TL;DR: Knowledge isn’t satisfied by having only a justified true belief, as shown by Gettier problems, but it also requires a correct belief in how that justification explains the truth of that belief, what I call a “connector.” A person is said to know "X" by evidence "Y" because the connector "Z", reasonably justifies believing "X" on the basis of "Y". Additionally, skepticism’s conception of knowledge would require that the concept of knowledge itself be meaningless. To the extent propositions regarding knowledge are meaningful, we should disregard skepticism.
Philosophy uneasily defines knowledge as a justified true belief (JTB). JTB consists of three parts: (1) a belief being justified, (2) that belief being true, (3) that belief being held.
First, a belief would be considered “justified” if it has reasons (inferential or non-inferential) to warrant that belief (J).
Second, that belief must actually reflect an objective state of the world. No matter how reasonable or well-informed a belief is, you wouldn’t have knowledge unless that belief were true. (T).
Third, a subject would need to hold that belief. (B).
In short, if you have a justified belief and that belief turns out to be true, then you can be ascribed “knowledge” about what it is you believe. While 99.99% of people can accept that JTB constitutes knowledge, for the 00.01% who have heard of Gettier problems, JTB is insufficient.
Gettier problems are essentially situations where justified beliefs are true only by coincidence.
For instance, if you look at a clock that shows it is 2:30 p.m., you have a justified belief that it is 2:30 p.m. However, unbeknownst to you, the clock is broken. Yet by sheer chance, it also happens to be 2:30 p.m. when you look at the clock and form your belief. Your justified belief is true.
So, based on the JTB definition of knowledge, you would have had knowledge that it was 2:30 p.m., even though you got lucky relying on a broken clock. But this doesn’t seem correct. You relied on a bad piece of evidence to form a belief, and only circumstances led to that belief being true.
Philosophers have proposed all sorts of additional conditions in an attempt to resolve the Geitter problem (no false lemmas, reliabilism, absence of defeaters, etc.). Yet other philosophers have successfully refuted these alternative definitions—often by showing how these formulations of knowledge are either too broad or too narrow, classifying too many circumstances as knowledge that we wouldn’t intuitively call knowledge, or failing to classify circumstances that we would intuitively call knowledge as knowledge. They create a lot of false positives and/or false negatives.
So far, we don’t have an agreed-upon definition of knowledge that corresponds to our intuitions. Something is missing from the JTB theory of knowledge. Let’s see if we can do any better.
Signal and Substance
The two parts of JTB, the justified belief and truth, can be viewed as two separate concepts: the signal and the substance.
The “justified belief” is based on a signal, while the truth is the actual substance. The signal is sufficient evidence of the substance. If you see a signal, then it’s fair to believe the substance, although the signal itself is not the substance. Just like how light from a lighthouse isn’t actually a rocky, ship-wrecking coast, it’s still an accurate signal of one.
A signal is publicly observable and conveys a fixed piece of information to reasonable observers that would warrant a certain belief. A signal couldn’t be a private intuition but is more like the type of evidence that you can present to a jury.
For example, if you were to say that you could predict that it will rain tomorrow based on nothing more than a gut feeling (you wake up one morning and are convinced you have that power), and it ends up raining the next day, then you still wouldn’t actually have had knowledge that it would rain. Your belief that it will rain could have been genuine, but it wasn’t “warranted,” since your gut feeling wouldn’t be a “reasonable” justification. Who else but you could be convinced that your bodily sensations are proof of future physical events?
However, if your gut feeling were to predict rain the next day persistently and reliably, that intuition may eventually be a reasonable justification, and serve as a sufficient “signal” of the “substance,” of future rain. The non-observable private, and unreasonable signal would then become an observable, public, and even reasonable signal that would warrant the belief of future rain (some things, we can’t explain yet).
Sometimes, signals can be misleading. For instance, if I wanted to know when Henry Ford was born, I could go online and discover that it was July 30, 1863. But let’s say that Henry Ford was actually born on July 29th, 1863, although available written records reported that he was born on July 30th, 1863. Then I wouldn’t have knowledge of Henry Ford’s birth based on his Wikipedia page. Although I would still have a justified belief despite that belief not turning out to be true.
