lecture 5 - ethics Flashcards
Obedience to Authority
1974 - Stanley Milgram
- hypothesis: individuals will obey requests by authority even if it is considered unethical
- cover story: learning by punishment (+ supposedly random selection between teacher and student)
was not an experimental design: no random assignment treatment and control
was a cross-sectional design
Stanford Prison Experiment
Philip Zimbardo
- proved that people could quickly become abusive: 1/3 guards exhibit sadistic tendencies
- experiment had to be stopped after 6 days (was intended to last 14)
- goal: investigate role of situation vs personality
*this experiment led to stronger regulation
(fraud with fake data + plagiarism)
fraud with fake data:
- Michael LaCour: fake data in research was found when it was replicated -> lost career
- Diederik Stapel: 58 (co-)authored papers had to be retracted
plagiarism (politician edition)
- Germany 2011: Karl Theodor zu Guttenberg
- Romania 2012: Victor Ponta
- Germany 2013: education minister Annette Schawan
- Spain 2018: Carmen Monton
- Norway, 2024: Sandra Borch (Higher Education Minister)
3 / 4 basic principles of research ethics
- do no harm
- informed consent
- protection of privacy/confidentiality
- (transparency and documentation)
research ethics
how should researchers behave
what procedures should be there to prevent the breaking of norms?
informed consent
5 + 4
requires
- topic and nature of questions (e.g. are they sensitive/intrusive)
- use of information
- purpose/goal (incl. disclosure financial sponsors/interests)
- participation requirements/expectations:
voluntary participation, freedom to stop, permission to record and quote, permission to use data - risks involved
further consideration:
- competence and comprehension (marginalized/vulnerable populations)
- incentives for participation?
- (unobtrusive) observation
- experimental manipulation (concealment vs deception -> debriefing necessary)
privacy and confidentiality
- public information = still rules
- confidential information = not always possible
- anonymous information = ideal, but rarely possible (e.g. surveys: phones can be traced)
challenge = transparency vs privacy
ethical behavior of researchers and sponsors
researchers:
- avoid bias
- no incorrect reporting
- no inappropriate use of information
sponsors:
- don’t impose restrictions on research
- no misuse of information
ethical principles and procedures
formal review of research programs
US: IRB (institutional review board)
Leiden: Ethics Committee
(Europe was later than US)
formal rules for documentation and archiving
- pre-registration: disclosing research design publicly (so that hypotheses can’t be modified e.g.)
- EU General Data Protection Regulation (GDPR)
- Data Access & Research Transparancy (DART)
- Data Management Plans (Leiden univ)
validity + reliability
internal = ability to draw causal inferences (confidence in relationship, control for confounding variables)
external = generalizability of research findings
reliability = consistence and replicability of findings
threats to internal validity
5
- history: anything that happens outside of the research project but has severe implications on the ability to do the research and to draw inferences
- maturation (subject effect): responses you get early on differ from later responses as units/actors have learning effects, they change internally (e.g. also population growth)
- testing/performance effect: measurement instruments create effects rather than measure them (question/research influences responses)
e.g. sensitization
e.g. order of questions in survey influences answers - instrumentation/instability: measurement instrument may not be reliable and/or change over time (e.g. less responses or changing of wording in repeated surveys)
e.g. effectiveness, bias - statistical regression
regression towards the mean (pretest can have accidental findings leading to using ‘‘wrong’’ cases in the actual study)
floor and ceiling effects
threats to external validity
- selection bias / generalizability = choosing cases that support a theory V ‘professional’ participants (people that voluntarily take part may not be representative)
- mortality/attrition (may lead to unrepresentative sample)
- statistical regression:
*ceiling & floor effects (when the outcome has reached ceiling or floor -> can’t measure change e.g.)
*regression-toward-the-mean: selection criterion can lead to incorrect selection due to accidental findings in the selection
statistical regression
= threat to external validity
ceiling & floor effects = when the outcome measure has hit the ceiling or floor
- e.g. can’t research effectiveness of climate change campaign when there has just been a natural disaster leading to interest in climate change hitting the ceiling
regression-towards-the-mean = when pre-test includes accidental findings resulting in an ‘‘incorrect’’ selection
- e.g. pre-test to select smart people, but a smart person can have a good day, whilst a dumb person can guess correctly -> unrepresentative sample
