Distinguishing Models and the Physical World

March 12, 2025

Let’s think about why we need models, and what is a model anyway?

Recently, I wrote about focus and its mastery. Starting with today’s material, we will move towards more practical aspects of this mastery of focused work. But first, we need to understand those very first principles that need to be mastered.

The main operation in focus is the ability to distinguish between models and reality (what is actually happening).

In order to work effectively and be “effectively focused,” we must not just know this difference intuitively, but descend to a more concrete level and note in each context these things:

1. Our personal (or collective) subjective opinions about what is happening. These are representations;
2. Factual data about reality that we obtain through “measurements,” somehow “feeling and touching” reality;
3. And what we should actually do to achieve the result we need, based on already selected effective models suitable for working with our circumstances (such models are either the essence of someone else’s experience — Scrum, Agile, Lean, Kanban, etc., or models developed by us independently).

Immediately, we want to understand what these “effective” models are, how to select and develop them? This is already a whole discipline of systems engineering that we are moving towards — methodology. First, we need to understand focus and rationality before proceeding further.

Nevertheless, now we need to define a conditionally “good” working model as one that helps our representations not to lag far behind reality, and by working according to which, we push this reality, direct it towards the result we are interested in (towards improvements).

The thing is that representation is already, essentially a model. But it’s not that simple.

The circus has left… Has it left?

Human thinking itself is a complex phenomenon, and we are not aiming to fall into endless philosophical reasoning; today we are much more interested in the representations themselves and their nature. We need to understand these things somehow applicably in our context.

Representations exist in a kinda similar way in all neural networks — both “wet” and “dead”, in the sense that some “assumptions/representations” are based on others.

And here lies an important point — reality as it is based on actual metrics that reflect it. You could even say that reality is based on itself!

The problem with representations is that they “freeze in time” in our heads (and as a “cutoff date” in ML neural networks), remaining locked in subjective considerations, “forever in the past”, if we are not going to refresh them.

Even if at the moment of forming some representation it was based on real data, over time (and quickly) reality changes, it is not constant. But our representations themselves do not change if they are not reinforced with new data from objective reality.

That’s how we live, exist. One reality is objective, based on facts, and “the other” is subjective — based on assumptions recorded in unreliable human memory.

Such “subjective reality” is a reflection of objective reality.

And this is a very (VERY) common, built in cognitive bug. Need I say how many representations at the moment of their formation are based not on real data, but on other representations — our own or those of other agents?

Ok, stop, as I said that we are not going to delve deep into philosophy. This idea of a gap is the only thing we need to understand.

The main relationship that we need to realize, check and accept is that the greater the gap between our representations and reality, the more inefficient work (and activity in general) we are engaged in.

For example — workaholism. This is a common representation in the post-Soviet space — “Take more with a shovel, throw it further, and while it’s flying – rest!”

It has nothing to do with efficiency. Working hard may make sense when starting some completely unfamiliar subject activity, but it makes no sense when the activity has already been mastered. And alternating hard work with periods of light work or rest is, on the contrary, strategically very promising in terms of improving overall productivity. So-called workaholism leads to little good.

Another common representation in organizations is the idea that “multitasking” is cool, fashionable, and working with several tasks at once is very effective, increases productivity — “We take 25 tasks simultaneously into the roadmap, semi-blindly set priorities and go.” Moreover, the “representer” here can be quite well-read about kanbans and other methods, but have almost no connection with reality, not set it up.

Any manager who has such a representation will defend it very painfully, and when asked “by what metrics is this increased productivity evaluated? how to understand that it has really increased?” will find it difficult to answer anything coherent.

Even if we agree that Kanban is excellent, without “feedback from reality,” how far will we get with it? We are considering precisely the option when there is an idea in the head, but it is only a representation, in the sense that we do not know how to use the idea, we cannot due to the gap with reality.

Tead abouot Tokaimura nuclear accident. The managers who made this fatal mistake very zealously adhered to their ideas about optimization. Managers who were very “effective,” but having absolutely no understanding of the subject area they were trying to optimize.

Model or not a model?

Representations can be either smeared across the unconscious, personal — “well, I feel it this way,” or even collectively unconscious — “after we made decision X, something seems to be not going as it should…”. And what exactly is going where — we cannot identify; we continue to simply suffer and ask questions that are constantly left with no answers, working harder, redoing things, just to push forward somehow.

Representations can be formally expressed, for example, in some company documentation, and still be detached from reality — not grounded in any way with the physical world and with what the company produces in this reality and where it is even moving. As a result, we sit with “papers,” and there seems to be no use from them at all.

Of course, having some knowledge base is already better than being in complete chaos of unconscious representations, but how much better?

Not much, until these representations are extracted into focused attention, our own and that of our colleagues.

Representations that are in the “scope of focused attention,” and not on the outskirts of consciousness or in dusty markdown that one and a half people look at once every one and a half months, can already be truly formalized, grounded in reality – checked how far these representations are detached from it, and what can be done with them. Perhaps they can be transformed, tightened up, and a suitable method chosen, or perhaps they need to be simply declared counter-productive and illiminated.

Can we say that representation is a type of model, and that a model, in principle, is a representation?

Why then do we need these two words if they are the same thing?

First, representations are a broader category. And representations can be formed in different ways — “absorbed” into the brain from context, adopted from other people, from communities, “mentality,” formed based on experience, and so on.

And it is also fair that we are not interested in all kinds of representations. I remind you that we are talking about rational and focused work here.

For example, if our technical director has an idea about how to boil sausages for breakfast and what music is best to do it to — such a representation does not interest us and does not particularly concern us.

If our technical director has the idea that the company should always work in “startup mode” — chaotically and in a constant rush in some direction (the intuitive understanding of which is also only in the director’s head), such a representation interests us very much, directly concerns us, because it affects the quality of our life and the fate of the company overall.

A model is something that helps us form the “right” representation of something from the real world, with minimal gap. Sorry for the tautology, but a model models this real object, albeit at the cost of simplifications.

A model of a car clearly does not have the same engine and all other parts inside as a real car, but it gives a very specific idea of this car, its brand, and so on.

You can model both things and processes, but we’ll talk about that later and a lot.

Managing with concrete data

In this smooth way, we have approached the very essence of Data-Driven-Development, which can be formulated as follows:

We are fully aware (composedly/attentively) that our representations (both personal and team representations) are subjective, and therefore can (and will) lie. Therefore, we begin to use metrics from the real world to ground these intuitive representations, to correct them towards what is actually happening.

All this seems fairly straightforward and understandable, not expensive to start applying (even if only by oneself), but as practice shows – few people actually bring this into practice.

Our engineering task consists of the systematic movement to correct such gap situations, if “engineer” interests us not as a bold title on LinkedIn, but as a life calling.