The Open Mind

Cogito Ergo Sum

Knowledge: An Expansion of the Platonic Definition

with 12 comments

In the first post I ever wrote on this blog, titled: Knowledge and the “Brain in a Vat” scenario, I discussed some elements concerning the Platonic definition of knowledge, that is, that knowledge is ultimately defined as “justified true belief”.  I further refined the Platonic definition (in order to account for the well-known Gettier Problem) such that knowledge could be better described as “justified non-coincidentally-true belief”.  Beyond that, I also discussed how one’s conception of knowledge (or how it should be defined) should consider the possibility that our reality may be nothing more than the product of a mad scientist feeding us illusory sensations/perceptions with our brain in a vat, and thus, that how we define things and adhere to those definitions plays a crucial role in our conception and mutual understanding of any kind of knowledge.  My concluding remarks in that post were:

“While I’m aware that anything discussed about the metaphysical is seen by some philosophers to be completely and utterly pointless, my goal in making the definition of knowledge compatible with the BIV scenario is merely to illustrate that if knowledge exists in both “worlds” (and our world is nothing but a simulation), then the only knowledge we can prove has to be based on definitions — which is a human construct based on hierarchical patterns observed in our reality.”

While my views on what knowledge is or how it should be defined have changed somewhat in the past three years or so since I wrote that first blog post, in this post, I’d like to elaborate on this key sentence, specifically with regard to how knowledge is ultimately dependent on the recall and use of previously observed patterns in our reality as I believe that this is the most important aspect regarding how to define knowledge.  After making a few related comments on another blog (https://nwrickert.wordpress.com/2015/03/07/knowledge-vs-belief/), I decided to elaborate on some of those comments accordingly.

I’ve elsewhere mentioned how there is a plethora of evidence that suggests that intelligence is ultimately a product of pattern recognition (1, 2, 3).  That is, if we recognize patterns in nature and then commit them to memory, we can later use those remembered patterns to our advantage in order to accomplish goals effectively.  The more patterns that we can recognize and remember, specifically those that do in fact correlate with reality (as opposed to erroneously “recognized” patterns that are actually non-existent), the better our chances of predicting the consequences of our actions accurately, and thus the better chances we have at obtaining our goals.  In short, the more patterns that we can recognize and remember, the greater our intelligence.  It is therefore no coincidence that intelligence tests are primarily based on gauging one’s ability to recognize patterns (e.g. solving Raven’s Progressive Matrices, puzzles, etc.).

To emphasize the role of pattern recognition as it applies to knowledge, if we use my previously modified Platonic definition of knowledge, that is,  that knowledge is defined as “justified, non-coincidentally-true belief”, then I must break down the individual terms of this definition as follows, starting with “belief”:

  • Belief = Recognized patterns of causality that are stored into memory for later recall and use.
  • Non-Coincidentally-True = The belief positively and consistently correlates with reality, and thus not just through luck or chance.
  • Justified = Empirical evidence exists to support said belief.

So in summary, I have defined knowledge (more specifically) as:

“Recognized patterns of causality that are stored into memory for later recall and use, that positively and consistently correlate with reality, and for which that correlation has been validated by empirical evidence (e.g. successful predictions made and/or goals accomplished through the use of said recalled patterns)”.

This means that if we believe something to be true that is unfalsifiable (such as religious beliefs that rely on faith), since it has not met the justification criteria, it fails to be considered knowledge (even if it is still considered a “belief”).  Also, if we are able to make a successful prediction with the patterns we’ve recognized, yet are only able to do so once, due to the lack of consistency, we likely just got lucky and didn’t actually correctly identify a pattern that correlates with reality, and thus this would fail to count as knowledge.  Finally, one should also note that the patterns that are recognized were not specifically defined as “consciously” recognized/remembered, nor was it specified that the patterns couldn’t be innately acquired/stored into memory (through DNA coded or other pre-sensory neural developmental mechanisms).  Thus, even procedural knowledge like learning to ride a bike or other forms of “muscle memory” used to complete a task, or any innate form of knowledge (acquired before/without sensory input) would be an example of unconscious or implicit knowledge that still fulfills this definition I’ve given above.  In the case of unconscious/implicit knowledge, we would have to accept that “beliefs” can also be unconscious/implicit (in order to remain consistent with the definition I’ve chosen), and I don’t see this as being a problem at all.  One just has to keep in mind that when people use the term “belief”, they are likely going to be referring to only those that are in our consciousness, a subset of all beliefs that exist, and thus still correct and adherent to the definition laid out here.

This is how I prefer to define “knowledge”, and I think it is a robust definition that successfully solves many (though certainly not all) of the philosophical problems that one tends to encounter in epistemology.

Advertisements

12 Responses

Subscribe to comments with RSS.

  1. I’ve elsewhere mentioned how there is a plethora of evidence that suggests that intelligence is ultimately a product of pattern recognition (1, 2, 3).

    There isn’t much evidence at all, as best I can tell. I’ll grant that it is a view that is widely believed. At one time, I held that view. But, as I studied the issues, I realized the serious problems facing that view.

    There’s a lot of machine learning research based on that view. Thus far, it has not produced anything resembling human learning.

