Causality in the Sciences:
a gentle introduction
Federica Russo
Center Leo Apostel, Vrije Universiteit Brussel &
Centre for Reasoning, University of Kent
Overview
Motivations to adopt a causal stance
The CitS spproach to causation
Towards a mosaic of causal theory
Accounts of cause / causation
2
WHY ADOPTING
A CAUSAL APPROACH
3
Goals of causal analysis
Knowledge-oriented
Understand and explain a
phenomenon of interest
Diagnose dysfunction a
designed system
Action-oriented
Predict, intervene on, control
a phenomenon of interest
Design / model / debug a
system
4
Causal assessment
‘What Causes What’
Establish
what’s the cause of a patient’s illness
who is (legally / morally) responsible for some state of affairs
what are the causes of a given phenomenon
what causes dysfunction in an organisation
which pathways explain some cellular behaviour
…
5
Do causes need to be causes?
Consider:
Smoking and cancer are associated. Should I quit smoking?
Smoking causes cancer. Should I quit smoking?
Causes trigger actions. Mere beliefs can’t, nor mere associations.
[What about risk factors, then? A debated issue in medical research.]
6
Source: http://xkcd.com/552/
Causes make things change
Demographic or economic trends
Social, economic or public health policy
The outcome of a physical theory
Interventions in a designed system
Interventions in a lab experiment
When we know the causes / mechanisms,
we can plan to make things change
within a certain margin, of course
7
THE CITS APPROACH TO CAUSATION
8
Analysis of scientific practice
Growing!
Examples
Causal assessment in medicine
Causal reasoning in quantitative social science
Diagnosis in enterprise engineering
…
The ‘Causality in the Sciences’ research trend
Philosophical questions about causation (and other topics) are motivated
by methodological and practical problems in real science.
Start from scientific practice to bottom up philosophy.
9
Examples of CitS theses
[Medicine]
Causal assessment has two evidential components:
mechanisms and difference-making
[Social Science]
‘Variation’ guides causal reasoning in model building and model testing
[Biology, Neuroscience]
Mechanisms are entities and activities organised in such a way that
they are responsible for a phenomenon
[Evidence-based approaches]
Evidence hierarchies should not neglect evidence of
mechanisms and expert knowledge
…
10
TOWARDS A CAUSAL MOSAIC
11
Biomarkers research
Environmental exposure and disease
Measurement at molecular level
Identify biomarkers of exposure, of early clinical changes, of disease
From GWAS to EWAS
Better understanding of disease mechanisms
Better prediction and health policy planning
12
The ‘biomarkers’ mosaic
Meeting in the middle methodology
Relation between background knowledge and new causal discoveries.
Exposome
New (causal) concept; redefines the causal context of causal relations at different
levels take place.
Processes
Studying the evolution of biomarkers is tracing processes of molecular changes
Difference-making and mechanisms
Their interplay is crucial to establish causal relations:
Biomarkers of disease make a difference in the probability of disease
This probability raising is substantiated by a plausible mechanism.
13
The biomarkers mosaic (cont’ed)
Production-Information
A test case for an account of production in terms of information
Levels of causation
EWAS for the population, what about the individual?
Integration of factors of different nature (social and biological) into the
same explanatory framework.
Capacity
A test case for capacities: is the predictive power of a biomarker due to a
capacity of some chemicals to induce molecular changes?
14
CONCEPTS OF CAUSE / CAUSATION
15
Difference-making:
probabilistic causality
Examples
Smoking increases the probability of developing cancer.
Physical exercise prevents heart attacks.
Cancer and yellow fingers are correlated, but both are effects of smoking.
Core ideas
P(A|B) > P(A) (positive cause)
P(A|B) < P(A) (negative cause)
Principle of common cause: if A and B are correlated but are not causes of each
other, there must be a third event C that causes both
Pioneered by P. Suppes. Still the basis of any account involving probabilities.
Helps with, e.g.:
Causal inference in medical datasets analysed with temporal probabilistic logic
16
Difference-making:
counterfactuals
Example
Missing the train caused me to miss the seminar.
Had I not missed the train, I would not have missed the seminar.
Core idea
A causes B iff, had A not been, B would not have been either.
Pioneered by D. Lewis. Still the basis of any account involving counterfactual, including
the “potential outcome” approach in statistics
Helps with, e.g.:
Heuristic search for causal hypotheses to test
17
Difference-making:
manipulability theories
Example
Consider the ideal gas law, were we to manipulate the pressure of the gas, the
volume would accordingly change
Core idea
A causes B iff, were we to manipulate A, B would accordingly change.
