Lecture 4 Flashcards
Are methods important?
They are the most important part of science
Thought experiment on simple vehicles
Construction
Observations with light
Interpretations (Psychological)
Machines developed which a sensor is either inhibitory or excitatory when light is shon on it and connects to an engine. There are two sensors at the front and two engines at the back, each connected to a wheel.
We observe the behaviour when light is shon on two vehicles. One turns away, one turns toward the light. If we did not know their design, how would psychologists describe this behaviour?
These behaviors could be interpreted as fear and aggression!
Thought experiment on simple vehicles: The problem with interpretative explanations
What kind of explanation would help with this?
If we understand their circuity, we will not find a section that facilitates fear or aggression.
We may be convinced that it is but we will not find evidence to support this
Very complex behaviour could theoretically arise from simple circuits. However, it is hard to get evidence for these. Better to create simple hypotheses first
ie they approach or turn away when they see light vs they are afraid or aggressive.
Complex vs simple hypotheses
When trying to understand cognitive or behavioral mechanisms, it is useful to consider an explanation that is simple because very complex behaviour can arise from a small set of principles.
Ockham’s razor
Why? They are easier to falsify or test because in light of contradicting outcomes, more complex explanations can generate ad hoc hypotheses to explain away inconsistencies
The limits of reductionism
Asking what level - why would you do it
If you do true reductionism, you always stop at matter/antimatter.
Does it help to do this in psychology? Is mind reducible to atoms? Is that useful to do?
It is helpful to ask when is my explanation sufficient and when I should go deeper. You can explain its behavior on one level but if we added a higher level of explanation such as psychodynamic or cognitive, would this add something about the true nature that reductionism would miss?
A problem for recutionism is
emergent behavior. Complex systems can be described in simple linear functions but the arising, emergent behavior might be unpredictable.
What is the appropriate approach to understand behaviour?
Useful to choose dependent on your question
ie for schizophrenia a psychologist might be happy at the behavioral level. A pharmacologist might want gene level or chemical interactions f they wanted to develop a drug for it.
The reductionist approach might miss something
.Does everyone think this is a problem
Such as emergent behaviour
Some don’t care and argue that if you cannot de=scribe it, you should not try Wittegstine
What does the robot experiment teach us?
To be careful about what we are thinking about because in the end there is rarely ONE approach that describes everything.
We also need to be careful what we assume, as what we observe can also be created by simple circuits hence, it is best to strive for simple, falsifiable theories.
Reductionism can be thought of as a
tool
arguably the best one but it is very rare anything gets fully reduced to matter. The level it needs to be reduced to depends on your question
ie schizophremia for a psychologist and a pharmacologist
Rationalism
Deduction via logic and reason leads to knowledge about the mind
Observation is not needed and potentially misleading
Empericism
Use observations to confirm or disconfirm hypotheses
Inductivism
Make observations which you use to induce theories, these induce hypotheses leading to more observations which arrive at laws
Falsificationism
A statement or hypothesis which is capable of being refuted is deducted from a theory. Test are designed to refute the predictions not to confirm the theory.
Theories allow for many predictions
If one is wrong, the theory is wrong
So you can strictly test them
If prediction is wrong, you must go back and amend the theory (Or scrap it).
Observations
Ideally unbiased start us off in science but these are never unbiased. We usually have an idea of what we are looking for. These are subjective
Status of data/facts
Data do not equal facts
They are usually produced by measurements and quantifications and hence are intertwined within a theory.
You have to interpret them with said theory.
Facts do not exist objectively but rather emerge from a context, that of the overlying theory
Fact = Data + Theory
Status of theory
Reflect research biases, prejudices, values and assumptions, the history of the scientist and the community and are embedded in a social context.
The social nature of science
Doing science is a human behavior - scientists are part of what they observe. Objectivity is a standard only achievable to a degree.
