Siderea

Agreed, this is something that has pissed me off for quite some time. Our schools actually teach kids NOT to think instead of HOW to think or how to learn. Just sit there and wait quietly until teacher explains...or complain and whine about not knowing how until she does it. It's rather disgusting.

I'm guessing I probably do use these categories...but not consciously. I think I just relate things to other things I understand, or use examples.

My biggest problem in learning (albeit quite small) is knowing what I'm supposed to be learning. All through undergrad I didn't feel like I learned a thing...all the concepts discussed were just commonsense with a fancy name attached to give some theorist an ego boost that he "originated" the idea.

I don't know. I learn just fine when I can put logic to it all. If I'm told "that's just the way it is" I'm not as apt to retain. When I was trying to learn C++ that was the problem I had..not with the programming, but that we had to start by invoking a library. It was never explained how the different libraries were defined and when I would need one versus another, and I had trouble simply trusting doing what they said.

In addition to the patterns you list, here are some of the patterns I've used recently. Philosophy, incidently, has dictionaries of these things (I think we'd call them "methods"). The Fifth Discipline also has a good collection. I might even claim that the underlying goal of all postmodern academic study is not knowledge, but discovering the contextified categories of sensemaking.

* Perceptual Patterns:
* the self-similar: Lots of natural systems organize themselves so they have self-similar processes at different scales.
* the chasm/blindside: Any body of understand that applies itself to the world incorporate a bit of magic to make its ideas applicable to the world (causality and faith are two such), and it will have a fundamentally important blindness around that jump.
* The inside, the outside, and the boundary: It's often useful to divide up the body into three segments. Something "core", something "antithetical", and the relation between them, and understand each individually.
* the rational and the irrational: these are complimentary ways to understand a thing. The rational is discrete and well-defined; the irrational is perceptually immediate and incomprehensible but modelable. The rational in/out/between division is One-Two-Many; the irrational division is Zero-One-Infinity.
* foreground and background: Is the a main subject and then supporting structure? If so, what happens when you switch the two?
* content and form: Does it make sense to think about there being a material here that acts very differently from the manner in which its contained? If so, to *do* better, concentrate on the content; to *understand* better, find the structure in the form.
* The Branches of Learning:
* the philosophy: Is the action here in ways of understanding, instead of the actual stuff being understood? Then concentrate on the form instead of the content of the materials.
* the applied math: Can this be well modeled as a behavior of equations? Huge branches of science and engineering are understandable as just the consequence of equations (note: equations are another category of sensemaking-structual-templates), and the way to understand them is to concetrate on the equations origins and surprising consquences, and not try to understand the whole from every "term" individually.
* the engineering: Is this all being driven by the desire to solve a problem? Then it probably is built as a many-layered collection of tools (functions) to solve different sub-problems that come up. A good way to learn this is by doing exercises (problems) and finding what tools you need.
* the liberal arts: Is the intended use of the thing to be made sense of to have an effect on you? Then let it have its effect, and study the thing as a different points of possible inspiration by concentrating on yourself-study-the-thing.

Nearly all my mental models are graphs or networks, whether I'm learning my way around a city, writing a game, or maintaining a computer system. Nodes are things: places, people, organizations, computers, programs, files. Edges are connections: routes, relationships, dependencies, dataflow, wires. How aware I am of my model varies a lot; the more clearly I can see it, the more comfortable I am with what I do and don't know. Getting started building a graph is hard -- there's a bunch of flailing about while I try to assemble enough pieces to put together into a sensible infrastructure that I can hang things off of. Once I've got that, it's all about finding new information and integrating it into the graph, which can range from the really obvious adding a new place and some ways to get there onto a mental map to gutting a node (or set of nodes and edges) and replacing it with a complicated sub-system with internal and external linkages of its own.

There are a number of known bugs.

My representation of time is lousy. When something happened or how long it took is data I rarely store. I can tell you that X happened and then Y happened and then Z happened if there are temporal (frequently causal) relationships between them, but I can't cross-reference that against another sequence of events or against calendar or clock time at all unless that's explicitly in the model in some way that's meaningful to me, or they're cross-linked to the rare things that do have time-references.

