Upload
nova
View
56
Download
0
Embed Size (px)
DESCRIPTION
Epistemology of natural sciences. Dr Tim Daw School of International Development University of East Anglia [email protected]. Overview. Epistemology – the nature of knowledge in natural sciences The ‘scientific method’ Popper – Falsification, Deduction Fisher - Statistical Hypothesis testing - PowerPoint PPT Presentation
Citation preview
Epistemology of natural sciences
Dr Tim DawSchool of International Development University
of East [email protected]
Overview• Epistemology – the nature of knowledge in
natural sciences• The ‘scientific method’– Popper – Falsification, Deduction– Fisher - Statistical Hypothesis testing– Quantification and statistics applied to Fisher and
Popper’s ideas• A typical ‘scientific study’– Tillman et al
• Problems with Null Hypothesis statistical testing• Observational and modelling studies• Complexity science, Systems ecology, Resilience
How do you know what you know?
How do know whether it is right?
What qualifies as Knowledge?
Monkeys can evaluate the reliability of their knowledge!How do scientists do it?
‘The scientific method’• Basically POSITIVIST– One reality is out there – there is a ‘truth’– Objective research is possible• Results depend on and reflect the nature of reality, not
the nature of the researcher
• Use of quantification and statistics to objectively describe reality
• Generally REDUCTIONIST– Examine the effects of one factor at a time...
Should the science of nature have a different epistemology to the science of human societies
or economies?
Induction
TheoryY is determined by X
What theory can explain the nature of the data?
Empirical research
Data
0
5
10
15
20
25
30
35
40
45
0 5 10 15 20 25
Y
X
Deduction
TheoryY is determined by X
Does this data support the theory?
Empirical research
Data
0
5
10
15
20
25
30
35
40
45
0 5 10 15 20 25
Y
X
Induction or Deduction?• Advantages of induction...• Disadvantages• What are you using in your research?• When would induction be useful?• When would deduction be useful?
Explanation Testing
Theory
Data
DEDUCTION
INDUCTION
Popper – Science is...• ‘Scientific’ and ‘unscientific statements’• Theory can’t be proved, only disproved
‘The sun will always rise’• Scientific statements must be falsifiable• Science should be the process
of trying to disprove theories• Natural selection of theories that
are not disproven
Karl Popper 1902 -1994
Applying numbers to deduction...
• Ronald Fisher• Provided mathematical framework to
implement Popper’s falsification– Null hypothesis, H0
– Statistical testing– ‘Significance’
Ronald Fisher 1890-1962
Deduction with null hypothesis testing
Theory/HypothesisY is determined by X
What is the probability of data if Ho is true?
e.g. P = 5% (unlikely)
Experimental research
Null Hypothesis H0
Y is unrelated to X
H0 is unlikely to be true...
Hypothesis is supported
Data
0
5
10
15
20
25
30
35
40
45
0 5 10 15 20 25
Y
X
X = treatmentY = response
How unlikely should the data be in order to reject the null
hypothesis?Why?
Tillman et al (1997)• Background: Species extinction rates are ~1000 higher
than background rates• Theory: Biodiversity is important for ecosystem function• Hypothesis: Changes in diversity will affect ecological
processes• Treatment variables – Spp diversity and Func diversity• Response variables – e.g. Biomass, Nutrients cycling etc• Experiment: Manually manipulate diversity (treatment
variable) and measure processes (response variables)
• H0 – There is no relationship between diversity and
processes
• Hypothesis: Biomass is a function of diversity• Biomass = an effect of Diversity + base level• Biomass ~ Diversity + Intercept• Null Hypothesis (H0):– The effect of diversity is zero
Treatment variable
Resp
onse
var
iabl
e
Response ~ Spp Div + Func diversity + ‘intercept’
‘Non significant’ – H0 not rejectedProbability (of the data) if H0 is true is low
p < 1%, result is ‘significant’H0 is rejected -> Theory is supported
But actually experiments are difficult in ecology?
Graham et al 2008
What are the epistemological implications of
- observational studies?- Modelling studies?
Null hypothesis statistical tests dominate the ecological literature
NHSTInf theoreticOtherNHST onobservations
Even though ecologists often have to rely on observational data
Issues with H0 statistical tests
• Mis-interpretation– ‘proving’ the null hypothesis– Focus on the ‘p-values’– Incomplete reporting and publication bias
• Philosophical issues– Binary approach – Significant or not
Is that really the important question?
Stephens et al (2006)
Use of modelling• Some of the most important Qs are not even observable• What will be the effect of ocean acidification on marine
fisheries?
Sumaila et al 2011
Alternative inferences• ‘Information theoretic’ approaches (Burnham
& Anderson 2002)• Compare alternative models (theories)...
Graham et al 2003
‘New Ecology’: changing epistemologies
• Complexity (see Berkes et al 2003)– Multiple interacting factors– Uncertainty– non-linear relationships
• Human environment linkages– Social-Ecological Systems (Berkes et al 2003)– Political Ecology– More holistic approaches– Epistemological pluralism (Miller et al 2008)– Broader range of knowledges used (e.g. Local
knowledge)
What are the epistemological implications for ecologists studying linked social-ecological systems?
Reductionism or Holism
Carpenter et al 2009
Salvador Dalí - Nature Morte Vivante (Still Life - Fast Moving) (1956) Oil on canvas
Range of epistemological approaches in natural sciences
Theoretical Empirical
AbstractGeneral
Context specific
Holistic Reductive
Experimental Observational
Holistic Reductive
Where does Tillman et al fit? For the most Natural scientist in your group – where does their study fit?
References• Berkes F, Colding J, Folke C (2003) Navigating social-ecological
systems: building resilience for complexity and change. Cambridge Univ Pr
• Carpenter SR, Folke C, Scheffer M, Westley F (2009) Resilience: accounting for the noncomputable. Ecology and Society 14:13
• Kornell N, Son LK, Terrace HS (2007) Transfer of Metacognitive Skills and Hint Seeking in Monkeys. Psychological Science 18:64 -71
• Miller TR, Baird TD, Littlefield CM, Kofinas G, Chapin III FS, Redman CL (2008) Epistemological pluralism: reorganizing interdisciplinary research. Ecology and Society 13:46
• Stephens PA, Buskirk SW, Rio CM del (2007) Inference in ecology and evolution. Trends in Ecology & Evolution 22:192-197
• Sumaila UR, Cheung WWL, Lam VWY, Pauly D, Herrick S (2011) Climate change impacts on the biophysics and economics of world fisheries. Nature Clim Change advance online publication
• Tilman D, Knops J, Wedin D, Reich P, Ritchie M, Siemann E (1997) The Influence of Functional Diversity and Composition on Ecosystem Processes. Science 277:1300 -1302