Basics of Causal Loop Diagrams: Structure & Behavior Nathaniel Osgood (osgood@cs.usask.ca) Institute for System Science & Health
Causal Loop Diagram Food Ingested Hunger Focuses on capturing causality and especially feedback effects Indicates sign of causal impact ( vs. )) x y indicates x y indicates yy 0 x > y < 0 x
Meaning of A B We are reasoning here about causal influences: The changes on B caused by changes in A We are interested in causal influence because we want to be able to reason about what if counterfactual situations This should not merely be An associational relationship (e.g. correlation absent causation) If unsure, remember that models are useful as capturing dynamic hypotheses about how things work and generally work well as long as we keep these limits in mind A matter of definition e.g. the larger the # afflicted by condition X, the smaller the number without X
Identifying Link Polarity Within a causal loop diagram, all linkages should be associated with a polarity ( vs ) Tips for identifying X Y link polarity Reason about this link in isolation Do not be concerned about links preceding X or following Y Ask if X were to INCREASE, would Y increase or decrease compared to what it would otherwise have been? a Increase in Y implies,decrease in Y implies If answer is not clear or depends on value of X, need to think about representing several paths between X and Y
Critical: Notion of Y Increases Must Clearly Distinguish Correct interpretation for causal loop diagrams: if X were to INCREASE, would Y increase or decrease compared to what it would have otherwise been? Different notion (common misunderstanding of link polarities): if X were to INCREASE, would Y increase or decrease over time?? i.e. if X were to INCREASE, would Y rise or fall over time?
Polarity A B Does not mean that if A rises then B will rise over time Just says that B will be higher than it would otherwise have been B may still be declining over time but is larger than it would otherwise have been, w/o increasing A A B Does not mean that if A rises then B will decline over time Just says that B will be lower than it would otherwise have been B may still be rising over time but is higher than it otherwise would have been without increasing A
Causal Pathways We can reason about the influence of one variable and another variable by examining the signs along their causal pathway Two negatives (whether adjacent or not) will act to reverse each other Consider A B C An increase to A leads B to be less than it otherwise would have been B being lower than it otherwise would have been causes C to be higher than it otherwise would have been (compared to what it otherwise would have been)
Tips Variables will often be noun phrases Variables should be at least ordinal Links should have unambiguous polarity Indicate pronounced delays Avoid megadiagrams g Label loops Distinguish perceived and actual situation Incorporate targets of balancing loops Try to stick to planar graphs Diagrams describe causal not casual factors!
Ambiguous Link Ambiguous Link: Sometimes, sometimes Food intake Energy egysupus Surplus Rate of Weight Gi Gain Replace this by disaggregating causal pathways by showing multiple links Food intake Calories Taken in Basal Metabolism Calories Burrt Energy egysupus Surplus Rate of Weight Gi Gain
Example 2 Ambiguous Link: Sometimes, sometimes Overtime Work Accomplished per Day Replace this by disaggregating causal pathways by showing multiple l links Overtime Fatigue Greater Incorporation of Outside Tasks at Work Efficiency Work Accomplished per Day More Time Working
Example 3 Ambiguous Link: Sometimes, sometimes Proportion of Fat Calories Ingested in Foods Replace this by disaggregating causal pathways by showing multiple links Proportion of Fat in Foods Satiety Diet Caloric Density Amount of Food Eaten Calories Ingested
Feedback Loops Loops in a causal loop diagram indicate feedback in the system being represented Qualitatively speaking, this indicates that a given change kicks k off a set of changes that t cascade through other factors so as to either amplify ( reinforce ) or push back against ( damp, balance ) the original change Loop classification: product of signs in loop (best to trace through conceptually) Balancing loop: Product of signs negative Reinforcing loop: Product of signs positive
Example Vicious/Virtuous Cycles Positive ( reinforcing, amplifying ) feedback can lead to extremely rapid changes in situation # of Infectives Individual Target Weight # New Infections Weight Perceived as Normal Prevalence of Obesity Prevalence of GDM Prevalence of Macrosomic Infants Mean Weight in Population # of Activated Memory Cells # of Clonal Expansions
Example Balancing Loops Balancing ( regulatory, negative feedback ) loops tend to be selfregulating # New Infections # of Susceptibles Adaptation Policy Reevaluation Policy Effectiveness Food Ingested Hunger Mistakes Learning from Mistakes
Best Practice: Incorporating Thresholds Balancing loops tend to be selfregulating Policy Reevaluation ad to Threshold Hunger to Motivate Eating Policy Adaptation Policy Effectiveness Food Ingested Treshold for Policy Dissatisfaction to Lead to Action Hunger
Best Practice: Indicating (Pronounced) Delays Balancing loops tend to be selfregulating Policy Reevaluation Threshold Hunger to Motivate Eating Policy Adaptation Policy Effectiveness Treshold for Policy Dissatisfaction to Lead to Action Food Ingested Hunger