Abstraction barriers are a little weird, because if my model is adequate I want to see them and see through them at the same time. (Here is a subgraph with a complex interior and a mostly well-defined interface, but it's rare that it's *actually* a sufficiently well-defined interface and clean implementation that I'm actually happy to ignore the innards while working with it.) Generalizing ... I can't build a satisfactory big picture without enough details to hold it up, and that's often a lot of details (that other people don't want to get into while talking about the larger scale), and I bounce back and forth between vast and trivial and run off on tangents all the time. Most of that complicates communicating with other people who don't happen to be following the same threads or want to talk about a particular level of detail or piece of something.

I'm not good at modeling human relationships this way, even though the obvious mapping is, well, obvious. I think that's a combination of my having decided I don't get it and blocking myself ("people are difficult!") and that I need to support nodes and edges with more annotations in lieu of being able to build reasonable sub-graphs (and/or learn to build better internal representations of individuals, knowing that they are terribly flawed).

(Yes, I do love hypertext, and I'm quite happy to have internal nodes that link to external documentation for details I haven't internalized -- that's one way to get a graph started. Gods forbid I should ever get involved in editing Wikipedia or the like, as I'd never escape. The one little wiki that I do work on eats my brain often enough.)

Continued:

* Mathematical Patterns:
* the dimensions: Do you have a well-defined frame of reference, with which you're measuring everything? Often basic behavior of physical systems is intuitive from the way the units work out (e.g., every variable has combinations of mass, distance, time, and similar quantities-- can you multiple two variables together and get the units of a third variable (Buckingham Pi method)?
* energy: Is there a conserved quantity, which different processes trade in?
* the discrete and continuous: Are there both discrete and continuous variables involved anywhere? Then you might be able to apply the principles of signals and systems (6.003). Many problems can be looked at both spatially and in frequency-space.
* The first-order and second-order systems: first order systems have one feedback loop, and they increase and decrease exponentially-- what's the time constant? Second order systems, with two feedback loops, oscillate-- what's the period?
* the linear system: Do different factors combine independently, so that if you have none of the factors, you get none of the result, but if you have some of two factors, you'll get as much as the sum of their results?

the balancing act, and the catastrophe: Are things moving toward greater-and-greater states of tension? Then everything will collapse and reorder itself eventually, unless that changes.
* the fractal: Is it complex on all scales? Then look for self-similarities between the levels, and follow garden-paths from the largest level to successively smaller scales.
* the chaotic system: Does it diverge quickly from the models used to describe it? Then it has many interacting feedback loops, and your best chance of understanding it is to look at the consequences of more than one model of the whole.
* People Patterns: I use another set for understanding people, including MBTI, sensory dominance, neurotransmitter effects, and another for groups, with social circles, tension, and games. As an INT, I have to remind myself that when it comes to people, the details matter enormously.
* Subconscious Help: Allow that some other part of your mind is way ahead of you on this sensemaking, and that it can help you anytime you're stuck or not progressing the way you want to. The subconsciousness doesn't work using the paradigms that both help and hinder conscious thought.
* Programming Languages: I have a further set of patterns that I use when I'm learning programming languages and systems: I look for object-orientedness, functionalness, proceduralness, events, libraries, and exceptions.

I use intuitively patterns a lot, but some of my favorite ways to learn have nothing to do with them. I look for individuals and groups who know the material, and I ask them to get me started, or immerse myself in the group centered around the body of knowledge. I "projectify" the sense: I find some interesting thing to *do* with the body of knowledge from the beginning, and that drives me to explore the pragmatically important parts and makes the information relevant.

I think knowledge "sticking" is another craft along the way: how do you make your mind receptive to facts, to the exploring process, to the learn-by-failing process?

Another place you might want to look for sensemaking ideas is the Gumptionolgy 101 section of Zen and the Art of Motorcycle maintenance. My first waitlisted project is a "Model Teacher", an automatically customized tutorial (of programming languages, at first), using patterns like this, so keep me informed.

The danger I see in the approach you describe is that it involves forming one's analysis at the very beginning of the learning process, and then accepting facts in that context subsequently. Everything I know about the way people learn suggests that the temptation to shave your pegs to fit the holes at that point is both pervasive and invisible.

