Uncertainty and Complexity in Clinical Decisions
Roger Strand, building upon work with Guri Rørtveit, Yngvild S. Hannestad & Edvin Schei University of Bergen, Norway
Part I: Risk, strict uncertainty and ignorance in clinical decisions
Risk, probability and uncertainty
Why probabilities are important
Why concepts of probability differ
Orthodox and Bayesian views on probability
Ignorance
B. Wynne (1992),
Global Environmental Change, 2:111-127.
B. Wynne (1992),
Global Environmental Change, 2:111-127.
Risk, strict uncertainty and ignorance
Risk: when we know (can quantify) the probabilities
(Strict) uncertainty: the event space is known, but the probabilities cannot be estimated
Frank Knight (1921): Risk, Uncertainty and Profit
There can be no stock market without strict uncertainty
Enlightenment model of reason
Difficult practical problem
Small and manageable technical problems Decomposition
+ + +
Modern rationality: predict-then-act
Big practical
problem
Technical problem #1
FACTS VALUES
Decision (Solution)
Facts of sufficient quality
Certainty? (Descartes)
Sources of uncertainty
Sources of uncertainty
Incomplete/imperfect observations
Incomplete conceptual frameworks
Inaccurate prescriptions of known processes (poor parameterisations etc)
Chaos
Lack of predictability
Uncertainty
Facts of sufficient quality
Certainty? (Descartes)
Probabilities
Pascal and the Jesuite solution
The plausibility of signs that are often seen (cf Ian Hacking)
The orthodox concept: Probability = Frequency
Frequencies may not exist
Bayesian methods: combining frequencies and degrees of belief
Problem: Why trust degrees of belief?
Ignorance
Unknown parts of event space
There are the known unknowns, and the unknown unknowns…
DDT, thalidomide, diethylstilbestrol (DES)
In clinical decisions: Indirect effects that are not categorised as «adverse effects»
Less energy, less initiative, «brain fog»!? (e.g. statins)
Promiscuity!? (e.g. anti-depressants)
B. Wynne (1992),
Global Environmental Change, 2:111-127.
Indeterminacy
In clinical decisions:
risks: probabilities known for patient groups
uncertainties: Is my patient representative? Or too
different? Of which relevant peculiarities am I ignorant?
ignorance: What other consequences will my decision
have, than main effect and medically studied side-effects?
Which of these feed back into health?
indeterminacy: How did we define the system and the problem?
Indeterminacy
Causal chains or networks are open
Different system definition → different
sources of risks
sources of uncertainties
border with ignorance
Trade-offs: narrowing the problem may decrease uncertainty at the expense of ignorance
Indeterminacy = Complexity ??
Causal chains or networks are open
Different system definition → different
sources of risks
sources of uncertainties
border with ignorance
R. Strand, G. Rørtveit
& E. Schei (2005), ComplexUs, 2:2-6.
Complex Systems & Human Complexity
Complex systems: “thin” complexity
nonlinear systems of many agents following rules
Agent-based models; complex adaptive systems
paradigms; neural networks; self-organised criticality…
Complex Systems & Human Complexity
Complex systems: “thin” complexity
nonlinear systems of many agents following rules
Human complexity: “thick” complexity
self-awareness, interpretation, self-deception, creativity
self-fulfilling and self-destructive prophecies…
Complex Systems & Human Complexity
Complex systems: “thin” complexity
nonlinear systems of many agents following rules
Human complexity: “thick” complexity
self-awareness, interpretation, self-deception, creativity
self-fulfilling and self-destructive prophecies…
Human complexity and medical care
simplicity: health by a technical fix
complex systems: healthy attractor patterns
human complexity: dialogue, negotiation, mutual learning
The ideal of algorithmic rationality
Big practical
problem
Technical problem #1
FACTS VALUES
Decision (Solution)