This post introduces the SOAR Cognition Loop. The post also explains how SOAR, when combined with analogy and introduction, and be used to automate human thought.
I compare SOAR, developed by the UNL educational psychologist Dr. K, to OODA, devised by the late USAF Colonel John Boyd. While SOAR comes from an educational psychology background, and OODA from a fighter pilot’s perspective, both are applicable to almost all of human thought.
The main difference between SOAR and OODA is that OODA focuses looks at how human thought is caused by outside forces, while SOAR looks at how humans actually think. In modeling lingo, SOAR is a logical representation while OODA is a physical representation.
And a last note before I begin: any mistakes, misunderstandings, or incorrect assertions in this post are mine, not Dr. K’s. I developed this post from my memories of his lectures, and I hope I captured and applies his ideas correctly.
SOAR has four parts
- Selection: Observing a text
- Organization: Creating internal connections within a text
- Association: Creating external connections, between parts of the text and prior knowledge
- Regulation: Testing the validity of the connections
This post focuses primarily on Organization and Association: connectivity building. As such, Selection and Regulation are glossed over. Interestingly, Dr. K’s Selection phase is more developed than Col. Boyd’s Observation phase, but as they say, that’s a post for another time…
SOAR can be diagrammed as:
Note that Association is optional. Internal connections are naturally developed because humans are pattern-seekers and pattern-recognizers. External connections are typically frail and loose, because they are expensive to create.
To see how SOAR, particularly Organization and Association, work in practice, consider the following bit of text:
Guerrilla warfare is old. It was used during the American Revolution and during the Iraq War. In 1775, after the Battles of Lexington and Concord, guerrillas hid in the trees. In 2005, guerrillas often use Improvised Explosive Devices (IEDs).
In a future post I will discuss the how the information in the above paragraph is Selected and Organized. Suffice it is say that the above paragraph may be Selected and Organized as:
This creates an organization of meaning, a “semantic network.” I earlier discussed semantic networks and how they relate to quality.
But this semantic net does not exist in isolation. As it is fed into long-term memory, it is appended onto the learner’s existing knowledge: his already developed semantic internet. In almost all cases, the optional Association subloop is skipped or just softly entered. For instance, the learner may make some external connections. In this case, consider what the learner already knows about the American Revolution and the Iraq War
so the semantic networks combine, Associate to give a better picture:
In turn, memory — the learner’s semantic internet — will further Associate the information. Along the lines of, “If the Battle of Lexington was in 1775, and the American Revolution was in 1775, then the Battle of Lexington was in the American Revolution”
This makes comprehension and good memories more likely. The “rich get richer” in learning, because the more background knowledge the learner has, the better he is able to learn new things.
Still, few insights have come from the passage. The added thoughts are automatic and vertical, not intuitive and horizontal. The added information is close to useless by itself, but this rarely matters at the moment information is learned, so regulation is successful (because failure is not observed) and the loop continues with new input.
However, if we want to change the learner’s behavior, we need to get him to associate the new knowledge with old knowledge in a meaningful way. Look above again at the semantic network the learner built from the passage. And now select another part of the learner’s existing knowledge, his semantic internet. Let’s say he already knew that the Battles of Concord and Lexington took place in New England:
and that the Iraq War takes places in Iraq:
Now he can Associate that with he new knowledge, adding a little meaning and making the information a bit more memorable. Just like before we can Associate these two Semantic Networks to grow the learner’s existing knowledge. But pay special attention to some connections in particular (highlighted in Red for your convenience). These connections are arbitrarily chosen, but this post will give an example of human thought using them:
Now the learner can use “triz” contradiction and analogy to develop new ideas, and make the passage much more memorable.
There’s a lot of relations to go through, but fortunately the human mind is a great parallel processor, so the learner can quickly match up two elements to create an analogy. To Associate these creatively, let’s create an analogy network with the highlighted connections
Now, to automate creative thought, to easily create new Associates that others have not, all we have to do is run through the analogy network, swapping out “is like” for some other node in our semantic internet that seems likely. For instance, instead of “Trees Are Like IEDs”
try an easy one: swap out “is like” for “Guerrilla Warfare”
And indeed, it’s true. Trees and IEDs are both used in Guerrilla Warfare. However, because the analogy has not yielded a contradiction, nothing creative has been learned. Let’s try again. Instead of “New England Is Like Iraq”
swap “Is Like” out with “Americans,” a node from our semantic internet that was added only after a previous Association
This is new information. This is what some would call “vertical learning.” We created an analogy based off a short paragraph and previous information, and suddenly know there are Americans in both New England and Iraq. Our learner could quickly verify this — and indeed it is a straightforward deduction from his new and old experience.
But still, this wasn’t too creative. All we did was tie trivia together in a Wikipedia-like manner. Creativity sure is hard! So many analogies already, and we just have information that is either obvious or quickly verifiable.
OK, one last one. For the New England : Iraq Pair, let’s swap out “Is Like” with another element from our semantic internet. So instead of:
ERROR! Regulation failure! Good!
The Iraq War has IEDs. The American Revolution is like the Iraq War. But the American Revolution did not have IEDs.
The failure of Regulation — the last step in SOAR — when we created this analogy is wonderful news. We can now think horizontally, Associate, and build new meanings.
Recall from just moments before that we compared Trees and IEDs. If the analog of IEDs, trees, also fails in our attempted analogy, we may learn something new about guerrilla warfare. So let’s try it: replacing “Is Like” with “Trees”
ERROR! Regulation failure! Double-Plus Good!!!
When we think of the Iraq War, we do not think of trees, even though the Iraq War is like the American Revolution, and the American Revolution prominently featured trees.
We know that the American Revolution is like the Iraq War, so there has to be a reason for these contradictions. We need to know what is the defect in the Trees-Guerrilla Warfare -IEDs chain.
We are now in the most expensive part of creativity. We need to Associate the portion of the semantic internet we have been dealing with to an unknown portion of our total knowledge. This requires a tremendous amount of parallel processing — it requires so much parallel processing that the best parallel computer yet devised – the human brain – often fails to complete the task of resolving contradictions into higher-order analogies.
(Fortunately, as the price of computing components continues to fall, we are approaching the day when we will be able to outsource creativity to computers just as we outsource complex mathematics to computers. This is a good thing. Economics tells us that when humans lose their comparative advantage in creative thought to machines, this will allow us to maximize production by focusing on something else.)
We have no logical place to start our exhaustive look for a higher order Trees : IEDs analogy that allows us to solve the contradiction that IEDs were not used in the American Revolution, and trees are not prominently used in the Iraq War. Nor can we use a nifty sorting algorithm, because knowledge has no natural beginning or end.
To avoid a self-referential paradigm , we will begin with a random semantic net already found in long-term memory, and try to tie it pack to the initial analogy network
So one has to start with something arbitrary. Why not “economics”? Say this is what we know about economics:
After attempting to associate these elements to our analogy network, and being frustrated by so many contradictions dead ends, our automated cognition system finds a very promising match.
“Humans” can be tied to almost every element. Because we are focusing on analogies, we try to make all the analogies parallel. Thankfully, we are able to do this. Otherwise, we would have to label this as yet another contradiction, and have even more calculation to do!
So far so good! We are getting closer to solving an contradiction and creating a higher-order analogy — in other words, original thinking — through our manipulations of semantic networks.
So using our random hunch that the solution is economics, let’s cycle around economics finding things that are also related to humans that can solve our contradiction with respect to IEDs, Trees, the American Revolution, and the Iraq War .l (If economics does not work, we will arbitrarily try some other concept. Because there are no true contradictions, an analogy will be discovered eventually. It just may take forever).
First, let’s randomly try “Adam Smith”:
But if we would go through these new analogies, we get a new contradiction: Adam Smith was alive during the American Revolution, but not the Iraq War. We will throw this new contradiction to the end of our stack of problems to solve later. Instead, we’ll try another element that’s related to both humans and economics
Profitability is more promising… the profit motive exists in both New England and Iraq, it exists during all wars, and the purchase of both IEDs and trees are governed by profitability concerns. Visually:
Now let’s take Profitability, and throw it into the analogy (“The American Revolution Is Like the Iraq War”), and the chain (Trees : Guerrilla Warfare, Guerrilla Warfare : IEDs) that gave us the contradictions in the first place.
Now find internal connections — Associate them
And then, build the analogies. Same way as before: swap out the relations with the new element; profitability.
Now, do the new chains
- American Revolution : Profitability : Guerrilla Warfare
- Iraq War : Profitability : Guerrilla Warfare
- Trees : Profitability : Guerrilla Warfare
- IEDs : Profitability : Guerrilla Warfare
In clearer English, we could say
- The American Revolution continued as long as it was potentially profitable for both sides to fight
- The Iraq War continued as long as it was potentially profitable for both sides to fight
- Trees are used in Guerrilla Warfare as long as it is potentially profitable to do so
- IEDs are used in Guerrilla Warfare as long as it is potentially profitable to do so
These are new facts. It is not even hinted at in the original example paragraph:
Guerrilla warfare is old. It was Patriots during the American Revolution and during the Iraq War. In 1775, after the Battles of Lexington and Concord, guerrillas hid in the trees. In 2005, guerrillas often use Improvised Explosive Devices (IEDs).
Nor is it obvious — except in the sense that it is obvious after it is known. Further, it is useful. It shows mathematically how to end the use of IEDs against soldiers
IEDs are used in Guerrilla Warfare as long as it is potentially profitable to do so
And of course, it gives the obvious hint to the British of how they could have avoided the costly retreat from Lexington: cut down the trees.
I hope this post is a good demonstration of the power of internal and external connectivity — Organization and Association — when combined with contradiction and analogy. As I mentioned, it left out almost all talk of Selection and Regulation. Implicitly, this post also used only the SOAR concept of representation, ignoring hierarchies, sequences, and matrices. But those are for the future 😉