I have two ways of making sense: seeing connection, and seeing disconnects. These are interaction patterns I make in my head, whether I'm thinking about history, math, or finding my way to someplace I haven't been before (I don't do very well with that last one).

For example, when I think about history, I look for patterns: how it repeats itself, and how it DOESN'T. And when it repeats itself, the ways in which it differs from the previous repetition and the one that follows it.

Memorizing straight data, I'm not so good at. I've never remembered names, dates, and places very well at all. But seeing them as patterns of interaction, I can do.

When I have an instructor that just throws a bunch of information at me, without at least pointing out some of the connections between one thing and the next, I'm sunk.

Hm.

So, this isn't something I've thought about too explicitly, but it's something I do a lot. So bear with me as I attempt to wrap some words around what I actually do with a minimum of confabulation and retrofitting.

Some of this is patterns, some is techniques, some is just attitude. Also, I don't actually usually do all of this, only when I'm tackling something I know will be too complicated to just "stare at until it makes sense" or something I'm intimindated enough by that I need to approach mechanically.

For pretty much anything involving an argument, three-pass processing: one fast pass to understand the conclusion (often this involves simply jumping to a summary section), a second semi-fast pass to follow the argument (sometimes writing it down as I go along), a third slow pass to actually look at details. Mostly, this is so that when I do the detailed pass I can answer the question "where is the author going with this?" which lets me prioritize concerns -- if something is poorly-supported but peripheral, I can put a question mark next to it and move on; if it's poorly-supported and central, I have to bang on it.

Also, if I'm really serious (which I almost never am) I'll do a fourth semi-fast pass where all I'm looking for is similarities in structure. I don't have a good definition of that, and I think that's OK... what I'm actually doing, I think, is letting the unconscious pattern-detecting parts of my brain have their shot at the data. Almost everything I get this way turns out to be trivial, since a lot of formal similarities don't really get at any important underlying facts about the domain... "gee, here's yet another example of someone thinking something, then other people arguing with them, and them turning out to be right!" and that sort of thing, but sometimes it's important... "hey, wait a second, what we're doing over here is basically the same kind of thing that we're doing over there, why should we build different gadgets for doing each?"

Also, my use of "something is poorly supported" as opposed to "I don't understand something" above is key too... it helps me to approach material as a reviewer rather than a student, even if I'm a student. Not so much because I find errors that way (although sometimes I do) but because the "looking for errors" mindset forces me to actually process the information, chunk it, tie it into other things I already know, rather than just skim it agreeably. It also helps me avoid "gee I must be stupid" self-talk when I don't make sense of something. I often forget to do this, though.

For any dynamic process, long-and-thin use cases: what does walking through this process actually look like, step by step, for a typical input? Mostly this is good for spotting places where "a miracle occurs", but also I've found that adopting the "point of view" of the input, if that makes any sense, helps me remember the process more cleanly. I suppose that's because it brings in episodic and kinesthetic memory, and I'm primarily a kinesthetic thinker. I often regret not having known this when we were studying the Krebs cycle back in biochem.

Again, for arguments: Look for the unstated assumptions, make them explicit. Usually they're trivial, sometimes they're important.

For understanding collections of data, many of the grouping tasks you cite above (grouping individuals into factions, events into time periods, etc.). I'll add to that that I try not to look for boundary definitions and instead concentrate on prototypical centerpoints... it keeps me from getting too bogged down in hairsplitting on whether something is an X or a Y. The consequence is sometimes a LOT of categories, which requires multi-pass categorization... in effect I chunk the data elements, then attempt to categorize my chunks. I've never actually done more than two passes, though I've often thought that it would be useful to.

Another pattern which shows up in a lot of places for me: optimization and game theory.

