Report

The Benefit of Shifting the Burden 20th MIT-UAlbany-WPI System Dynamics PhD Colloquium April 30, 2010 Joe Hsueh MIT Sloan School of Management Research Questions • Can a formal simulation model reproduce the dynamic behavior as described in the shift the burden archetype? – Dowling, MacDonald and Richardson (1995) • What can formal modeling add to qualitative systems archetype? In particular, should one always strive to eliminate shift the burden as prescribed? Acknowledgement: Nelson Repenning, Peter Senge, John Sterman Shift the Burden Behavior • “An underlying problem generates symptoms that demand attention. But the underlying problem is difficult for people to address, either because it is obscure or costly to confront. So people “shift the burden” of their problem to other solutions – well-intended, easy fixes which seem extremely efficient. Unfortunately, the easier “solutions” only ameliorate the symptoms; they leave the underlying problem unaltered. The underlying problem grows worse, unnoticed because the symptoms apparently clear up, and the system loses whatever abilities it had to solve the underlying problem.” (Senge 1990, p.104) Shift the Burden Structure • “The shift the burden is composed of two balancing (stabilizing) processes. Both are trying to adjust or correct the same problem symptom. The top circle represents the symptomatic intervention; the “quick fix.” It solves the problem symptom quickly, but only temporarily. The bottom circle has a delay. It represents a more fundamental response to the problem, one whose effects take longer to become evident. However, the fundamental solution works far more effectively – it may the only enduring way to deal with the problem.” • “ Often (but not always), in shifting the burden structures there is also an additional reinforcing (amplifying) process created by “side effects” of the symptomatic solution. When this happens, the side effects often make it more difficult to invoke the fundamental solution.” (Senge 1990, p.106) Effect of Problem Symptom on Symptomatic Solution Symptomatic Solution Sensitivity + + B1 Reference Problem Symptom Normal Symptomatic Solution + Symptomatic Solution OUTSOURCING Symptomatic Solution Problem Symptom Relative to Reference Problem Symptom WORKLOAD Problem Creation Rate Problem Solving Rate R1 + Side Effect Fundamental Solution Sensitivity + Available Fundamental Problem Solving Rate + Effect of Problem Symptom on Fundamental Solution Buildup Problem Solving Per Fundamental Solution B2 Fundamental Solution + <Normal Fundamental Solution Buildup> Fundamental Solution Buildup TRAINING Fundamental Solution EMPLOYEE CAPABILITY Fundamental Solution Atrophy SKILL OBSOLESCENCE + + - Atrophy Time <Normal Symptomatic Solution> + Effect of Symptomatic Solution on Fundamental Solution Erosion + Fundamental Solution Erosion from Side Effect CONFIDENCE EROSION Normal Erosion from Side Effect Side Effect Sensitivity Symptomatic Solution = Normal Symptomatic Solution * Effect of Problem Symptom on Symptomatic Solution Effect of Problem Symptom on Symptomatic Solution = (Problem Symptom/Reference Problem Symptom) ^ Symptomatic Solution Sensitivity Effect of Problem Symptom on Symptomatic Solution 2 1.5 1 0.5 0 0 0.20 0.40 0.60 0.80 1 1.20 1.40 1.60 1.80 Problem Symptom / Reference Problem Symptom Sensitivity : 0 Sensitivity : point 3 Sensitivity : point 7 Sensitivity : 1 2 Normal Fraction of Problem Solved by Symptomatic Solution Effect of Problem Symptom on Symptomatic Solution Symptomatic Solution Sensitivity + + + B1 Reference Problem Symptom Normal Symptomatic Solution + + <Exogenous Input Rate> Symptomatic Solution OUTSOURCING Symptomatic Solution Problem Symptom Relative to Reference Fundamental Solution Sensitivity + Problem