consciousness. If the puppy was merely a mindless automaton that was not capable of having experiences but still acted hurt, we wouldn’t think that kicking it was wrong in the same way. (Though we might still think things like, “there’s something sadistic and wrong about wanting to induce even fake pain behavior”.)
Thus, many people think that having the capacity for this kind of empirically undetectable consciousness is necessary (and perhaps sufficient) for being the kind of entity that deserves moral consideration. If you are conscious in this empirically undetectable way, then you can experience things (e.g., pain, pleasure) and you have interests, desires, hopes, etc., that ought to be considered.
So, a problem with redefining consciousness in some other terms is that we run the risk of misclassifying which entities matter morally. If our new, empirical definition doesn’t give exactly the same answers to the question “Which entities are conscious?” as the empirically undetectable definition, then we’ll either attribute suffering to things that don’t really experience it or end up ignoring real suffering. The first of those outcomes could lead to unnecessary restrictions on our actions, while the latter could lead to moral catastrophe! By itself, this is enough to show that some questions that aren’t empirically answerable matter greatly. That’s enough to show that Practical Empiricists are wrong.
What’s the takeaway?
I have argued that Practical Empiricism threatens to impoverish your worldview: as we have seen in the arguments above, by committing most of the tools of “inference to the best explanation” to the flames, it can cut you off from realms of fascinating explanations, make it harder to reject conspiracy theories, and potentially even lead to moral catastrophe (by leading us to misclassify which entities matter morally)!
Along the way, we’ve seen a framework for rationally deciding between different answers to empirically unanswerable questions: the framework of "theoretical virtues". You can use this framework in your own life, to help guide your inferences towards the best explanations.
Much more can (and has been) said on this topic. There are lots more interesting arguments to consider, but regardless of your view, I hope it is no surprise that at Clearer Thinking we're passionate about the value of scientific thinking and empirical testing where possible. We believe that the methods of science are among the best methods we have for gaining knowledge. But I have argued that the methods of science include using the method of “inference to the best explanation” to answer questions that can’t be settled empirically. Hence, plenty of empirically unanswerable questions are useful, interesting, and even (sometimes) scientific.
If you enjoyed this article, you might enjoy our “Can You Detect Weak Arguments?” quiz. Find out how well you can spot misleading rhetoric and recognize your own bias using real-world examples:
