Academic and Engineering Walkthrough
A concept unites objects by characteristics. A term is a linguistic sign
by which a concept is expressed. One term can be polysemantic, and one concept
can have several names.
«клиент»
├── пользователь продукта
├── организация-покупатель
├── HTTP client
└── приложение, вызывающее API
The phrase “the client must repeat the request” without context almost certainly
will create a discrepancy between the documentation, code, and support.
2. Content and Scope
In classical logic, a concept distinguishes:
- content — a set of features thinkable in a concept;
- extent — a class of objects to which the concept is applicable.
понятие «идемпотентная HTTP-операция»
содержание:
повторение одинакового запроса имеет тот же предполагаемый эффект,
что и один запрос
объём:
конкретные операции API, удовлетворяющие этому условию
The more required properties we add, the narrower the volume usually becomes:
сервис
⊃ stateful-сервис
⊃ stateful-сервис с PostgreSQL
⊃ stateful-сервис с PostgreSQL и синхронной репликацией
This is called the inverse relation of content and volume. But do not turn it into
mechanical arithmetic: attributes can be dependent, incompatible, or
add nothing to the already given definition.
3. Unit, general, collective and empty concepts
The Particular and the General
- singular: «
courses repository»;
- general: «static site»;
- collective: «team», considered as a whole;
- empty: a class that has no instances in the selected domain.
Empty classes are important. In modern logic from:
Все единороги проходят code review.
should not:
Некоторые единороги проходят code review.
A universal statement is vacuously true in a model without unicorns. Many
traditional square opposition schemes assumed existential import:
the subject class is non-empty. This assumption needs to be added explicitly.
Distributive and Collective Reading
Команда подняла сервис за день.
This does not mean that each participant raised the service in a day. Attributing the property of the whole to each participant is a future division fallacy; attributing the property of parts to the whole is a composition fallacy.
4. Genus, species and specific difference
The classical definition is built through the nearest genus and specific difference:
Квадрат — прямоугольник с равными сторонами.
род видовое отличие
Engineering example:
Circuit breaker — механизм устойчивости, который временно прекращает вызовы
зависимости после достижения порога ошибок и затем контролируемо проверяет
восстановление.
«Resilience mechanism» defines the family, the rest separates circuit breaker from retry, timeout and bulkhead.
Essential and accidental feature
For the current purpose, the attribute by which an object belongs to the required
class is essential. The button color is not essential for the concept of «idempotent operation», but
may be essential for the concept of «control with sufficient contrast».
Relevance depends on the model and the task. The fallacy begins when a local
property is declared the nature of the object:
All our successful services are written in Go; therefore, Go is an essential characteristic
of a successful service.
Here technology is probably mixed with the team size, competencies,
workload, and operation quality.
5. Relations between Concepts
flowchart LR
A["Совместимые"] --> A1["Тождественные
один объём"]
A --> A2["Подчинённые
S ⊂ P"]
A --> A3["Пересекающиеся
S ∩ P ≠ ∅"]
B["Несовместимые"] --> B1["Соподчинённые
виды одного рода"]
B --> B2["Противоположные
допускают промежуточное"]
B --> B3["Противоречащие
P и не-P"]
Opposition and contradiction are especially easy to confuse:
- «fast» and «slow» — opposites; there is a mean between them;
- «cached» and «not cached» at a fixed moment and in a fixed
sense — contradictory attributes.
From “the solution is not centralized” it does not follow “the solution is decentralized”, unless
it is proven that the scale is exhausted by two options.
6. What is a Good Definition
Definition answers the question of how to recognize membership in a class.
6.1. Nominal and Objective
- nominal explains the use of the sign: «SLO means service level
objective»;
- real defines the properties of the object: «SLO — target value of a measurable
service level on a specified window».
Decoding an abbreviation does not yet define the entity.
6.2. Explicit, Operational, and Recursive
Explicit:
Чётное целое — целое n, для которого существует k: n = 2k.
Operational:
Запрос считается успешным в этом отчёте, если сервер вернул 2xx не позднее
500 мс и асинхронная операция завершилась в течение 30 секунд.
Recursive:
Пустой список — список.
Если xs — список, то cons(x, xs) — список.
Ничто другое списком не является.
Operational definition is useful for measurement, but does not necessarily exhaust the concept. «Intelligence = result of a specific test» is convenient for research and too narrow for a philosophical thesis.
7. Definition Rules
7.1. Proportionality
The volumes of the definable and the definition must coincide.
Too broad:
База данных — программа, которая хранит информацию.
A text editor is suitable.
Too narrow:
База данных — сервер PostgreSQL, хранящий строки в таблицах.
Other relational, document, and embedded systems are excluded.
Check: find a false positive and a false negative example.
7.2. Absence of a Cycle
Надёжная система — система, которая работает надёжно.
Hidden circle:
Хорошая архитектура следует правильным архитектурным принципам.
Правильные принципы — те, которые создают хорошую архитектуру.
7.3. Clarity
Unknown cannot be defined through even less known:
Согласованность — отсутствие аномалий сериализуемости в префиксно-замкнутом
множестве историй.