Solving Delay <Initial Problem + Symptom> Problem Problem Solving Rate Symptom Problem WORKLOAD Creation Rate + Max Problem Min Problem Symptom Exogenous Symptom Solving Rate Input Rate Solving Time - Effect of Problem Symptom on Fundamental Solution Buildup Slack Problem Solving Capability R1 Side Effect Available Fundamental Problem Solving Rate + + Problem Solving Per Fundamental Solution B2 Fundamental Solution + <Normal Fundamental Solution Buildup> Fundamental Solution Buildup + TRAINING <Initial Fundamental Solution> Fundamental Solution EMPLOYEE CAPABILITY Fundamental Solution Slack Buildup + Maximum Erosion Rate Minimum Erosion Time + Fundamental Solution Atrophy SKILL OBSOLESCENCE + + - Atrophy Time <Normal Symptomatic Solution> + Effect of Symptomatic Solution on Fundamental Solution Erosion + Fundamental Solution Erosion from Side Effect CONFIDENCE EROSION Normal Erosion from Side Effect Side Effect Sensitivity Cost Accounting <Symptomatic Solution OUTSOURCING> + Symptomatic Solution Unit Cost Fundamental Solution Buildup Cost as Multiple of Symptomatic Solution Symptomatic Solution Cost Fundamental Solution Cost <Fundamental Solution Buildup TRAINING> <Normal Problem Solving Delay> Problem Solving Delay Unit Cost + Cumulative Total Solution Cost + Fundamental Solution Buildup Unit Cost <Problem Solving Delay> Cumulative Symptomatic Solution Cost Cumulative Fundamental Solution Cost + Cumulative Total Cost + Cumulative Problem Solving Delay Cost Problem Solving Delay Cost <Problem Symptom WORKLOAD> Partial Model Test 1: B1 Symptomatic Solution only • • • Cut B2 Fundamental Solutions and R1 Side Effect loops. Test: Pulse problem creation rate from 10 to 15 at time 10 for 10 months Adjusting Symptomatic Solution (SS) Sensitivity from 0 to 0.5 to 1 • Question: Is shift the burden intended rational? Problem Creation Rate symptom/Month 20 17 14 11 8 0 10 20 30 Problem Creation Rate : SS=1 Problem Creation Rate : SS=point 5 Problem Creation Rate : SS=0 40 50 60 Time (Month) 70 80 90 100 SS sensitivity = 0 SS sensitivity = 0.5 Main Variables Main Variables Main Variables 100 sym 15 cap 2 cap/m 100 sym 15 cap 2 cap/m 100 sym 15 cap 2 cap/m 50 sym 7.5 cap 1 cap/m 50 sym 7.5 cap 1 cap/m 50 sym 7.5 cap 1 cap/m 0 sym 0 cap 0 cap/m 80 100 0 0 sym 0 cap 0 cap/m 80 100 0 0 sym 0 cap 0 cap/m 0 20 40 60 Time (Month) Problem Symptom WORKLOAD : SS=0 Symptomatic Solution OUTSOURCING : SS=0 Fundamental Solution EMPLOYEE CAPABILITY : SS=0 Fundamental Solution Buildup TRAINING : SS=0 20 40 60 Time (Month) SS sensitivity = 1 Problem Symptom sym WORKLOAD : SS=point 5 Symptomatic cap Solution OUTSOURCING : SS=point 5 Fundamental capSolution EMPLOYEE CAPABILITY : SS=point Fundamental cap/mSolution Buildup TRAINING : SS=point 5 20 40 60 Time (Month) 80 Problem Symptom sym WORKLOAD : SS=1 Symptomatic cap Solution OUTSOURCING : SS=1 5Fundamental capSolution EMPLOYEE CAPABILITY : SS=1 Fundamental cap/mSolution Buildup TRAINING : SS=1 100 sym cap cap cap/m Decision Rule: higher problem symptoms (PS), enact higher symptomatic solutions (SS) to reduce the problems. Intended Rationality: Higher SS sensitivity from 0 to 0.5 to 1, higher SS (red) which reduces PS (blue) and restores to equilibrium faster. Insight 1: Without fundamental solution and side effect, shift the burden is intended rational as it reduces problem symptoms as expected. Partial Model Test 2: B1 + B2, No Side Effect • • • Cut R1 Side Effect loop Test: Pulse problem creation rate from 10 to 15 at time 10 for 10 months Adjusting Symptomatic Solution (SS) Sensitivity from 0 to 0.3 to 0.6 to 0.9 • • Question 2: Without side effect, do we see shift the burden dynamics as described? Question 3: Without side effect, is shift the burden always bad? Problem Creation Rate symptom/Month 