Main supporter: J. Woodward. Widely adopted in current philosophy of science.
Helps with, e.g.:
Explanation of structure of lab experiments
18
Physical connections:
physical processes
Example
Billiard balls colliding (causal process)
Airplane shadows crossing (pseudo-process)
Core idea
A causes B if there is a physical process connecting the two points.
The transmission of extensive quantities discriminate between a causal and a
pseudo-process
Main supporters: W. Salmon, P. Dowe. More recently: G. Boniolo, R. Faraldo and A.
Saggion
Helps with, e.g.:
Causal inference in physics contexts
19
Physical connections:
mechanisms
Examples
Protein synthesis
Circadian rhythms
Core ideas
A causes B iff there is mechanism linking A to B
A mechanism is an arrangement of entities and activities producing some
phenomenon
Main contemporary supporters: Machamer et al, Bechtel et al, Glennan, …
Remote supporters: Decartes, Newton, …
Helps with, e.g.:
Conceptualising interactions in complex systems
20
Regularity
Example:
Every time I push the button the bulb lights up.
Core ideas
Causes are events that regularly precede their effect
in space and time.
We infer that A causes B from the observation
that B regularly follows A.
Most famously: D. Hume. More recently: S. Psillos, M. Baumgartner, …
Helps with, e.g.:
Causal inference of generic causal claims
21
Necessary and sufficient components
Example:
Short circuit caused house fire. Not on its own, but in conjunction with other
factors and in a given background. It is however not redundant because the
other parts are not sufficient to cause fire. The whole thing is itself not
necessary. [Analogous: gene knockout experiments]
Core ideas
Causes are, at minimum, INUS conditions:
“Insufficient but Non-redundant part of a condition
which is itself Unnecessary but Sufficient”
Most famously: J.L. Mackie. In epidemiology: Rothman’s pie charts.
Helps with, e.g.:
Conceptualisation of complex causal settings (disease, concurrent systems, …)
22
Capacities, powers, dispositions
Example
Aspirin has the capacity to relieve headache
Cells have the capacity to metabolise lactose
Core idea
Entities or systems have the capacity, power or disposition to bring about effects
Main supporter: N. Cartwright, S. Mumford, A. Bird, A. Chakravartti. Remote
supporter: an Aristotelian concept!
Helps with, e.g.:
Explain properties of biological systems
23
Inferentialism
Example
“H. Pylori causes gastric ulcer” is inferred from multifaceted evidence; it allows
certain kinds of inferences. But it does not correspond to anything ‘out there’
Core idea
Causation is an inferential map by means of which we chart the world
Main supporter: J. Williamson; J. Reiss
Helps with, e.g.:
Account for inferential practices in science
24
Some causal riddles
Are omissions causes?
The gardener failed to water my plant, that subsequently died.
What entity is not watering?
What process can there be from ‘not watering’ to ‘dying’?
The King didn’t water the plant either. Is he also a cause of my plant dying?
Are non-manipulable factors causes?
Gender is a cause of salary discrimination;
Ethnicity is a cause of HIV infections is sub-Saharan Africa.
But such factors cannot undergo experimental manipulation.
Are they rightly called ‘causes’?
25
TO SUM UP AND CONCLUDE
26
Why going ‘CitS’
The philosophy of causality is becoming a full-blown discipline.
‘Causality in the Sciences’ has a strong empirical and
methodological basis.
Whence its usefulness and importance for science
Time for a reassessment of the numerous causal theories
Abandon the causal theory
Embrace the CAUSAL MOSAIC
27
The tiles in the mosaic
Causal notions are placed in the causal mosaic according to their role
Conceptual / metaphysical
Epistemological
Methodological
Use causal accounts to complement scientific tasks
Inference
Explanation
Prediction / control
Conceptualisation
28
Useful resources
Causality in the Sciences Conference Series
http://www.kent.ac.uk/secl/philosophy/jw/cits.htm
Stay tuned for Paris conference in July 2013
Causality in the Sciences, OUP 2011 (Illari, Russo, Williamson eds)
Research papers on causality from the sciences2929
The Oxford Handbook of Causation, OUP 2010 (Beebee, Hitchcock, Menzies eds)
Introduction to ‘traditional’ causal theories
Judea Pearl’s causality blog
http://www.mii.ucla.edu/causality/
A lot on Bayesian nets approach
29
Phyllis Illari and Federica Russo
CAUSALITY:
PHILOSOPHICAL THEORY MEETS SCIENTIFIC PRACTICE
Coming soon at
Oxford University Press

Causality in the sciences: a gentle introduction.