All science is a social activity. Conventions, traditions, shared assumptions and things like peer review which ensures controversial petitions are mediated out
There might not be a pure science and neither a purely subjective science: what is scientifically valid is a matter of conventions. Sticking to these ensures the validity of the science (across time) but not its objectivity.
By using the same conventions we can keep comparisons with older material but this just reflects continued cultural situation. Sometimes retards science by things such as peer reviews rejecting new perspectives.
Constructivism in science, the Khun cycle of revolutions
preparagdim period
Contending schools
Random fact generating
No science (No agreed set of methods)
Normal science
One paradigm
No schools
Puzzle solving within this methodology
Anomaly
Anomaly appears
Important and not compatible with paradigm
Could be shelved or might lead to
Crisis insecurity loosening or paradigm conditions Contending theories Old successful scientists oppose new thoughts
Revolution
Younger scientists adhere to new paradigms
Some older scientists (often less successful) switch sides
Old ones die
Go back to Normal science with the young rebels now representing the establishment
Science advances funeral by funeral
Observer effects (Quantum physics)
If you preform the double split experiment you will see evidence of light acting as a wave. If you now observe from the side, it begins to act as a photon again.
Does this mean reality is not real?
This shows that the act of measurement changes what we are measuring. What is observed is the observed thing and the observer combined. The world is generated by you at the moment of observation.
Others agree because we all use similar methods to observe and measure. If you changed these, you might change the observation.
The experiment as a social sitiuation
There is an interaction between the observer and the observed. The underlying assumption is typically that the researcher does not influence the participant but this is not true.
Two underlying assumptions in psychological experiemtents
1) researchers only influence the participants behavior to the extent that they decide what hypothesis is tested and how to operationalize the variables etc.
2) The only factors influencing the behavior od participants are the objectively defined variables manipulated by the researcher.
Not true
Experimenter bias
Exposure of male rats and mice to male experiments but not female experimenters produces pain inhibition.
Probably cos testosterones in pheromones’.
Pain intolerance might make the mice/rats vulnerable to attack so they hide it.
Shows how the experimenters might influence the participants.
Hawthorne effect
Hypothesis: being observed leads people c=to change their behaviour
Method (observational) researchers manipulated many independent variables and measured the DV: rate of work
Results: workers increased output irrespective of what the changes to the conditions, whenever they were observed
Conc being observed can change behavior
Expectancy effect
Hyp: researchers behaviour will be impacted by experimenters biases
Research method (experiment with two groups)
2 groups given rats and told to train them to run a maze.
One group told rats super smart, one group told rats bred to be bad at mazes. In reality, no difference.
Results: “good at mazes” group outperformed bad group
Conc: Results different because the students expectations caused behaviors which changed the rats’ behavior
Demand characteristics
Explanation + 2 Examples
The person being studied is not only a passive responder, but might
engage in the experiment actively, e.g., trying to solve the problem
what the experiment is actually about.
This can lead participants to respond in a way to confirm the assumed hypothesis, in order to please the experimenter.
The sum total of cues of the experimental situation that convey the experimental hypothesis to participants are called the demand characteristics.
Example 1 (Orne, 1962): subjects were asked to add sheets of random numbers, then tear them up into at least 32 pieces. Over 5 h later, subjects were still doing it. *Shows they want to please experimenter*
‣ Example 2 (Milgram, 1963): video
Shows extreme compliance
REPRESENTATIVENESS
Psychological research is done predominantly on white North-American undergraduate students enrolled in psychology courses at colleges and
universities.
‣ The bias is Anglocentric, Eurocentric, Androcentric, and, used to be Masculinist.
‣ The data obtained do not represent humanity in general.
‣ It is doubtful, whether obtained effects can be found in other populations.
ARTIFICIALITY
DRAWBACKS OF LABORATORY RESEARCH
‣ Psychological research usually unfolds inside research laboratories located at research institutes and university departments
.
‣ Participants are subjected to often bizarre tasks, which they are asked to perform in the name of science.
‣ Often these tasks are the result of a reductionist approach, aimed at identifying mechanisms of cognition/behavior.
