(Last updated: 2025/03/30 4:05)
There are several methods we can use to build knowledge systems. For example, we can classify our knowledge based on theoretical hierarchy (meta/grand/middle-range/applied/empirical generalization). Also, we can build the system based on the disciplinary division. According to different purposes, we have different methods.

However, it is too difficult for personal users to build knowledge systems, if we are exposed to multiple disciplines, diverse work experience and life experience. So, this article aims to find an approach that is suitable for me to build a knowledge system. The framework will be shown first, and then I will introduce how I create it.

How to Build Knowledge Systems

Theoretical foundations

DIKW Model

The DIKW Model (Data, Information, Knowledge, Wisdom) is a hierarchical framework that illustrates how raw data evolves into actionable wisdom through contextualization, analysis, and ethical application. Here’s a structured breakdown:

  1. Data: Raw, unprocessed facts or symbols (e.g., numbers, text, sensor readings). The foundational layer; lacks meaning without context.
  2. Information: Data organized or interpreted with context. Answers “who, what, when, where” by adding relevance.
  3. Knowledge: Information synthesized to identify patterns, relationships, or principles. Enables prediction, decision-making, or problem-solving (e.g., forecasting tomorrow’s temperature).
  4. Wisdom: Judicious application of knowledge with ethical and experiential insight. Guides long-term, value-driven actions.

Applications:

  • Business: Transforming sales data (data) into strategic inventory plans (wisdom).
  • AI/ML: Data → information (processed datasets) → knowledge (predictive models) → wisdom (ethical AI deployment).
  • Knowledge Management: Emphasizes converting raw data into actionable organizational insights.

Key Characteristics:
Hierarchy: Each layer builds on the prior, adding value through processing and insight.
Criticisms:

  • Boundaries between layers can be ambiguous (e.g., information vs. knowledge).
  • Wisdom is abstract and context-dependent, challenging to operationalize.
  • Oversimplifies nonlinear, iterative learning processes.

OECD knowledge classification framework

The OECD knowledge classification framework, introduced in the 1996 report “The Knowledge-Based Economy”, categorizes knowledge into four distinct types: know-what, know-why, know-how, and know-who. This taxonomy emphasizes how different forms of knowledge contribute to economic and organizational development. Below is a detailed breakdown of each category, including examples, characteristics, and their roles in knowledge management:

  1. Know-What: Factual knowledge about observable phenomena, events, or data. Forms the foundational layer of knowledge, critical for decision-making based on verified facts.
  2. Know-Why: Knowledge of principles, laws, and scientific theories explaining natural or social phenomena. Drives innovation and problem-solving by explaining underlying mechanisms.
  3. Know-How: Practical skills or expertise to perform tasks effectively. Key to operational efficiency and competitive advantage, as it is harder to replicate than explicit knowledge.
  4. Know-Who: Knowledge of social networks, including who possesses specific expertise or resources. Enhances collaboration and accelerates problem-solving by connecting knowledge sources.

PS: A structured classification of knowledge type in educational psychology

  1. Declarative Knowledge: Knowledge of facts, concepts, or information that can be explicitly stated or recalled (“knowing what”).
  2. Procedural Knowledge: Knowledge of how to perform tasks, procedures, or skills (“knowing how”).
  3. Conceptual Knowledge: Deep understanding of interconnected ideas, principles, or theories (“knowing why”).
  4. Metacognitive Knowledge: Awareness and regulation of one’s own thinking and learning processes (“knowing about knowing”).

Criticisms: Boundaries between categories (e.g., know-how vs. know-who) can blur in practice. Real-world problems often require blended knowledge types, which rigid classifications may overlook.

Framework Construction

Dual pathways of knowledge acquisition

Generally speaking, I tend to divide the ways of acquiring knowledge into two categories - structured and unstructured. Also, we can use other dualistic concepts, like institutionalized and practical. Institutionalized channels include formal education system (school/university), digital learning platform, etc. Practical channles cover professional experience, social observation, cross-scene experience, etc.

In my opinion, if the process of knowledge acquistion could cover ‘DIKW’, it is structured. Otherwisw, it is unstructured. There are 2 steps to achieve the goal of organizing the acquired knowledge.

For structured parts, we should clarify the knowledge framework;
For unstructured parts, we should transform data or “know-what” into knowledge “know-why”, with the details of “know-how”.

Propose the framework

Based on theoretical foundations and personal experience, I propose the framework of building knowledge systems. It is dual with five steps.