  • 1.
    Causality in theSciences: a gentle introduction Federica Russo Center Leo Apostel, Vrije Universiteit Brussel & Centre for Reasoning, University of Kent
  • 2.
    Overview Motivations to adopta causal stance The CitS spproach to causation Towards a mosaic of causal theory Accounts of cause / causation 2
  • 3.
  • 4.
    Goals of causalanalysis Knowledge-oriented Understand and explain a phenomenon of interest Diagnose dysfunction a designed system Action-oriented Predict, intervene on, control a phenomenon of interest Design / model / debug a system 4
  • 5.
    Causal assessment ‘What CausesWhat’ Establish what’s the cause of a patient’s illness who is (legally / morally) responsible for some state of affairs what are the causes of a given phenomenon what causes dysfunction in an organisation which pathways explain some cellular behaviour … 5
  • 6.
    Do causes needto be causes? Consider: Smoking and cancer are associated. Should I quit smoking? Smoking causes cancer. Should I quit smoking? Causes trigger actions. Mere beliefs can’t, nor mere associations. [What about risk factors, then? A debated issue in medical research.] 6 Source: http://xkcd.com/552/
  • 7.
    Causes make thingschange Demographic or economic trends Social, economic or public health policy The outcome of a physical theory Interventions in a designed system Interventions in a lab experiment When we know the causes / mechanisms, we can plan to make things change within a certain margin, of course 7
  • 8.
    THE CITS APPROACHTO CAUSATION 8
  • 9.
    Analysis of scientificpractice Growing! Examples Causal assessment in medicine Causal reasoning in quantitative social science Diagnosis in enterprise engineering … The ‘Causality in the Sciences’ research trend Philosophical questions about causation (and other topics) are motivated by methodological and practical problems in real science. Start from scientific practice to bottom up philosophy. 9
  • 10.
    Examples of CitStheses [Medicine] Causal assessment has two evidential components: mechanisms and difference-making [Social Science] ‘Variation’ guides causal reasoning in model building and model testing [Biology, Neuroscience] Mechanisms are entities and activities organised in such a way that they are responsible for a phenomenon [Evidence-based approaches] Evidence hierarchies should not neglect evidence of mechanisms and expert knowledge … 10
  • 11.
  • 12.
    Biomarkers research Environmental exposureand disease Measurement at molecular level Identify biomarkers of exposure, of early clinical changes, of disease From GWAS to EWAS Better understanding of disease mechanisms Better prediction and health policy planning 12
  • 13.
    The ‘biomarkers’ mosaic Meetingin the middle methodology Relation between background knowledge and new causal discoveries. Exposome New (causal) concept; redefines the causal context of causal relations at different levels take place. Processes Studying the evolution of biomarkers is tracing processes of molecular changes Difference-making and mechanisms Their interplay is crucial to establish causal relations: Biomarkers of disease make a difference in the probability of disease This probability raising is substantiated by a plausible mechanism. 13
  • 14.
    The biomarkers mosaic(cont’ed) Production-Information A test case for an account of production in terms of information Levels of causation EWAS for the population, what about the individual? Integration of factors of different nature (social and biological) into the same explanatory framework. Capacity A test case for capacities: is the predictive power of a biomarker due to a capacity of some chemicals to induce molecular changes? 14
  • 15.
    CONCEPTS OF CAUSE/ CAUSATION 15
  • 16.
    Difference-making: probabilistic causality Examples Smoking increasesthe probability of developing cancer. Physical exercise prevents heart attacks. Cancer and yellow fingers are correlated, but both are effects of smoking. Core ideas P(A|B) > P(A) (positive cause) P(A|B) < P(A) (negative cause) Principle of common cause: if A and B are correlated but are not causes of each other, there must be a third event C that causes both Pioneered by P. Suppes. Still the basis of any account involving probabilities. Helps with, e.g.: Causal inference in medical datasets analysed with temporal probabilistic logic 16
  • 17.
    Difference-making: counterfactuals Example Missing the traincaused me to miss the seminar. Had I not missed the train, I would not have missed the seminar. Core idea A causes B iff, had A not been, B would not have been either. Pioneered by D. Lewis. Still the basis of any account involving counterfactual, including the “potential outcome” approach in statistics Helps with, e.g.: Heuristic search for causal hypotheses to test 17
  • 18.