I wouldn’t need to have been directly present during Henry Ford’s birth to have knowledge of the “substance,” (the date of his birth). Rather, I can rely on a good signal.
Meanwhile, if Henry Ford’s Wikipedia entry also includes a quote attributed to Henry Ford, but was unwritten and reported only secondhand, with other sources questioning the validity of the quote, then the Wikipedia entry doesn’t seem to provide sufficient justification that Henry Ford actually said that quote. Some substances require stronger signals.
Sometimes signals are bad and don’t justify beliefs that turn out to be true. Sometimes signals are misleading and justify beliefs that aren’t true. Knowledge is only possible when the signal warrants belief in a true thing, although the signal and the substance alone are not sufficient to constitute knowledge, as Gettier problems show.
The Signal
The substance is pretty straightforward. It’s just something objectively true as corresponding to a fact in the world.1 Although the signal may require a bit more explaining.
The signal can be thought of as a sufficient condition for a warranted belief in the substance. If the signal exists, then believing in the substance would be justified.
Once you have the signal, you can have a justifiable belief. I emphasize “can” because people are presented with signals all the time, but may still not have the belief that the signal would warrant, for one reason or another (personal biases, cognitive dissonance, etc.). Yet the signal at least allows a belief to be justified.
A factual substance can have multiple signals. If multiple reputable news sources report on and show visual evidence of an event, each of those news sources could be a signal that would warrant a belief in that event.
The above raises the question: what level of probability would the signal need to ensure that the substance exists? It depends on the context (see contextualism). In some contexts, saying that I know something may require a lot of evidence and taking little for granted, whereas to know something in another context may be flexible with our assumptions.
If I search online to check whether a restaurant is open, and then I see that Google shows me that the restaurant is currently open, the report by Google may be a good enough signal that would allow me to know whether it is open, although it doesn’t guarantee that fact.
Google may not properly reflect the restaurant’s hours of operation, which I know based on my previous experiences of disappointingly traveling to venues that Google falsely reported were open. Google may therefore be only a weak signal that the restaurant is open.
However, if I were to call up the restaurant, recognize the voice on the other end as an employee of the restaurant I know, and hear the employee state that the restaurant is open, then the employee’s testimony would be a very strong signal that the restaurant is open.
Therefore, whether I can say that I “know” the restaurant is open depends on how strong I need a signal to be in a certain context. If I’m just casually considering going to the restaurant, which, based on my experience, is usually open at the time in question, and there would be no real consequence if the restaurant wasn’t open, then the Google check would be fine.
But if it’s a holiday and I have relatives visiting who really want to try this restaurant, then a Google check likely wouldn’t be enough for me to confidently say that I know the restaurant is open. I may need to call first to have a signal that would justify my belief that the restaurant is open.
This contextual understanding of knowledge is reflected in our legal system. In American criminal courts, the prosecution must satisfy a “beyond reasonable doubt” standard for guilty rulings (commonly understood as greater than 90%). Meanwhile, in civil courts, the plaintiff only needs to show a “preponderance of the evidence." This is because the defendant’s liberty is at stake in the former case, whereas only property is at stake in the latter.
Our legal system operates with two different standards of knowledge, depending upon what’s at stake.
As Edmund Gettier showed us with his dilemmas, we can have justified signals for true substances without knowledge. I argue here that while JTB only incorporates the signal and the substance, it misses the relationship between the two, what I will call the “connector.”
The Connector
The connector is the explanation that ties the signal and the substance. To have knowledge, we would need to correctly believe that a connector exists. I’ll call this definition of knowledge a justified belief connected to a truth (or JBCT). The only difference between JBCT and standard JTB is the connector condition.
Imagine a bright, star-like signal at one end connected to a dimmer, more mysterious substance at the other end. We may not have direct experience of the dimmer substance, but we may for the publicly observable signal.
