Petrova digitální zahrada 🚀

EDSM - 1. lecture - A1

5 min čtení

Science is done by questioning

The Scientific method

  • The Scientific method = the empirical method
  • components:
    1. observe a phenomenon
    2. find patterns in observations
    3. develop fitting descriptions and/or equations (= models or hypotheses)
    4. conduct experiments (empirical verification) to verify to what extent the models are able to predict the future observations
    5. if it predicts the future observations successfully, it becomes a law or scientific theory

Models

  • a model is a description/simplification of reality
  • they describe, they do not explain (e.g. gravitational law)
  • fundamental laws are laws of nature
    • are very thoroughly tested, but not fully verified (that is not possible - to test out every available option)
    • they should be questioned and tested regularly (auto-regulating system of science)

No overfitting

  • you want to fit the data, but not overfit them
    • overfitting means fitting too much on the past data - which explains them really well, but on the other side, the future observations are skewed (and we don’t want that)
    • the most important things in the model are “predictions”
      • and empirical verifications of those predictions
  • the model function has to be as simple as possible (Occam’s razor)
  • models are often valid within some boundaries and have deviations and exceptions
    • e.g. the Newton’s Gravitational Law is valid within the General Relativity Theory

Self-fulfilling models and why we need to be careful

  • we need to be careful of the models we create
    • models tend to validate themselves, because with creating models, we do not describe reality, but we create/shape it (and the creation of reality can then shape the behavior and outcomes in a way it confirms the model instead of questioning it)
    • people tend to see the created models to be axiomatic and irrefutable/undisputed
      • the models are accepted without questions, which can then create flawed assumptions
      • empirical observations (experiment results) are viewed through the “lens” of the existing model, they are not used to question the model itself
  • design choices create reality
    • if I assume that users will do a lot of errors, I will design a more restrictive UI → users are than more passive (and less skilled) → which “proves” that the restrictive UI was needed
  • big danger is when observing the data:
    • if I observe data that match my model, I risk treating that data as proof of the reality
    • BUT with a self-fulfilling model, we may have created/shaped the reality to produce this data (e.g. through our design choices)
      • so it creates a sort of feedback loop: the model influences, how we treat and interpret data, we take actions based on this data and then we get results which confirm the original idea of the model/paradigm
    • to mitigate the risk:
      • ask: is this data showing the inherent truth or something my model created?
      • look for counter examples (where different model produced different outcomes?)
      • create conditions where it is possible to disprove my model
      • try different models and observe data
  • we should use observations to question the model

Do you realise that you are in a paradigm?

  • we always operate in some paradigm:
    • how the world works? what is good, what is bad? which questions should we ask? etc.
    • we do not live in objective reality and the paradigm we currently live in shapes our thinking and our research direction
  • we had few scientific paradigms in the past:
    • Earth center of the universe
    • Sun is the center (Copernicus)
    • Mechanistic physics with absolute time/space (Newton)
    • Relativity, spacetime (Einstein)
    • with each paradigm shift, we didn’t just add new facts, we had to reorganize everything and start working with completely new facts (and discarding many old ones)
  • as a designer, I can live in a paradigm, that users are emotional vs. that users are rational decision makers
    • and each paradigm will lead to completely different results and we should be aware of that
  • changing a paradigm means reframing the way we see the world

“Scientific progress involves both working within paradigms (normal science) and occasionally stepping outside them (revolutionary science). While science has self-correcting mechanisms like peer review and replication that allow for error correction, these mechanisms are imperfect and often slow. Paradigm shifts do occur but typically face significant institutional resistance, as scientists are invested in existing frameworks. The capacity for eventual paradigm change—even if difficult—distinguishes science from more dogmatic systems, but we shouldn’t romanticize how easily or frequently this happens.”

  • Thomas Kuhn’s theory on scientific revolutions

Falsifiability

  • a scientific principle saying that “no theory cannot be definitely proven true, it must be capable of proven false”
  • every theory, model, principle (anything scientific) should be falsifiable, that means that a potential experiment/observation could be made to contradict it
    • the argument does not have to be valid at the time, but it has to exist
    • e.g. the statement “All swans are white” is falsifiable by an argument “Here is a black swan” - it does not matter if the black swan is really there or not - the argument exists logically
  • this principle comes from Karl Popper
    • “no number of experiments can ever prove a theory, but a reproducible experiment or observation can refute one”
    • Popper distinguishes between scientific and non-scientific theories with this (the demarcation criterion)

Graf

Příchozí odkazy