    Difference-making: manipulability theories Example Consider theideal gas law, were we to manipulate the pressure of the gas, the volume would accordingly change Core idea A causes B iff, were we to manipulate A, B would accordingly change. Main supporter: J. Woodward. Widely adopted in current philosophy of science. Helps with, e.g.: Explanation of structure of lab experiments 18
  • 19.
    Physical connections: physical processes Example Billiardballs colliding (causal process) Airplane shadows crossing (pseudo-process) Core idea A causes B if there is a physical process connecting the two points. The transmission of extensive quantities discriminate between a causal and a pseudo-process Main supporters: W. Salmon, P. Dowe. More recently: G. Boniolo, R. Faraldo and A. Saggion Helps with, e.g.: Causal inference in physics contexts 19
  • 20.
    Physical connections: mechanisms Examples Protein synthesis Circadianrhythms Core ideas A causes B iff there is mechanism linking A to B A mechanism is an arrangement of entities and activities producing some phenomenon Main contemporary supporters: Machamer et al, Bechtel et al, Glennan, … Remote supporters: Decartes, Newton, … Helps with, e.g.: Conceptualising interactions in complex systems 20
  • 21.
    Regularity Example: Every time Ipush the button the bulb lights up. Core ideas Causes are events that regularly precede their effect in space and time. We infer that A causes B from the observation that B regularly follows A. Most famously: D. Hume. More recently: S. Psillos, M. Baumgartner, … Helps with, e.g.: Causal inference of generic causal claims 21
  • 22.
    Necessary and sufficientcomponents Example: Short circuit caused house fire. Not on its own, but in conjunction with other factors and in a given background. It is however not redundant because the other parts are not sufficient to cause fire. The whole thing is itself not necessary. [Analogous: gene knockout experiments] Core ideas Causes are, at minimum, INUS conditions: “Insufficient but Non-redundant part of a condition which is itself Unnecessary but Sufficient” Most famously: J.L. Mackie. In epidemiology: Rothman’s pie charts. Helps with, e.g.: Conceptualisation of complex causal settings (disease, concurrent systems, …) 22
  • 23.
    Capacities, powers, dispositions Example Aspirinhas the capacity to relieve headache Cells have the capacity to metabolise lactose Core idea Entities or systems have the capacity, power or disposition to bring about effects Main supporter: N. Cartwright, S. Mumford, A. Bird, A. Chakravartti. Remote supporter: an Aristotelian concept! Helps with, e.g.: Explain properties of biological systems 23
  • 24.
    Inferentialism Example “H. Pylori causesgastric ulcer” is inferred from multifaceted evidence; it allows certain kinds of inferences. But it does not correspond to anything ‘out there’ Core idea Causation is an inferential map by means of which we chart the world Main supporter: J. Williamson; J. Reiss Helps with, e.g.: Account for inferential practices in science 24
  • 25.
    Some causal riddles Areomissions causes? The gardener failed to water my plant, that subsequently died. What entity is not watering? What process can there be from ‘not watering’ to ‘dying’? The King didn’t water the plant either. Is he also a cause of my plant dying? Are non-manipulable factors causes? Gender is a cause of salary discrimination; Ethnicity is a cause of HIV infections is sub-Saharan Africa. But such factors cannot undergo experimental manipulation. Are they rightly called ‘causes’? 25
  • 26.
    TO SUM UPAND CONCLUDE 26
  • 27.
    Why going ‘CitS’ Thephilosophy of causality is becoming a full-blown discipline. ‘Causality in the Sciences’ has a strong empirical and methodological basis. Whence its usefulness and importance for science Time for a reassessment of the numerous causal theories Abandon the causal theory Embrace the CAUSAL MOSAIC 27
  • 28.
    The tiles inthe mosaic Causal notions are placed in the causal mosaic according to their role Conceptual / metaphysical Epistemological Methodological Use causal accounts to complement scientific tasks Inference Explanation Prediction / control Conceptualisation 28
  • 29.
    Useful resources Causality inthe Sciences Conference Series http://www.kent.ac.uk/secl/philosophy/jw/cits.htm Stay tuned for Paris conference in July 2013 Causality in the Sciences, OUP 2011 (Illari, Russo, Williamson eds) Research papers on causality from the sciences2929 The Oxford Handbook of Causation, OUP 2010 (Beebee, Hitchcock, Menzies eds) Introduction to ‘traditional’ causal theories Judea Pearl’s causality blog http://www.mii.ucla.edu/causality/ A lot on Bayesian nets approach 29
  • 30.
    Phyllis Illari andFederica Russo CAUSALITY: PHILOSOPHICAL THEORY MEETS SCIENTIFIC PRACTICE Coming soon at Oxford University Press