This may be an exact formula for a specialist and a useless first
definition for a novice. A definition has an audience.
Where possible, state what an object is, not just what it is not:
Микросервис — не монолит.
The definition allows infinitely many irrelevant objects. A negative
boundary is appropriate when negation is an essential feature: «odd number».
7.5. Essential criterion
Senior-разработчик — сотрудник с должностью Senior.
This is a definition of a record in the HR system, not engineering maturity. For a career
conversation, observable responsibility zones and autonomy level are needed.
8. Definition Fallacies
| Fallacy |
Pattern |
Diagnosis |
| Too Broad |
D ⊃ X |
find an object that fits the definition but not the concept |
| Too Narrow |
D ⊂ X |
find an accepted instance excluded by the definition |
| Circular |
X explained through X |
replace terms with variables and check dependencies |
| Obscurity |
unknown through more unknown |
ask for an operational criterion |
| Metaphor instead of boundary |
“architecture is the skeleton of the system” |
ask what is included and excluded |
| Semantic Shift |
term changes meaning along the way |
rewrite each meaning with a separate word |
| Reification |
abstracts are attributed independent causality |
find active people, processes and mechanisms |
Materialization
The market decided the product is not needed.
«Market» is not a single agent. Useful decomposition: which user segments,
through which channels, at what prices did not perform what action?
9. Concept Division
Division reveals the concept volume through types. It has:
- divisible concept;
- basis of division;
- members of division.
Запросы по результату:
успешные
клиентские ошибки
серверные ошибки
Basis — result class. If “mobile requests” are added, the bases
will be mixed: a mobile request can be successful or erroneous.
Division Rules
- One basis per step.
- Members are mutually exclusive in the chosen sense.
- Division is exhaustive or explicitly includes “other”.
- No leap: move to the nearest useful subclasses.
- Levels are not mixed: genus does not stand next to a species of species.
Bad incident taxonomy:
backend
frontend
PostgreSQL
критические
Here the layer, technology, and severity are mixed. One incident will fall into
multiple buckets, and the report will lose its meaning.
A good model stores independent facets:
layer: backend
component: database
technology: postgresql
severity: critical
cause: capacity
10. Dichotomy and False Dilemma
Dichotomy divides the class by the P / не-P criterion and is therefore formally exhaustive:
запрос аутентифицирован / не аутентифицирован
But natural language often substitutes contradiction with opposites:
либо переписываем систему, либо ничего не делаем
«Do not rewrite entirely» includes modular replacement, strangler pattern,
dependency updates, legacy isolation, and many other options.
11. Classification and Purpose
Classification is not a photograph of the “true world structure”, but a tool.
Classification of books by author is good for bibliography; by topic — for learning;
by format — for the warehouse. You cannot declare one axis naturally correct without
a goal.
In engineering ask:
- What decision will be made according to this classification?
- Is the feature stable enough?
- Can an object have multiple labels?
- What will happen with the edge case?
- Who and to what extent will classify similarly?
12. Practical walkthrough: “technical debt”
Definition:
Technical debt is everything developers dislike in the code.
Problems:
- subjective criterion;
- too broad a scope;
- no connection to future value;
- the solution cannot be measured.
Working version:
Technical debt is an adopted or inherited technical decision that
accelerated a previous delivery or arose from constraints, but now creates
recurring cost of change, risk or operational losses.
Operationalization:
карточка долга =
контекст происхождения
+ наблюдаемая повторяющаяся стоимость
+ затронутые изменения
+ риск бездействия
+ варианты погашения
Now one can argue about priority, not about taste.
13. Checklist
- Is the term used in one sense?
- Is the subject area specified?
- Is there a membership criterion?
- Is the definition neither too broad nor too narrow?
- Is there no direct or hidden circularity?
- Is the metaphor supplemented with a testable boundary?
- Does the division use one basis?
- Do the categories mutually exclude or explicitly allow multiple labels?
- Are all important cases covered?
- Are empty classes and boundary cases handled?
- Does the classification help make a specific decision?
14. Tasks
- Give two definitions of a “ready task”: for a team and for a contract with
a client. Explain the differences.
- Find a false positive and a false negative example for the definition
“microservice — independently deployable service”.
- Fix the bug classification:
UI, backend, критический, регрессия,
Safari.
- Restore different meanings of the word “platform” in your project.
- Explain why “AI — a system that behaves intelligently”
is a circular definition without an independent definition of intelligent behavior.
Sources
- Georgy Chelpanov.
Textbook of Logic, chapters II–VI and XXIII.
- Aristotle. Categories.
- Irving Copi, Carl Cohen. Introduction to Logic — traditional work with
terms, definitions, and classification.
- George Lakoff. Women, Fire, and Dangerous Things — why real
categories often cannot be reduced to a set of necessary and sufficient features.
What’s next
Concepts become true or false not by themselves, but within judgments.
The next chapter discusses four categorical forms, term distribution,
the square of opposition, negation, and quantifiers.
Judgments and Square of Opposition