Some questions it raises and gives tools to deal with:

What are the controllable features?
What are the constraints?
What exactly is to be maximized or minimized?
Are there multiple independent actors?
Do they share the same objectives?
Do they share the same understanding of the situation?
Is there hidden information?
Does the situation reduce to a version of a standard combinatorial game?
Is there a saddle point (natural attractor)?
What are the dominated options (options for which another choice will always yield a better outcome for the controller)?
What solutions (sets of choices) are Pareto-optimal (there is no solution better for at least one actor and at least as good for all actors)?
What solutions are stable (any actor's change of option is detrimental to that actor if nothing else changes)?
Is there a difference between apparent and actual objectives (are people behaving in a way consistent with a rational view of the explicit factors)?

I don't think I set out to find specific patterns in information before I've started to try to learn the information.


I think first I dive into the information and kind of swim through it and see what sticks to what I already know.

In the process, if something seemed unintuitive, I focus on that part and analyze all aspects of it until I have identified the parts that run contrary to my intuition and have determined why they are that way if they really are that way or if I misunderstood something about them at first

I think only at this point do I start to say "do I see any patterns forming? What are those patterns and are they extensible?"

So then I take the patterns I think I see forming and see if further data also fits into those patterns. Any time I find a situation that does not feel intuitive or doesn't fit the patterns I'm trying to fit things into or receive feedback from someone else that the data doesn't really fit the pattern I'm trying to fit it to I do the analyzing and troubleshooting again to find out if that is truly the case, and if so.... why. Then I try and see if my patterns are still valid or need to be changed in the light of that information.

But I don't think I go into any sets of data with preconceived patterns like "if it has to do with subject A then I should try pattern X" (I might, but if I do, I don't do it conciously)

As an analogy to your factions pattern, I seem likely to look for a dominance hierachy. This includes analyzing who cares/doesn't care about their status.

"Animal and trainer" is another social pattern; you can iterate through the players and see whether they're getting punished or rewarded for particular actions. Neither party need be aware of the relationship for it to exist, but it's better not to assume that delayed gratification is involved.

I'm more aware of the patterns that I'm using, when it's subject matter that I find difficult; I think that's because applying an unsuitable pattern, and getting wacky results is more likely. Both of the ones I've offered can be applied by rote, without requiring that you have much initial insight.

Hmm. I'm sure I do this, but I do it much more at the intuitive level than the intellectual one. Never sat down and tried to create a taxonomy of the techniques I use, although it's something to pay attention to.

But a question occurs to me. (And please note that the following is random musings, not even remotely well-formed yet.) I seem to recall (correct me if I'm wrong) that abstraction is a skill that the INTJ is generally strong in. This whole technique seems based on that: deriving successful abstractions from previous learning experiences, and applying them as appropriate to the new learning experiences.

So the question is: is there likely to be a type correlation in being good at *using* this technique? Are different types likely to have different levels of success, both in deriving and applying these patterns?

Related to this: how applicable are other peoples' patterns? That is, I understand, at a pretty deep gut level, how to derive and apply my *own* patterns, but I am far less confident that a pattern that is *explained* to me is likely to occur to me as something to apply to a situation. How communicable are the patterns, and how do you communicate them effectively? (And do the answers to these questions vary by type?)

The thing is, I do know the programming patterns world moderately well, and the thing that's interesting about Patterns is that, when you scratch only a little below the surface, the interesting sources tend to emphasize that they really *aren't* very important. The key thing is understanding how to decompose and break down a problem: the Patterns are mainly artifacts and examples of that process. Insofar as specific Patterns are considered important, it's mainly as a communications tool -- a shorthand taxonomy that programmers can use to talk *about* their programming with each other. You can use them to program with, but the main point there is to teach you formalisms that will then lead you to understand the more useful higher-level abstractions.

Hmm. All of which argues that the technique you use may well be *quite* useful, albeit perhaps more as a means of learning how to think than an end unto itself. Okay, I'm convinced.

But another thing that's clear from the Pattern literature is that it's hard to really learn Patterns without concrete examples of how to apply them to various problem domains -- if you pursue "How to Learn" further, you should ponder illustrative examples. I commend the book "Head-First Design Patterns", from O'Reilly, as an interesting example of how to make this stuff really accessible, strongly focused on teaching patterns to people who have no idea what any of this stuff is about. It comes across as a bit fluffy, but it's immensely readable and really does seem to get at the meat of the problem...