Saturday, February 29, 2020

Symmetry Breaking

It's common knowledge that Socionics is a "Ti system." That is to say, Model A and the socion itself are very beautiful, symmetric systems that are, apparently, all-encompassing. This is all well and good, but the theory is also severely imbalanced in this way. It needs to be Te too, not just Ti.

What is a symmetry? It's some kind of change or process that leaves a system looking the same as when you started. Essentially, it's "a change that's not a change". If that sounds like a paradox, it's because it is.

The classic example of symmetry is rotations of a shape: if you rotate a square by 90° it will look the same as when you started — as long as the different parts of the square have no other distinguishing features (such as the numbers in the picture below). It won't be deformed or broken, and will occupy the exact same location in space.

image from

But the right side of the square cannot literally be the exact same as the left side, otherwise it wouldn't have four sides, it would have at most three. If the parts of the square weren't separate and distinct we couldn't even recognize a process as different from doing nothing — except maybe by looking at intermediate states of the rotation, to see that yes, the square is moving from one place to another. In other words, the parts have distinct identities. But in reality identities are only distinguished by observable properties — by Leibniz's identity of indiscernibles, two different things must differ on some property.

The same is true of Model A: EIIs and SLEs are "the same" by virtue of being socionic types, which have strengths, weaknesses, values, etc. described in terms of the same functions and IM elements, but they clearly don't have the same thinking and behavior. If all the types were the same then socionics wouldn't be very interesting. In fact they wouldn't even be types, as everyone would be the same.

Of course socionics acknowledges these differences already — in type descriptions the differences are described, albeit often not very well, or using overly-specific examples drawn from the author's own limited and potentially flawed observations of the type. The real issue is that the differences are not explicit in the theory. There is a mathematical structure describing precisely how the relationships and functions fit together, but there is no mathematical description of, e.g. what makes Te different from Ti. This means that anyone and everyone can come up with their own definitions and say that they're right. Certain ones may fit together better and describe reality better, but good luck trying to convince anyone else to change their minds. So we have a community which is becoming ever-more fractured in its theoretical foundations, with any number of baseless hypotheses being accepted as fact, and passed off as legitimate theory to unwitting beginners.

But in fact there are beginnings of such a mathematical description, partially described right here on this blog. The elements are described by explicit geometric properties like extension (extroversion) and limitation (introversion). A distinct hierarchy emerges: the irrational elements are more fundamental and "wider" ontologically than the rational elements. Irrationality deals with direct apprehension, both physical (the senses) and mental (the imagination and memory). In other words irrationality is prelinguistic while rationality is linguistic. But rationality can be seen as the culmination of the system (and the intellect it describes), and perhaps gives it a greater degree of closure. All of the IM elements are good without a doubt in their own way, but they do play different roles and some are preferred over others in particular contexts.

As mentioned in another article, the contrary elements (Ne and Ni, Se and Si, etc.) are in fact the same information but opposite "vectors" or preferences within it. The dual elements are different, seemingly alien perspectives on the same reality, and our duals open us up to this perspective. The superego elements are arguably the "most opposite" from a higher perspective, but they share certain deceptive similarities and are "far apart" enough that contrary elements will conflict more readily. Conflicting elements can be reconciled in the long-term view but they certainly conflict in the here-and-now, as opposite states or choices (which produce said states).

That's roughly how it goes for IM elements. What about types? One benefit of 16-function models is that they identify types with IM elements, and therefore also functions with relationships. "Introverted socionics" further identifies types with relationships. Taken together, they will theoretically unite the basic elements of socionics into one fundamental reality, which manifests itself in different forms. It just so happens that IM elements are at the forefront of what we can describe in detail, because they are what we actually observe (i.e., information). So if you want to understand socionics, understand the IM elements.

Monday, February 10, 2020

Types are not boxes!

People sometimes object to socionics or typology in general, thinking that having a type limits you to being in a particular "box". First of all, typology is not prescriptive: it only claims to describe differences that are already there. Assuming that you're typed correctly, it's just a new kind of knowledge about yourself, not a new law that you have to follow. Surely if you had, say, a disease or a special talent you would want to know about it, right? Then you could seek out a cure for the disease, or actualize the talent and use it for good. Going to the doctor doesn't make you sick, it just tells you if/how you are sick — or in the words of Tupac, "I didn't create thug life, I diagnosed it."

But socionics does one better: Model A essentially says that we have all of the types within us, all of their basic functions and skills (as the eight IM elements) — only differing levels of access to them, with some being easier to use and others harder or more difficult to keep up. The leading function is "you", but it's merely the baseline state — you can "fake" being another type for a limited period of time, and in a sense we do this all the time when we use our other functions.
(from Wikimedia)

It's like the idea of a harmonic oscillator from physics. One example of a harmonic oscillator is a mass at the end of a spring: it has a normal state where it extends out a certain distance X. You can stretch it to extend beyond X, but if you let go it will contract, oscillating until it gets back to the equilibrium point X. Similarly you can compress it to a smaller distance than X, but it takes work. So the leading function is like the distance X, with other functions being closer or further from that equilibrium (and particularly the bold functions will be closer to it).