Unstructured

For the unstructured part, firstly, we should record experience objectively from a third perspective. Secondly, we should extract abstract concepts from records, such as pattern features and key variables. Thirdly, we should coupling the information moduels with academic theories. There are some tools could help the process. For example, the STAR-T method: Situation, Task, Action, Result, Thinking, Conceptualization.

Case:
Step 1. Record (STAR)
Situation: In Beijing, 2015, I joined in a startup project called ‘Chenxing Talent Plan’. This program aims to provide loans for college students and recover loans based on the percentage of students’ salaries after graduation.

Task: My job is to promote this project to target customers.

Action: There are two options for brand promotion - online and offline. I tried three methods to make offline campus promotions. The first is to communicate with the official to get supports, like establishing a enterprise-school organization, establishing a foundation, sponsorship for campus activities. Also, I tried online methods. I communicated with students’ union and organize an online speech through WeChat community.

Result: Obviously, the offline promotion failed in my school, because the enterprise did not provide sufficient resources. And the online speech succeeded. By the way, establishing a foundation and sponsorship for campus activities were successful in the other university, but I was just involved in and not primarily responsible for.

Step 2. Extraction (Thinking)
Why it failed? The enterprise should consider costs and benefits (short term and long term). For example, establishing a enterprise-school organization was a long term favourable plan, but it needed more funds in the short term. ‘Capital is short-sighted’, especially for the startup project.
Why it succeeded? In Beijing Jiaotong University, establishing a foundation and sponsorship for campus activities were successful, because the management had a relationship with the school. And the online speech was successful because of the low cost.

With limited resources in a startup project, we should choose low cost promotion channels at first. Appropriate content and KOL can be effective to attract the first batch of users. However, the content needs to be linked to the project.

Step 3. Conceptualization (MFUs)
Concepts: resources, costs, benefits, long term, shor term, channles, content, content carrier, customers/users, stakeholders, decision making
Relationships:

Step 4. Reconfiguration

Step 5. Creative Application

(See details in the next section)

Structured

For the structured part, in most cases, there are knowledge systems or knowledge maps while you are studying. For example, when you read articles, authors will list their hypotheses, methods, conclusions, etc. While you take courses or read books, teachers and authors will also introduce the outline, theories, practical methods, etc.

The problem is how you integrate this knowledge into your own system. Aiming to solve this problem, we should ‘split’ the information into smaller units. For instance, when we conduct literature review research, we will use framewroks to extract key information, such as TACI (Targets, Antecedents, Consequences, Interventions), SPIDER (Sample, Phenomenon of Interest, Design, Evaluation, Research type), etc.

In this part, I wanna introduce a word from the business research, MVP (Minimum Viable Product). MVP is the simplest version of a product that can be released to the market with just enough core features to satisfy early customers and validate a business idea. By analogy, when we process the structured knowledge, we should find the MFU, minimum feasible unit as the foundation of your knowledge system. When referencing MFU, we should understand it by common sense or MNK, minimum necessary knowledge, which should be easy to be measured or quantified.

Why should we use MFU? Let’s back to the case mentioned above. In terms of brand promotion, we divided it into online and offline. As for the online promotion, we chose online speech to attract customers. In this experience, online speech is the MFU. We can answer the question ‘How should we organize an online speech?’ through common sense. ‘Speech content, speaker, venue (WeChat) and audience (students).’

How about the courses, articles, books? We extract theories and ideas from them. For example, the ‘Easterlin Paradox’ states that at a point in time happiness varies directly with income, both among and within nations, but over time the long-term growth rates of happiness and income are not significantly related. In this case, MFUs are happiness and income.

As a matter of fact, in the academical field, we usually use CDI (Concept-Dimension-Index) to make research designs. But in different cases, MFUs could be concepts or indexes. The boundary is blur in practice. When we link different concepts (or MFUs) together, theories are formed. It could be linear or non-linear relationship, could be hierarchy, could be framework, could be workflow, etc. Now, I wanna propose steps to process structured knowledge.

Step 1. Outline
Outline the knowledge from authors’ perspectives.

Step 2. Distillation
Identify useful theories and ideas from the outline, including relationship between concepts, framework, models, etc.

Step 3. MFUs
Find MFUs of these theories and ideas, and make tags on them.

Step 4. Reconfiguration
Combine MFUs with distillation and extraction parts, to understand the relationship, hierarchy, framework, etc. Transform empirical knowledge and others’ knowledge into your own’s.

Step 5. Creative Application
Try to apply your knowledge in the real world!