As its name suggests, the connector connects the signal to the substance. It is the explanation of why the signal warrants the substance. For the connector condition to be satisfied, (1) the connector must exist, and (2) the subject must believe in the connector, thus understanding how the evidence actually relates to the truth. Only with a true belief in a connector between the signal and the substance, (or the justified belief and truth), could a subject state that he has knowledge.
Apply the connector requirement to the classic “clock” Gettier problem example provided above, where you relied on a broken clock to accurately believe that it is 2:30 p.m. In that case, there is no connector between the time at that moment (the substance) and the clock (the signal). You may believe that there is a connector because clocks tend to track time. However, because that connector doesn’t truly exist (since the clock is broken and therefore unrelated to the time), the connector condition isn’t satisfied. Therefore, your belief that it is 2:30 based on a broken clock doesn’t constitute knowledge.
And connectors can be of all different types. In the “gut feeling of rain” example, the connector would be one of timing and reliability. While we may not be able to explain the relationship between your bodily sensations and meteorology, reliable past experiences (which, given the mysterious relationship between the signal and substance, would need to be very reliable) can serve as a connector.
In the “Henry Ford Birthday” example above, Wikipedia is connected to Henry Ford’s date of birth by the ease of certification and recordkeeping of birthdays as well as the ease of correcting any inaccuracies on birthdays on Wikipedia. An easily verifiable fact can be expected to be found on Wikipedia, whereas quotes reported secondhand are subject to more doubt. So while I may be using a reliable process to obtain a signal, I would also have to know why the signal is accurate.
Let’s take another example. Say someone were to stop me at gunpoint and demand that I agree that his socks are orange, which is in fact true. However, I have no idea what color my attacker’s socks are, but I placate and say that I believe his socks are orange. I have a justified belief, (my not wanting to be killed should be justification enough). And that belief happens to be true. But there is no connector. The justification for my belief does not sufficiently explain its relationship to the applicable truth. Therefore, I don’t have knowledge that my attacker’s socks are orange. Not until I look at his socks that my visual experience could serve as that connector.
So to have knowledge, there must be:
a justified true belief based on a signal,
a connector between the justified belief and the truth (where the connector is the fact that explains the relationship between the signal and substance so that the signal would be a sufficient condition of the substance), and
a belief in the connector.
Condition (1) is JTB while conditions (2) and (3) are conditions of the connector.
In sum, knowledge is reasonably believing that a true thing is true, and correctly believing how the evidence which warrants that belief is connected to the true thing.
On its face, this seems like a high bar for knowledge. However, I believe that connectors are fairly intuitive. To know something is true based on evidence, you would also need to explain how that evidence connects to that truth accurately. Knowledge isn’t about reiterating facts but knowing why something is a fact.
To address a concern with JBCT, it seems like knowledge is being pushed off to the connector. Wouldn’t we need to know about the connector, which itself would require a connector? Yes, knowledge assumes other knowledge. If you keep on asking “why?” to something, eventually, you’d get down to self-evident axioms that are not justified by anything else (i.e., the “foundations” in “foundationalism”). So when we say, I know the substance “X” because of the signal “Y,” and “Y” warrants the truths of “X” because of the connector between them, “Z”.
How do you know “Z”? It may be because of “A”, “B”, and “C.” All justifications for beliefs would need to assume some background knowledge.
“I know that a mug is on the table.”
“How do you know that you can rely on your perception to say that a mug is on the table?” We would have to fall back on certain foundational beliefs—beliefs that don’t require justifications themselves—at some point.
A connector is only a justification of a certain kind —the relationship between the signal and the substance. The connector isn’t too demanding or complicated. We would either know this relationship between the signal and the substance, or we wouldn’t. And we may place different standards on our connectors in different contexts.
In some contexts, we demand a “conenctor” to be infallible, like in mathematics. In other contexts, a “connector” may provide much less certainty, like the social sciences.
Really, the only difference between JTB and JBCT is that JBCT requires that (1) there is a certain justification for the belief (the explanation between the signal and substance) and (2) that justification also be true.