The functional, fractal (self-similar) nature of Model A — touched on but not elaborated by Jung — is so popular that other typologies are now adopting it: Beebe's 8-function model for MBTI, and tritype for Enneagram where people speak of "my 9 [fix]" as if it were an aspect of their personality, like the Model A functions.

Comparison Typing

Socionics has only 16 types, but clearly there are more than 16 kinds of people. Each sociotype therefore has to contain a wide variety of people, so as to cover approximately 6% of the world's population.

Socionics enthusiasts sometimes use "comparison typing", i.e. typing others based on comparison with other, already-typed examples. So if your brother is IEE you might determine whether others are IEE based on whether they are similar to him, whether they have a similar energy level, interests, level of skepticism, etc.

To some extent this is unavoidable and actually beneficial: comparing with known examples is a quick and easy way to type. BUT, it should only be used positively: if someone is similar to a known example then they are likely to be the same type. But if they are dissimilar then they may simply be a variant of the type that you haven't seen yet. If you are relatively inexperienced in socionics then the probability of this is higher — but no socionist can assume that they've seen all variants of the types.

Secondly, even when you do comparison-type, you should always attempt to describe clearly what similarities you're seeing, and how they're related to socionics. Not all similarities are related to socionics, and in fact, energy level and interests vary widely within types. If you don't give a theoretical explanation then you're essentially building your own idiosyncratic typology, which is probably going to be less coherent than Model A, not to mention impossible to discuss with other socionists.

Negative Typing

A more subtle issue is when there is not an explicit comparison, but when someone types based on what is absent from a person rather than what is present. This is again negative typing, and generally speaking not the right way to type. It can trip up beginners even when they have a solid understanding of the theory.

For example, "mobilizing Se wants to look fierce, [but he doesn't]". This is based on a specific attribute that the typer associates with mobilizing Se.

This is an argument based on what is not there, not on what is there. It's essentially appealing to comparison with known examples. You may not know any mobilizing-Se types who don't want to look fierce, but that doesn't mean that they don't exist.

If you use a general trait that really does get close to the essence of mobilizing Se (such as restlessness or desire for impact), then it would be valid to use negatively. But this example seems too specific for that. It would be better to look at what is present, and see what that says about the subject's relation to Se, where it would best fit in their Model A.


People often complain about "stereotypes". But stereotypes are stereotypes for a reason — they just shouldn't be used negatively, to exclude people from a type. For example, if someone greatly enjoys participating in the performing arts then it's evidence towards being EIE, or some type with high Fe. But clearly the majority of EIEs are not performing artists, and no one in their right mind would say "he's not an actor, so he's not EIE". And if you dig deeper, of course you might find the performer is not really motivated by the emotional interaction he's having on the audience, but something else. Maybe he got into the field by accident (like Harrison Ford). But this is less likely, because performance is closely tied with Fe.

Appendix: Converse Error and Bayesian Reasoning

Now, a bit of math!

Converse error is a logical fallacy: if A implies B, it doesn't necessarily mean that B implies A. This is essentially why negative typing doesn't work. But socionics is empirical (although not quantitative), so rarely do we have strict implications going from observation to type. As mentioned above, just as not every EIE is a performer, not every performer is EIE even though it's more likely for them to be EIE than SLI.

Technically the right framework to use is Bayesian inference, which tells us how to update probabilities using observation. We start by assuming a roughly equal probability for any typing to be correct. If the person to be typed is in a community that skews a certain way type-wise, then you might not assign equal probabilities. Or if you believe that the type distribution is heavily skewed (which I don't).

Anyways, the formula goes like this:

Here P(X) means the probability of X, H means the hypothesis (the typing in question, e.g. EIE), and E is the evidence (the subject in question being a performer).

Let's rewrite the formula using the example:

Now suppose

Then we get

In other words: if 2% of the general population are performers, and 4% of EIEs are performers, then we'll double our probability of the subject being EIE, to being 1/8 = 12.5%. And if three times as many EIEs are performers as everyone else, then we'll triple it, to get 18.75%. So the small percentage 4% becomes significant when considered relative to the other types. 12.5% is still somewhat low, but if we have many pieces of evidence like this we can form a good case.

Of course, it's typically very hard to assign exact probabilities to a claim — in life in general, much less in the notoriously ambiguous field of socionics diagnostics, where you may not even be sure what you think. Nevertheless, the general principle still applies.

So, no more criticizing stereotypes! Negative arguments and comparison typing are the real issue.