Elaborating Causal Loops Creation of Nutrition and Exercise Programs Prevalence of Obesity Prevalence of GDM Study of Obesity Prevalence of Macrosomic Infants # of Infectives #N New Infections # of Susceptibles
More Elaborate Diagrams Karanfil, 2008
More Elaborate Diagrams 2 LaVallee & Osgood, 2008
Causal Loop Structure : Dynamic Implications Each loop in a causal loop diagram is associated with qualitative dynamic behavior Most Common SingleLoop Modes of Dynamic Behavior Exponential growth Goal Seeking Adjustment Oscillation When composed, get novel behaviors due to shifting loop dominance Behaviour of system more than sum of parts e.g. Growth & Plateau, Boom & Bust, Lockin
CL Dynamics: Exponential Growth (First Order Reinforcing i Loop) # of Infectives Individual Target Example # New Infections Dynamic implications Weight Perceived as Normal Weight Mean Weight in Population
CL Dynamics: Goal Seeking ad to (Balancing Loop) Example: Food Ingested Threshold Hunger to Motivate Eating Dynamic behavior Hunger
Causal Loop Dynamics: Oscillation (Balancing Loop with Delay) ) Causal Structure Dynamic Behavior:
Growth and Plateau Potential Existing Users Loop structure: Reinforcing Loop Balancing Loop Dynamic Behavior: Customers Word of Mouth Sales 100,000 Customers Graph for Customer New Users Discovering Site Likelihood of Cross Listing and Listing on Search Engines Internet Users Yet to Discover Site 75,000 50,000 From Tsai 25,000 0 0 10 20 30 40 50 60 70 80 90 100 Time (Month) Customer : Current
Complexities & Regularities Department of Computer Science
Feedbacks Driving Infectious Disease Dynamics S u s c e p t ib le s C o n t a c t s b e t w e e n S u s c e p t ib le s a n d Infe c tive s N e w I n fe c t io n s I n fe c t iv e s N e w R e c o v e r ie s
2,000 people 600 people 10,000 people Example Dynamics of SIR Model (No Births or Deaths) SIR Example 1,500 people 450 people 9,500 people 1,000 people 300 people 9,000 people 500 people 150 people 8,500 people 0 people 0 people 8,000 people 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 Time (days) Susceptible Population S : SIR example Infectious Population I : SIR example Recovered Population R : SIR example people people people
Shifting Feedback Dominance 2,000 people 600 people 10,000 people S u s c e p t ib le s SIR Example 1,500 people 450 people 9,500 people o n ta c ts b e tw e e n S uscep tib les and I n fe c t iv e s In fe c tiv e s N e w I n fe c t io n s 1,000 people 300 people 9,000 people S u s c e p t ib le s N e w R e c o v e r ie s S u s c e p t ib le s 500 people 150 people 8,500 people C o n t a c t s b e t w e e n S u s c e p t ib le s a n d Infe c tive s I n fe c t iv e s N e w I n f e c t io n s o n t a c t s b e t w e e n S u s c e p t ib le s a n d In fe c tiv e s I n fe c t iv e s N e w I n fe c t io N e w R e c o v e r ie s 0 people 0 people 8,000 people 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 Time (days) N e w R e c o v e r ie s Susceptible Population S : SIR example Infectious Population I : SIR example Recovered Population R : SIR example people people people
Shifting Feedback Dominance 2,000 people 600 people 10,000 people S u s c e p t ib le s SIR Example 1,500 people 450 people 9,500 people o n ta c ts b e tw e e n S uscep tib les and I n fe c t iv e s In fe c tiv e s N e w I n fe c t io n s 1,000 people 300 people 9,000 people S u s c e p t ib le s N e w R e c o v e r ie s S u s c e p t ib le s 500 people 150 people 8,500 people C o n t a c t s b e t w e e n S u s c e p t ib le s a n d Infe c tive s I n fe c t iv e s N e w I n f e c t io n s o n t a c t s b e t w e e n S u s c e p t ib le s a n d In fe c tiv e s I n fe c t iv e s N e w I n fe c t io N e w R e c o v e r ie s 0 people 0 people 8,000 people 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 Time (days) N e w R e c o v e r ie s Susceptible Population S : SIR example Infectious Population I : SIR example Recovered Population R : SIR example people people people
Common Phenomena In Complex Systems Counterintuitive behaviour (Often the result of interactions between feedbacks [fb]) Snowballing: When things go bad, they often go very bad very quickly Vicious cycles lead to cascading of problems (Due to positive feedback) Path dependence : Different starting points can lead to divergence in project progress (Due to positive feedback interacting w/ mult. negative fb) Policy resistance: Situation can be unexpectedly difficult to change (Typically due to negative feedbacks that resist change) Lockin: Waiting raises barriers to improvement (Due to positive feedback interacting w/ mult. negative fb)
Issues with Causal Loop Unclear variables Diagrams Diagrams can become very large Confusion regarding polarity Noncausal relationship Conservation not captured Behavior not always same as archetype Missing causal factors Missing i links Asymmetry in direction of change
Unclear Variables Variables Lacking Clear Polarity Gender Ethnicity Shape Often categorical & nonordinal Implicit Polarity Population (size) Revenue (amount of) Sound, Color (more of) Socioeconomic status (greater, lesser) Ask whether more X is Meaningful Unambiguous
Very Large Diagrams http://kim.foresight.gov.uk/obesity/obesity.html Still useful for getting big picture identifying where research fits in, research gaps
Artifactual Loop
Artifactual Loop 2
Artifactual Loop 3