‣ It is unclear, however, to what extent the observed behaviour reflects the normal operation of the brain in natural situations and under natural conditions. Animal research (in psychology) faces the same problem.
‣ One solution are field studies in which the experimenter tries to observe natural behavior in the wild without being noticed by the observed
population.
CORRELATIONAL RESEARCH
DESCRIBE AND PREDICT HOW VARIABLES ARE RELATED
‣ Correlational studies explore how variables are naturally related, describing and predicting relationships between the variables.
‣ Correlational studies cannot detect causal relationships between the variables.
‣ Example: most colleges require SAT scores in their student applications because it has been found that SAT scores and academic performance positively correlate – the higher the SAT score, the better academic performance.
‣ BUT: Scoring high on the test does not cause better academic performance. High academic performance also does not cause higher test scores. These variables are correlated, but not related in a causal way (A causes B).
‣ Correlational studies allow making predictions, and these predictions can be tested in controlled experiments to search for causal relationships between identified variables and observational outcomes
‣ Positive correlation: both variables move in the same direction (e.g., the higher the level of education, the higher the salary)
‣ Negative correlation: variables have an inverse relationship, thus moving in different directions – as one variable increases, the other one decreases
(e.g., the less day missed in school, the higher the GPA)
‣ Zero correlation: the variables are not predictably related.
REASONS WHY CAUSATION CANNOT BE INFERRED FROM CORRELATIONS
‣ Directionality problem: the direction of the relationship between variables can appear ambiguous. Causations cannot be determined, therefore it remains unclear whether a positive or negative correlation
results from the increase in one or the other measured variable.
For example, if you sleep less, your stress levels increase (negative correlation). But it is possible that increased stress levels make you sleep less.
‣ Third variable problem: this is a basic problem of all correlational studies. The relationship between the two measured correlated variables might be dependent on a third, not measured, variable (confounding variable).
Example, texting while driving (A) is correlated with driving dangerously (B):
‣ Risk taking (C) causes some people to text while driving: C→A
AND
‣ Risk taking (C) causes some people to drive dangerously: C→B
Experimental studies
METHOD USED TO CONTROL AND EXPLAIN PHENOMENA
‣ Scientists are interested in explanation phenomena, in establishing cause and effect relationships.
‣ The experimental method is a way to detect such causalities.
‣ In an experiment, in order to test a hypothesis, experimenters manipulate one or several variables, and, while trying to keep everything else constant, they
observe the effect of this manipulation on one or several other variables.
‣ Definition: An experiment is a research method that tests causal hypotheses by manipulating and measuring variables.
‣ The manipulated variables are called independent variables (IV), the variables to measure the effect of the manipulations are called dependent variables (DV).
Experimental studies 5 stages
1) manipulate IV
2) Randomly assign groups to (a) control or (b) experimental groups
3) researcher measures DV
4) researcher assesses result; is there a difference between the groups?
5) Conclusion: data either does or does not support the hypothesis, this could lead to new explanations
Descriptive statistics
Allow us to organize, summarize and describe data
Histograms
Assess the:
Shape of a distribution
Where is the center of the data
What does the spread look like
Skews
Positive (right) tail
Negative (left) tail
Normal distribution
Symmetrical
Mode=mean=median
Positive skew (order of central tendency measures)
Mode < Median < Mean
Negative skew (order of central tendency measures)
Mode > Median > Mean
Variability
SD
Degree of spread in the distribution of the data
SD is a measure of the variability of the data. It is the average distance of all data points from the mean
Percentile scores
68% of values within 1SD
95% within 2
Inferential statistics
Gather info about a subset of a population
Infer something about a population based on the sample
Do results represent meaningful findings (likely real, true findings) about the population or might they be due to chance alone
Less variability implies
greater likeliness that the difference between the group means is meaningful, statistically significant. More variability means more overlap between the distributions and less likelihood that the difference between the group means is significant.