20 17 14 11 8 0 10 20 30 40 50 60 Time (Month) Problem Creation Rate : No SE, SS=0 Problem Creation Rate : No SE, SS=pt3 Problem Creation Rate : No SE, SS=pt6 Problem Creation Rate : No SE, SS=pt9 70 80 90 100 Higher Symptomatic Solution Sensitivity, higher symptomatic solution (red) and lower fundamental solution (green) Shift the burden does occur. Problem symptom (blue) restores faster to equilibrium. Main Variables Main Variables 100 sym 8 cap 0.2 cap/m 100 sym 8 cap 0.2 cap/m 50 sym 4 cap 0.1 cap/m 50 sym 4 cap 0.1 cap/m 0 sym 0 cap 0 cap/m 0 sym 0 cap 0 cap/m 0 20 40 60 Time (Month) 80 100 Problem Symptom WORKLOAD : No SE, SS=0 Symptomatic Solution OUTSOURCING : No SE, SS=0 Fundamental Solution EMPLOYEE CAPABILITY : No SE, SS=0 Fundamental Solution Buildup TRAINING : No SE, SS=0 sym cap cap cap/m Main Variables 100 sym 8 cap 0.2 cap/m 0 100 sym 8 cap 0.2 cap/m 0 sym 0 cap 0 cap/m 0 sym 0 cap 0 cap/m 40 60 Time (Month) 80 Problem Symptom WORKLOAD : No SE, SS=pt3 Symptomatic Solution OUTSOURCING : No SE, SS=pt3 Fundamental Solution EMPLOYEE CAPABILITY : No SE, SS=pt3 Fundamental Solution Buildup TRAINING : No SE, SS=pt3 100 sym cap cap cap/m 80 Main Variables 50 sym 4 cap 0.1 cap/m 20 40 60 Time (Month) Problem Symptom WORKLOAD : No SE, SS=pt6 Symptomatic Solution OUTSOURCING : No SE, SS=pt6 Fundamental Solution EMPLOYEE CAPABILITY : No SE, SS=pt6 Fundamental Solution Buildup TRAINING : No SE, SS=pt6 50 sym 4 cap 0.1 cap/m 0 20 0 20 40 60 Time (Month) 80 Problem Symptom WORKLOAD : No SE, SS=pt9 Symptomatic Solution OUTSOURCING : No SE, SS=pt9 Fundamental Solution EMPLOYEE CAPABILITY : No SE, SS=pt9 Fundamental Solution Buildup TRAINING : No SE, SS=pt9 100 sym cap cap cap/m 100 sym cap cap cap/m Is shifting the burden necessary bad? It depends. If there is a cost for solving problems slower than expected (Problem Solving Delay Cost), then higher symptomatic solution helps restore equilibrium faster, thus reduces the cumulative costs. Insight 2: Shift the burden is beneficial when (1) no or low enough side effect from symptomatic solution and (2) high enough problem solving delay cost. Costs 2,000 2,000 1,000 1,000 $ $ Costs 0 0 0 20 40 60 Costs Time (Month) 80 100 20 40 60 Costs Time (Month) 80 100 Cumulative Total Cost : No SE, SS=pt6 2,000Symptomatic Solution Cost : No SE, SS=pt6 Cumulative Cumulative Fundamental Solution Cost : No SE, SS=pt6 Cumulative Problem Solving Delay Cost : No SE, SS=pt6 1,000 $ $ Cumulative Total Cost : No SE, SS=0 2,000Symptomatic Solution Cost : No SE, SS=0 Cumulative Cumulative Fundamental Solution Cost : No SE, SS=0 Cumulative Problem Solving Delay Cost : No SE, SS=0 0 1,000 0 0 0 20 40 60 80 Time (Month) Cumulative Total Cost : No SE, SS=pt3 Cumulative Symptomatic Solution Cost : No SE, SS=pt3 Cumulative Fundamental Solution Cost : No SE, SS=pt3 Cumulative Problem Solving Delay Cost : No SE, SS=pt3 100 0 20 40 60 80 Time (Month) Cumulative Total Cost : No SE, SS=pt9 Cumulative Symptomatic Solution Cost : No SE, SS=pt9 Cumulative Fundamental Solution Cost : No SE, SS=pt9 Cumulative Problem Solving Delay Cost : No SE, SS=pt9 100 Full Model: B1+B2+R1 with Side Effect Sensitivity = 1 • • Test: Pulse problem creation rate from 10 to 15 at time 10 for 10 months Adjusting Symptomatic Solution (SS) Sensitivity from 0 to 0.2 to 0.4 to 0.6 to 0.8 to 1 • • Question 4: With side effect, do we see shift the burden dynamics as described? Question 5: With side effect, is shift the burden always bad? Problem Creation Rate 20 symptom/Month 17 14 11 8 0 Problem Problem Problem Problem Problem Problem 10 Creation Rate : Creation Rate : Creation Rate : Creation Rate : Creation Rate : Creation Rate : 20 SS=0 SS=pt2 SS=pt4 SS=pt6 SS=pt8 SS=1 30 40 50 60 Time (Month) 70 80 90 100 SS sensitivity = 0.6 SS