Please enable JavaScript.

Coggle requires JavaScript to display documents.

To what extent do you agree with the claim “all models are wrong, but some…

- - - - Explanation: By reducing complex mathematical relationships (i.e.: astronomical calculations, particle physics) to a few variables allows us to focus more on the relationships without the interference of unpredictable variables. This makes models both accessible and functional.
        
        Example: Linear approximation in multivariable calculus - how a complex function in a 3D plane is approximated using partial derivatives without dealing with the entire function. It provides a good enough approximation but ignores the higher order complexities, extreneous solutions, etc.
        
        Example 2: Linear regression
        
        Example 3: the value of pi (or any irrational number). Using pi in calculations such as those concerning circles, will give as answer for the circumference of area but it would not be correct, as we fundamentally approximated the value of pi to 10, 20, or a 1000 digits. We cannot truly do any "correct" calculation with irrational numbers because we round off irrational numbers to a certain degree, hence making the answer very close to 100%, but not exactly.
  - - - Explanation: Models, in order to be interpretable, do have to define certain relationships and trends between variables. Models cannot account for the millions of parameters that are present, limiting their accessibility and functionality.
        
        Example: Neural Networks: neural networks take into account millions and billions of parameters concerning a specific model in order to make a prediction, yet no model can attain 100% accuracy, because even though it is a very complicated model with billions of variables (weights, biases, etc.), it still cannot account for the complexities of reality. Makes you wonder then how can a mathematical model be true even if billions of varaibles cannot define it.
        
        Euclidean Geometry.
- - - - This means that in the human sciences, often the models are built across trends from large populations and data sets. These models may be useful in providing an indication as to what is going to happen, but are not correct as they dont consider individual psychologicla, social, and cultural factors.
        
        Example: Taylors Theory of Motivation: a broadly generalized theory in work-place psychology which suggests that workers are only motivated by money and nothing else. While this might be a useful model to depend on, driving payment schemes, it is hardly correct as there are individual worker nuances which arent conisderd, like importance of job security, relationships, etc.
  - - - This counterclaim talks about how models in human sciences are based on studies and evidence that provide definitive results, often pairing with natural sciences or historical evidence, which overshadow the effects of personal nuances to a degree where they are inconsequential to the truth. Analogically, in the battle between evidence and biases, evidences can sometimes pile up, rendering the vast individual differences moot.
        
        Example: Theory of Supply and Demand; this simplistic model of the supply and demand of goods in a market is always true no matter individual differences and biases. When the price increases, it can said that all consumers will be less willing to buy that product, considering the theory of rational choice (this can be contradictory). Hence, this example shows how the validity and evidence-based proof of the supply-demand model is so storng that it overshadows personal differences in spending patterns, culture, race, education, etc.
      - Can argue that these personal biases can reduce the truth from 100% even if they are inconsequential.