Module 1 Lectures Flashcards
What are descriptive statements?
They give an account of how the world actually is, without evaluating it
ex: Russia is at war with Ukraine
What are normative statements?
Express how the world ought to be and/or to express an evaluation if it is good or bad
ex: Russia should not have started the war with Ukraine
What are scientists reasoning towards in general?
Scientists aim to build models that capture worldly phenomena
What is a model?
A model is a representation of a worldly target system
ex: a street map of eindhoven is a model of the road infrastructure
Do models accurately display the world?
They try to accurately display certain aspects of the world but ignore other parts of the world.
Every model has a specific goal
What is an argument?
A connected series of statements (premises) intended to give reason for another statement (conclusion)
Which 2 types of arguments exist?
Deductive and inductive arguments
What are deductive arguments and their characteristics?
Arguments in which the premises are intended to guarantee the conclusion. If premises are true then the conclusion must be true
All A’s are B
Vaios is A
Vaios is B
The strongest arguments
Scientifically often unattainable, except math-based science
What does it mean if an deductive argument is valid?
The logical form of the argument is such that if the premises are true, the conclusion must be true
Does not specify if premises are true or false, only on the structure of the argument
ex. statistics
What does it mean if an deductive argument is sound?
Valid argument that have true premises and therefore true conclusion
What are inductive arguments?
Argument in which the truth of the premises doesn’t guarantee the truth of the conclusion. The premises make the conclusion probable (occur with some probability)
Weaker than deductive but most common in science
Range in probability but always less than 1
Often draw conclusions on larger group with small sample
What are the three reasoning strategies in science?
Hypothesis-driven approach (HD): test a theory
Data-driven approach (DD): observe reality
Application-oriented approach (AO): build towards an application
What are the 5 steps in the hypothesis-driven approach?
- Look at the real-world
- Create a model based on the hypothesis from the real world
- Create predictions generated from the model
- Check if the predictions from the model agree with observations (data) made in the real-world
- Make a conclusion based on the agreement of data and prediction to see if the model fits the real world
Which step is deduction in a hypothesis-driven approach?
The step of using the model to get predictions is deduction
If model (M) then prediction (P)
What are the different parts of the model in the hypothesis-driven approach?
Hypothesis (H)
Auxiliary assumptions (AA): assumptions connected to the hypothesis which are known DO NOT fit the world (ex. no friction)
Initial conditions (IC): the initial state of the system (ex. starting variables)
What are the characteristics the predictions (P) need to have?
Sufficiently suprising: it should not be too easy for P to be true if H is false
Sufficiently precise: P should be precise enough to test H
A single prediction: only a single prediction should follow the hypothesis. More predictions make it easier to pass tests.
What is the inductive step in the hypothesis driven approach?
Step 5 where it is checked if the model fits the real world according to if the real-world data and predictions agree
How does positive and negative evidence look like in the inductive step of the hypothesis driven approach?
Positive evidence:
If M, then P*
P*, so M (with prob)
Future evidence can still disprove it
Negative evidence:
If (H & AA & IC), then P*
Not P*, so (not H or not AA or not IC)
Stronger than positive evidence
*initial conditions only for dynamic model
What are the steps of analyzing the deductive step in the hypothesis-driven approach?
Step 1: identify H, AA, and IC associated with model
Step 2: identify P
Step 3: Evaluate if plausible that if (H & AA & IC), P follows
Step 4: Evaluate surprisingness, precision, and singularity of P
What are the steps of analyzing the inductive step in the hypothesis-driven approach?
Decision tree:
Does data and prediction agree?
No: there is evidene that the model fails to fit the real world objects
Yes:
Are there plausible alternative models?
No: there is evidence that the model fits the real world objects
yes: The data are inconclusive regarding the fit between model and real world
What is the objective of the hypothesis driven scheme?
Is this objective always achieved?
Objective:
standard of how to do hypothesis driven research
Much research does not achieve this due to unfeasibility, sloppiness, faulty argumentation, biases, and more
What are the steps in the data-driven approach?
- look at real world
- Get data from the real world
- create a model based on the data from the real world which agrees with eachother
- Check if the model fits the real world according to the agreement of the model and the data.
What is the deductive step in the data driven approach?
There is none
What is the inductive step in the data driven approach?
Step 4 which checks if the model fits the real world by seeing the agreement of the data and the model.
D, so M (with some prob)
Where does the probability of M depend on in the inductive step of the data driven approach?
formula: D, so M (with some prob)
Reasons:
The existence of other plausible models/hypotheses consistent with D
Quality, amount, conditions, and deviations of observations
Plausibility of auxiliary assumptions (AA)
What are the steps to evaluate the data driven approach?
- identify D
- identify M
- evaluate the three points determining the probability of M (plausible other models, observations, AA)
What are the steps and stages in the application-oriented approach?
Problem definition:
0. define the design specification for the application
Model stage
1. Create a model
2. Make model according to design specs
3. Get model data and look if the data matches the design specs
Artifact stage:
4. Use analogical reasoning to go from the model to the actual artifact
5. Make the artifact according to the design specs
6. Get artifact data and look if the data matches the design specs
At step 3 and 6, the approach can go back to adjust the design specs.
How do you evaluate the application oriented approach (globally)?
- Evaluate design specs
- Evaluate the model stage and artifact stage according to hypothesis driven approach
- Evaluate the analogy between the model and artifact
What are the evaluation metrics of the design specs?
SMART
Specific
Measurable
Acceptable
Realistic
Time-related
What are the steps to evaluate the model stage of the application-oriented approach?
Evaluate the model -> prediction -> data as in hypothesis driven approach
What are the steps to evaluate the analogical argument between the model and artifact stage of the application-oriented approach?
Is it inductive or deductive?
The model and artifact ought to be sufficiently similar in the relevant respects
Inductive formula:
M (model) has properties f, g, h. It meets z (design specs)
A (artifact) will have properties f, g, h
Then, A will meet z (with prob)
On what characteristics does the confidence in the interference depend in the analogical in evaluation for the application oriented approach?
Inductive formula:
M (model) has properties f, g, h. It meets z (design specs)
A (artifact) will have properties f, g, h
Then, A will meet z (with prob)
characteristics:
The relevance of the similarities of f, g, h
The number of relevant similarities
The degree of similarity
The relevance, number, and degree of dissimilarities
The existence of other, independent models yielding the same conclusion
What are the steps to evaluate the artifact stage of the application-oriented approach?
- Evaluate the artifact-> prediction -> data as in hypothesis driven approach
- Evaluate the fit of the artifact to the design specs
What is the qualitative and quantitative version of the data driven approach?
Qualitative: Historical texts
Quantitative: statistics
