The Four Learning Paradigms
Banasiewicz distinguishes four overlapping approaches to learning: observation, theory, computation, and simulation. The framework describes different ways to connect experience, abstract explanations, and models built with data.
THE IDEA
The Four Learning Paradigms
Banasiewicz groups learning into observational, theoretical, computational, and simulational paradigms. He presents them as a historical progression shaped by technology. In practice, the approaches overlap: observations inform theories, and theories guide the models used in computation and simulation.
Observational Learning
Observation and experimentation provide evidence from experience. A learner examines what happens, looks for patterns, and develops generalisations. The author connects this approach with empiricism, the philosophical view that gives sensory experience a central role in knowledge.
Theoretical Learning
Theory uses abstract relationships to explain observations and derive consequences. Tycho Brahe’s detailed celestial observations in the 16th century supplied data that Johannes Kepler used to develop laws of planetary motion. Newton’s laws and Maxwell’s equations likewise show how a compact set of relationships can explain many phenomena.
Reasoning from a theory can suggest new predictions without first collecting a separate observation for every step. Those predictions still need to be compared with evidence.
Computational Learning
Building on ideas associated with Babbage and Turing, the growth of computing in the mid-20th century made it possible to study models too demanding to calculate by hand. Researchers combine measurements with theoretical relationships to estimate how a system behaves.
This is useful for systems that are difficult to observe directly or experiment on, including subatomic processes and distant galaxies. What the model reveals depends on its assumptions and the data used to constrain it.
Simulational Learning
The author uses this term for increasingly detailed representations built from large, varied, and continuously updated data. Such models let investigators explore scenarios and interactions. A model of conditions shortly after the Big Bang, for example, lets researchers examine consequences of theories under conditions they cannot recreate directly.
Virtual reality offers one way to present a simulated environment. Machine learning serves a different function: it updates a model from data. Both can support investigation, but a convincing representation still needs to be evaluated against the physical system or task it represents.
When to use it
Using computational simulations to predict weather patterns and climate change.
Applying theoretical models from physics to engineering design (e.g., aerodynamics).
Analysing business data to identify patterns in customer behaviour.
Using virtual reality for immersive training in medical or vocational education.
FROM THEORY TO EVERYDAY LIFE
In practice
Tycho Brahe and Kepler
Tycho Brahe meticulously recorded celestial observations, which his assistant Johannes Kepler used to derive the laws of planetary motion—a clear transition from observational to theoretical learning.
Simulating the Big Bang
Researchers combine measurements and physical theories in computational models of conditions shortly after the Big Bang. Running the models helps them explore the consequences of those theories under conditions that cannot be observed directly in a laboratory.
A MOMENT TO REFLECT
Put it to work.
Which learning paradigm do you rely on most in your daily problem-solving? Can you identify a situation where a different paradigm would have been more effective?
When a model learns from data, what can the human investigator learn from its results, and what still needs to be checked?
Can you think of a real-world problem that would benefit from combining multiple learning paradigms?
Often confused with
CONTINUE THE CONVERSATION
Source texts
- Evidence-Based Decision-Making How to Leverage Available Data and Avoid Cognitive Biases (Andrew D. Banasiewicz)