Artikelen Flashcards
Baron (202) ch 3 in Sternberg & Halpern (2020).pdf:
mensen negeren belangrijke psychologische bevindingen, zoals kinderen die moeten leren hoe je engels schrijft en dat computers veel betere decisions kunnen maken dan mensen
oke
wat voor studies hebben vooral last van de replication crisis
priming studies (moral disgust and cleaniness, people who were primed with older words walked slower) en studies of ego depletion
studies of ego depletion =
idea: doing mentally tiring tasks reduces mental resources and causes a lack of self-control, so that subjects are more impulsive.
In another example, initial studies found that preventing subjects from thinking about a complex decision task resulted in better decisions than if the subjects were allowed to think, an “unconscious thought advantage
wat zeggen Baron et al dat het lastige is aan psychologie
we study ourselves: As a result of being human, and existing in a social context of other humans, we already know a lot about why we do things, how we feel, and what we believe. Although it is possible to find errors in
our understanding of our own psychology, most of it is pretty accurate.
but… science is focused on teaching us things that we did not already know. but surprising results tend to be false.
+ small samples -> low power
wat weten we dmv NHST (formule)
p(D|Ho)
en wat willen we weten? (formule)
p(H1|D)
wat moeten we weten om er achter te komen wat p(H1|D) is
P(H1) en p(D|H1)
The hypothesis of interest is less likely to be true, given the data, if it was less likely to be true from the outset. Surprising results are less likely to be real even when the significance level (.05) is held constant.
oke
kleine samples zorgen vooor….
lagere p(H1|D) en dus lagere p(H|D)
(t) these fields put a premium on surprising results, and (2) experiments use small samples with high variability.
oke
p hacking
A second reason has to do with the behavior of researchers. Notice the criterion for statistical significance is p(DIHo ). If we look at all the studies
with that number (often called just p) equal to .05, and in which Ho is true, we can expect that rout of 20 to be “significant.” If you keep doing the same experiment over and over, eventually you are likely to get a significant result, just by chance. The same thing happens if you test a lot of hypotheses on the
same study. Or test them in different ways. Thus, the significance level is distorted because researchers do not take into account the other things they did that didn’t work
file drawer effect
One form of statistical distortion can happen when several studies, undertaken by the same researcher or different researchers, test the same hypothesis.
When one of them doesn’t work, the researcher sighs and puts the data in a “file drawer”, never to be published. Perhaps she tries again, a different way, or
some other researcher tries, with or without knowing about the initial failure. Eventually, even if the hypothesis is false, one of these studies will probably be significant. At that point, whoever does it, will exclaim “Eureka!” (or words to that effect) and send the result to a journal to be published (she hopes). The
result is that the studies published may be a small sample of those that have been done, and the published p-values may not truly reflect what they are
supposed to reflect. This is called the “file drawer effect.”
bible code
where researchers convinced themselves (and others) that the Hebrew Bible contained hidden messages if you read it from top to bottom instead of right to left (as Hebrew is written).
-> it is easy for researchers to convince themselves that the method that happened to yield a significant result was really the most obvious method to use, even if this was not immediately apparent
cognitive dissonance
when scientists started to defend a certain point, they cannot stop, even though they might not have defended it if they had never advocated it at all
wat is lastig aan de relatie tussen vaccineren en later dood gaan
- causation of correlation?
-The problem here is that their measure of the relevant aspects of health are not perfect. Even a standard full physical examination does not pick up all the relevant risk factors. When you include an imperfect measure in a regression, you fail to eliminate the effect of interest.
wat was er mis bij de kinderen met error rates
But what is really happening is that the
problems were too easy for the older children and they made very few
wat is het kutte aan de media
- only report surprising findings: meestal fout
- geven alleen de positieve resultaten, niet replicaties die geen effect vonden
- rapporteren correlatie als causatie
- news reports may emphasize agreement or disagreement, without explaining where either one comes from. Some reports (e.g., about climate change) emphasize a “consensus of scientists,” without telling us how chis consensus was achieved.
actively open-minded thinking (AOT)
thinkers must not only be open to challenges that come their way but must also seek such challenges actively, by thinking of alternative possible conclusions, by looking for reasons that favor the alternatives or impugn their favored conclusions, and by asking questions about ultimate criteria for evaluating conclusions.
wat voor bias voorkom je door middel van AOT
myside bias
eerste stap AOT
searching: possibilities, evidence and goals
3 general properties of AOT
- search must be sufficient
- search and inference must be fair
- confidence should be responsive to the thinking done so far (heel veel thinking, goede evidence en inference = veel confidence)