And the connector, like justified beliefs generally, would have to rest on foundational beliefs.
How JBCT Relates to Other Definitions of Knowledge
This connector is similar to the causal theory of knowledge, where a belief of X constitutes knowledge of X if the fact of X causes the belief of X. The causal theory replaces the “justification” requirement in JTB with a “cause” requirement.
However, in the JBCT argument I make here, the connector doesn’t have to be causal. And JBCT still retains the “justification” requirement (whereas the causal theory removes it). That justification must only provide how the evidence used to arrive at the belief explains the truth of the belief. And that justification itself must be true.
JBCT is also similar to reliabilist accounts of knowledge, where true beliefs constitute knowledge if those beliefs were formed through a reliable process. Yet unlike reliablism, the relationship between the signal and the substance must actually be true, not just generally true. JBCT also adds in the requirement that the subject correctly believes how the reliable process relates to the true fact.
JBCT is also similar to the “no false lemmas” response to Gettier problems, where knowledge is classified as JTB where a belief is not inferred from something false. JBCT is a bit more specific, requiring only that the connector not be false.
Overall, JBCT is somewhat broader than the causal theory, since connectors don’t need to be causal—just any reasonable explanation which ties the signal to the substance. It is somewhat more restrictive than the reliability theory since JBCT requires the subject to correctly believe why a reliable process justifies a belief. And it’s more specific than “no false lemmas” since it requires that a connector be both true and believed in.
On Skepticism
As an aside, one can claim that we can never have “knowledge,” since we can’t have any justified beliefs. We can’t exclude the possibility that we are living in a simulation or are brains in vats (which I will call “illusion theories”) since our perceptions and sensations are entirely compatible with what we call reality being only an illusion. Therefore, the mind-independent “substance” can never be truly knowable. This is the argument for skepticism and it’s pretty strong.
However, skepticism is a cheap dodge at the problem of knowledge. Under skepticism, where all of our beliefs are unjustified, the concept of knowledge couldn’t meaningfully exist. Because illusion theories are unfalsifiable, beliefs could never be justified.
So a skeptic would say that we could never have knowledge, since no amount of evidence could warrant entirely disproving illusions theories. But what would knowledge mean if it’s something that is held, but that’s impossible to hold? The skeptic’s conception of knowledge is self-defeating.
In order for us to be able to communicate, there has to be a shared understanding of sensible concepts. We can’t use terms that are meaningless to convey meaningful sentences. If a skeptic argues in favor of an illusion theory that cannot be rebutted, so that none of our beliefs can be justified, then we can’t have knowledge under skepticism.
Under skepticism, knowledge itself would be a meaningless concept, since there isn’t a world where someone could claim that they have knowledge. This means that questions that inquire about knowledge, like “Do you know what color my shirt is?” or “Do you know who won the 2020 Presidential Election?” aren’t real questions since the only possible answer is no. And simply stating “no” conveys no new information. But clearly, we ascribe meaning to these questions, and answers to them can convey new information.
Answering these inquiries with “No, because we could be brains in vats” is itself not meaningful, whereas responding based on the context in which the question was asked does convey the information that the questioner was looking for.
For us to have a meaningful discourse on knowledge, knowledge must mean something obtainable—something that is possible to possess or not possess. And our discourse shows that we can know things with “reasonable certainty.”
It would be like if we had a term called “Schmowledge,” defined as information that no one could ever possess (or schmow). The only correct way to use “Schmowledge” is to make statements like “I don’t have Scholwedge about X” or “He doesn’t Schmow anything about Y.” Schmowledge can never be possessed, and it would never make sense to ask about it, since no one, by definition, schmows anything. This would make the very concept of Schmowledge meaningless, and the term itself can be eliminated from our vocabulary without consequence. But the same can’t be said for knowledge. It’s clearly not a meaningless concept, so let’s not understand it in a way to make it meaningless.
