lecture 23 (reasoning)

Question 1

what’s a theory?

Answer 1

statements which organize how we believe the world to function and which allow us to make predictions about future functioning

Question 2

formal vs verbal theory?

Answer 2

• formal theories: use mathematical equations and rigorous calculations, being more precise
• verbal theories: use descriptive language, risking lower clarity and ambiguity

Question 3

why do we need theory?

Answer 3

• observations are theory laden: assumptions about world/humans drive your observations; a good theory makes this assumptions explicit and verifiable
• the way I do research (make observations) influences the way I reason about my theory
• we need more integration in psychology (confetti explotion analogy) -> theory helps connect research findings

Question 4

reasoning from theory to experiment?

Answer 4

• the ideal is a clear theory from which an experimental and statistical hypothesis is distilled by means of a number of logically powerful steps (check notebook)
• in practice we see argumentation steps that are left implicit or are completely absent, and misleading forms of argumentation

Question 5

Circular reasoning?

Answer 5

• theory as an almost literal reformulation of experimental findings, without further explanation
• the arguments might be valid, but they are not useful
• 3 types:
  1. repeat the premise
  2. premise presupposes the truth of the conclusion
  3. premise is logically/semantically equal to conclusion

Question 6

equivocation?

Answer 6

• the same word (without further additional assumptions)
  describes several underlying processes -> more detail is needed for a good theory
• hot hand fallacy: Representativeness causes people to expect more good throws from people who have thrown a good series
• gambler’s fallacy: representativeness causes people to expect ‘black’ to be more likely after a long series of ‘red’

Question 7

the null ritual?

Answer 7

• common practice fallacy: everyone is doing it so it must be right (“always perform this procedure”)
• false dillema: “either my theory is right or the null will not be rejected”
• straw man fallacy: exagerating the point of the opponent so it can be easier to refute it (“no mean difference” or “zero correlation”) -> any deviation from zero difference will be significant with sufficient power

Question 8

errors in deductive logic?

Answer 8

• confimation bias
• confirming the consequent of a conditional: if my theory is true than this observation, this observation therefore my theory is true

Question 9

Duhem-Quine problem?

Answer 9

you predict something on the basis of your theory and a set of assumptions which is falsified -> but you don’t know exactly what is untrue, the theory or the assumptions

Question 10

discovery oriented vs theory testing research?

Answer 10

• “discovery oriented research weakly implies an hypothesis from theory”
• “theory tetsing research strongly implies an hypothesis from theory”

Question 11

model fit?

Answer 11

in what way does your theory need to outshine other theories to be the one we should prefer in our abductive argument

Question 12

descriptive adequacy?

Answer 12

• how well does the data fit the theory -> distance between the values predicted by the model and the values found is small
• most common way of testing data is NHST, however, this gives limited information as it always tests only two competing hypotheses and does not consider alternatives
• the available data can be explained by an infinite number of models (identification problem) -> in the case of equal adequacy, other factors must be taken into account

Question 13

precison and interpretability?

Answer 13

• how precisely is the data described/ are the relationships described in the model or in the theory
• formal models up precision

Question 14

what is a model?

Answer 14

• simplified representation of the world aiming to explain the observations
• models are formal (as opposed to verbal theories which are informal) representations of the world, usually including mathematical equations or computer code

Question 15

advantages of formally specifying theories?

Answer 15

1. Models allow the creation of strong tests for theories
2. formal theories are more specific and can be tested exactly
3. Models can lead to better theories
4. Modeling helps address real-world problems by increasing ecological validity

Question 16

model selection and its approaches?

Answer 16

• comparing alternative models and choosing one given the data
• Practical approach: the model that better predicts the unseen data on the validation set should be chosen
• Simulation approach: models could be run to see how well they can predict the outcomes and then tests can be designed to discriminate between them
• Theoretical approach: generalizability = goodness-of-fit + complexity

Question 17

problems with model selection?

Answer 17

1. Irrelevant specification problem: the precision of a model can be challenged if it is inappropriately translated from informal to formal and differences arrise
2. Bonini paradox: more complete and complex models simultaneously become harder to understand
3. data can be explained by an infinite number of models

Question 18

parsimony and generalizability?

Answer 18

• parsimony: only the constructs that are necessary for explaining phenomena should be used not more
• overfitting: the better the model fits the current data, the more likely it is that it does not fit the new data so well since not only the signal is fitting in the model but also the noise, which will be different in a different data set -> decreases generability
• trade off between precision and parsiomony

Question 19

prediction and postdiction?

Answer 19

• what is the value of being able to predict results in advance
• Postdictive explanations: after the fact explanations of why something occurred
• the timing of deriving an expected result from the theory (and alternative competing accounts) should not matter

Question 20

coherence and consistency?

Answer 20

is the theory logical and internally consistent

Question 21

originality?

Answer 21

is the theory new

Question 22

breadth?

Answer 22

does the theory apply to a wide range of phenomena

Question 23

usability?

Answer 23

does the theory have real applications

Question 24

falsifiability?

Answer 24

• theories can be compared on falsifiability
• degrees of falsifiability: the number of different ways in which a theory can be wrong -> this is perferred all else being equal
• theories with more precise predictions and that cover a broad range of situations are more falsifiable
• more falsifiable models say more about the world