sensitivity = 0.8 Main Variables Main Variables Main Variables 120 sym 10 cap 2 cap/m 120 sym 10 cap 2 cap/m 120 sym 10 cap 2 cap/m 60 sym 5 cap 1 cap/m 60 sym 5 cap 1 cap/m 60 sym 5 cap 1 cap/m 0 sym 0 cap 0 cap/m 0 sym 0 cap 0 cap/m 80 100 0 0 sym 0 cap 0 cap/m 80 100 0 0 20 40 60 Time (Month) Problem Symptom WORKLOAD : SS=pt6 Symptomatic Solution OUTSOURCING : SS=pt6 Fundamental Solution EMPLOYEE CAPABILITY : SS=pt6 Fundamental Solution Buildup TRAINING : SS=pt6 20 40 60 Time (Month) Problem Symptom sym WORKLOAD : SS=pt8 Symptomatic cap Solution OUTSOURCING : SS=pt8 Fundamental capSolution EMPLOYEE CAPABILITY : SS=pt8 Fundamental cap/mSolution Buildup TRAINING : SS=pt8 SS sensitivity = 1 20 40 60 Time (Month) Problem Symptom sym WORKLOAD : SS=1 Symptomatic cap Solution OUTSOURCING : SS=1 Fundamental capSolution EMPLOYEE CAPABILITY : SS=1 Fundamental cap/mSolution Buildup TRAINING : SS=1 80 100 sym cap cap cap/m Higher SS sensitivity, higher SS (red) and lower FS (green) Shift the burden does occur. When SS sensitivity = 0.8 SS increases (red) which lowers PS temporarily (blue) however, side effect induced by too many SS causes FS to erode PS increases rapidly need more SS even faster FS erosion … Insight 3: With side effect, shift the burden too much can tip the system into vicious circle. SS sensitivity = 0.4 SS sensitivity = 0.2 Main Variables Main Variables Main Variables 120 sym 10 cap 2 cap/m 120 sym 10 cap 2 cap/m 100 sym 10 cap 2 cap/m 60 sym 5 cap 1 cap/m 60 sym 5 cap 1 cap/m 50 sym 5 cap 1 cap/m 0 sym 0 cap 0 cap/m 0 sym 0 cap 0 cap/m 80 100 0 0 sym 0 cap 0 cap/m 80 100 0 0 20 40 60 Time (Month) Problem Symptom WORKLOAD : SS=pt4 Symptomatic Solution OUTSOURCING : SS=pt4 Fundamental Solution EMPLOYEE CAPABILITY : SS=pt4 Fundamental Solution Buildup TRAINING : SS=pt4 20 40 60 Time (Month) Problem Symptom sym WORKLOAD : SS=pt2 Symptomatic cap Solution OUTSOURCING : SS=pt2 Fundamental capSolution EMPLOYEE CAPABILITY : SS=pt2 Fundamental cap/mSolution Buildup TRAINING : SS=pt2 SS sensitivity = 0 20 40 60 Time (Month) Problem Symptom sym WORKLOAD : SS=0 Symptomatic cap Solution OUTSOURCING : SS=0 Fundamental capSolution EMPLOYEE CAPABILITY : SS=0 Fundamental cap/mSolution Buildup TRAINING : SS=0 80 100 sym cap cap cap/m As SS sensitivity reduces from 0.4 to 0.2 to 0, there is less SS (red), however FS (green) and PS (blue) oscillate! When SS sensitivity = 0 a pulse shock of PS increases FS buildup (grey) FS increases with time delay FS overshoots PS drops significantly cut back FS buildup FS erodes with time delay FS undershoots PS increases significantly …. Insight 4: Appropriate level of shift the burden can dampen FS oscillation, stabilize PS and reduce problem solving delay costs. Cumulative Total Cost 6,000 $ 4,500 3,000 1,500 0 0 10 20 30 40 50 60 Time (Month) 70 80 90 100 Cumulative Total Cost : SS=0 Cumulative Total Cost : SS=pt2 Cumulative Total Cost : SS=pt4 Cumulative Total Cost : SS=pt6 Cumulative Total Cost : SS=pt8 As SS sensitivity goes from 0 (blue) to 0.2 (red) and 0.4 (green), the cumulative total cost actually decreases due to SS helps dampen FS and PS oscillations, and reduces problem solving delay costs. As SS sensitivity increases to 0.6 (grey) and 0.8 (black), the cumulative total cost increases dramatically due to too high the shift the burden tips the system into a vicious circle. Further Work • Other shocks: step, sine wave, noise etc • Add and analyze proactive FS buildup policy (a fixed training) in combination with existing reactive SS and FS buildup policies. • Add and analyze subjective performance measure (Gibbons), a goal-and-gap structure for FS. • Analytic analysis beyond simulation runs Questions • Publishable? Where? How to frame? • Real world cases resemble to the described insights?