Just as determinists argue that free will does not exist because their concept of free will would be impossible (which they may define as not being subject to the laws of causation—see my argument for free will here), similarly, skeptics’ conception of knowledge make any claims to knowledge conceptually impossible.
Some people aren’t satisfied with this contextualist response to global skepticism. Winning by semantics is never satisfying. However, the consequences to public discourse under the skeptic’s idea of knowledge are unacceptable. If we can use the term “knowledge” in a meaningful way, then it can’t be true that the term “knowledge” lacks meaning as the skeptics’ conception of it would conclude. If knowledge is an idea that’s worth keeping around, we should adopt an epistemology that is compatible with it.
Intuitively, you might ask why this matters. Isn’t this just philosophers spilling ink over a topic whose answer has no consequences? Everyone basically knows what “knowledge” means through ordinary parlance, so why does it matter if we have a meaning that satisfies the philosophy geeks?
Defining our terms is critical when discussing more complicated concepts. Too often, philosophical debates get stuck as participants use the same term to mean different things. (See my discussion on what morality is here.)
Given our exciting age of change and uncertainty, a clear definition of knowledge would help us distinguish between the charlatans who got lucky in their predictions on politics, COVID-19, or AI and those to whom we can truly ascribe knowledge.
Conclusion
I hope this article has satisfied the 00.01% of people who care about Gettier problems. But even for everyone else, I hope I at least explained what I consider to be a “justified belief” and “knowledge” reasonably.
I don’t expect this article to have resolved Gettier problems and the decades-old debate over the definition of knowledge. JBCT has worked fine in the classic Gettier cases I have considered, at least upon my own examination. However, I’m looking forward to reading their counter-examples where even having a JTB and a connector wouldn’t constitute knowledge, or where having a JTB without a connector would still constitute knowledge. Discourse and feedback are the purposes of this substack after all. But I do hope that the “connector” gets us one step closer to a complete conception of knowledge.
Or fits in our mind-dependent socially constructed model of the world, if you don’t believe in a mind-independent reality. Definitions of truth are subject to debate and warrant a separate article.
UPDATE: unfortunately, you have fallen for the same confusion as the other proponents of your theory. You’re confusing the input of algorithm with the algorithm itself. The context affects the INPUT of the algorithm, without affecting the ALGORITHM itself. However, you haven’t gave much detail beyond “the output depends on the input.”. The problem is that there is cluelessness about the algorithm itself IN ADDITION to cluelessness about its inputs. The connectedness requirement looks less like an independent vindication of our epistemic intuitions about particular cases, and looks more like an attempt (successful or otherwise) to reverse engineer a general principle to accommodate our epistemic intuitions about particular cases. As a result, I would NOT be surprised if justified true belief really is sufficient for knowledge, despite of all our ( and Edmund Gettier’s) epistemic intuitions about particular cases to the contrary. As a result, I want to suspend judgment (at least temporarily) about which theory of knowledge is true.
There’s a trilemma regarding the relation between a belief being true on one hand, and being connected to the truth Eiither being true is:
Horn #1: SUFFICENT to be connected to the truth, in which case we’ve returned to the justified true belief theory of knowledge, or
Horn #2: UNNECESSARY to be connected to the truth, in which case a justified false belief could (at least in principle) count as knowledge, or
Horn #3: NECESSARY but INSUFFICIENT to be connected to truth, in which case it is hard to distinguish justified true belief without knowledge from unjustified true belief.
Also, some proponents of your theory of knowledge (possibly but not necessarily including you) insist that whether or not a justified true belief is connected to the truth depends on context. However, those proponents of your theory of knowledge (other than you) confused the input of the algorithm with the algorithm itself. For example, for the vast majority of those algorithms that have input, the output of the algorithm depends on the input of the algorithm. Nonetheless, if the algorithm itself is known, then any remaining cluelessness about the output of the algorithm would be reducible without residue to cluelessness about the input of algorithm. Unfortunately, those other proponents of your theory of knowledge haven’t gave any additional information about the algorithm itself. I’ll check your article to see if you do any better.