    Best wishes on this. I won’t be arguing with you about it, because it’s clear that you won’t be persuaded.

    Neil Rickert

    March 10, 2015 at 8:45 pm

    • The evidence speaks for itself and is plentiful. Research in neuroscience as well as AI have thoroughly demonstrated it, as well as the very tests we use to test for general intelligence. They all support the idea that intelligence is a result of pattern recognition. Cheers!

      Lage

      March 10, 2015 at 11:48 pm

  2. Non-Coincidentally-True = The belief positively and consistently correlates with reality, and thus not just through luck or chance.

    What does it mean to say that a belief (i.e. a recorded pattern) correlates with reality? How would this be tested?

    Neil Rickert

    March 10, 2015 at 8:51 pm

    • A belief (pattern) correlates with reality if it allows one to use it to make predictions successfully, even if just in order to accomplish a particular goal. The better it is at allowing us to do this, the more correlated with reality it is. That’s the simple way to put it anyway, and how it would be tested (consistent, successful predictions? Or not?)

      Lage

      March 10, 2015 at 11:52 pm

      • Given a pattern (whatever that is – no definition has been provided), how do you use that pattern to make predictions?

        Neil Rickert

        March 11, 2015 at 6:09 am

  3. “Confirmation bias.”

    Explain how my pointing out that the pattern recognition correlation with intelligence is supported by the findings in neuroscience (related to the neo-cortex in particular), AI research, and that it is implied by the way we design tests to quantify intelligence, suggests that I’m guilty of confirmation bias. I’m all ears…

    Lage

    March 11, 2015 at 8:41 am

    • You seem to have a warm fuzzy feeling that something that can be vaguely called a pattern (but which is undefined) has some sort of relation that can be vaguely called correlation (also undefined). You need clear definitions that allow you to get at the causal structure.

      Neil Rickert

      March 11, 2015 at 4:32 pm

  4. “Given a pattern (whatever that is – no definition has been provided), how do you use that pattern to make predictions?”

    For example, if my brain recognizes the pattern that every time I release an object from some non-zero height it falls to the ground, I can predict that the next time I release the ball from a non-zero height, it will again fall to the ground. Thus, I would successfully “know” a pattern that explains this causal event and assists me in making predictions and accomplish goals (in this case goals pertaining to dropping objects). I would classify the recognition and recall of this pattern as “knowledge” as it fulfills the definition I gave.

    Lage

    March 11, 2015 at 8:53 am

    • For example, if my brain recognizes the pattern that every time I release an object from some non-zero height it falls to the ground, I can predict that the next time I release the ball from a non-zero height, it will again fall to the ground.

      Okay. But that is better described as reinforcement learning, rather than pattern recognition.

      I’ll agree that reinforcement learning works. What I question is the more general pattern recognition that you seem to call on.

      With reinforcement learning (as in your example), the pattern is a pattern in your behavior, though it still needs a definition of “pattern”.

      That’s one example. It falls far short of what is needed to explain what you mean by “true” for whatever it is that you mean by “belief”.

      Neil Rickert

      March 11, 2015 at 4:27 pm

  5. Okay. But that is better described as reinforcement learning, rather than pattern recognition. I’ll agree that reinforcement learning works. What I question is the more general pattern recognition that you seem to call on. With reinforcement learning (as in your example), the pattern is a pattern in your behavior, though it still needs a definition of “pattern”. That’s one example. It falls far short of what is needed to explain what you mean by “true” for whatever it is that you mean by “belief”.

    Reinforcement learning is just an example of a process that utilizes pattern recognition. In any case, it is the pattern recognition that makes it possible (e.g. “If I do X, based on a pattern I’ve learned, this will result in Y, which is my goal”). Also, there are many different levels of complexity for patterns that we identify. Some are 3D and some are 4D, for example, where our visual perception of a stationary object in our frame of reference is 3D and only requires things like shape, texture, color, and brightness patterns (which are themselves composed of more fundamental patterns such as vertical, horizontal, diagonal, and curved lines). On the other hand, our visual perception of a falling object (as in my example given previously) is effectively adding on the component of time introducing causal patterns (which are really just large sets of incrementally different 3D stationary patterns). So I first recognize the ball and environment I’m in, mostly based on 3D patterns, and then I recognize how those patterns change over time, which is itself a pattern, one with a temporal dimension.

    My example didn’t fall short at all. I gave an example of a recognized pattern, and how the predictive capability (when demonstrated successfully) illustrates that it is true, as it correlates with reality (demonstrated by the successful predictions made with the recall of such patterns). In my example, a person gains the “belief” (as I defined it) that a ball will drop if released from a non-zero height. This belief consists of recognized patterns (and many different patterns at that, with varying levels of complexity), as per my definition.

    Lage

    March 11, 2015 at 5:16 pm

  6. You seem to have a warm fuzzy feeling that something that can be vaguely called a pattern (but which is undefined) has some sort of relation that can be vaguely called correlation (also undefined). You need clear definitions that allow you to get at the causal structure.

    I’ve already given some examples of what I mean by “pattern”, as well as examples of how they can be shown to correlate with reality (predictive capability), which should be adequate, combined with my definition of knowledge given in the post.

    Lage

    March 11, 2015 at 5:19 pm


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: