Systems Thinking: A Comprehensive Approach to Complex Socio-Technical Systems, Lecture notes of Computer Science

Download Systems Thinking: A Comprehensive Approach to Complex Socio-Technical Systems and more Lecture notes Computer Science in PDF only on Docsity!

Systems Thinking

MS-

Revision

Course Outline

• Systems Concepts
  • Introduction to systems thinking
  • Systems thinking stories
  • Illities
  • Special focus on Robustness and Flexibility
  • Emergent Systems and Swarm intelligence
• Systems Modelling and Analysis
  • Systems Dynamics
  • Networks
  • Simulation
  • Decision-making in complex systems
  • Multi Objective Optimization
  • Data analysis
  • Experimentation

System thinking stories

• Some class stories:
  • Cricket versus basketball metrics
  • Toilets (footfalls as a proxy for measuring cleanliness, safety, usability, etc.)
  • Fixing congestion in a road
  • Waiting for the elevator
  • Deaths from safety road regulations. No free right turn.
  • Minimum wage (Video)
  • The electric car
  • The restaurant that looked at its trash
  • Sendhil Mullainathan’s ted talk: (Video)
  • Puzzle of Motivation (Dan Pink): (Video)

System thinking stories

• What is common in these stories:
  • Needs to be about metrics (modelling effort), or methods of analysis, solving a problem, or achieving an objective
  • Have many interconnected parts, ideally with lots of interactions
  • Improvement in any one part need not necessarily mean improvement to the whole system. Individual performance is not system performance
  • Metrics/results/behaviour associated with individual parts need not be very representative of the whole system
  • It might not make any sense to modularize the system into parts. It might be suboptimal to break down the system into parts

Illities

• The old illities:
  • Quality
  • Safety
  • Usability
  • Reliability

The new illities:

• Flexibility
  • Reconfigurability
  • Redesign
    • Evolvability (ability to change its core DNA, the system’s very purpose)
    • Adaptability (ability to change in response to external stimuli)
    • Modularity (ability to be modular and therefore facilitate plug and play)
    • Agility (ability to change quickly)
    • Scalability (ability to grow to accommodate more volume)
    • Extensibility (ability to grow to accommodate new functions)
• Interoperability – closely linked to compatibility and modularity
• Resilience
• Reliability
  • Maintainability
  • Availability
• Durability
• Robustness
• Sustainability
• Testability

illities

• Focus on difference between operability and usability
• Focus on the nuances of different types of Flexibility through redesign
• Focus on difference between Reliability, durability and robustness
• De Weck, Olivier L., Daniel Roos, and Christopher L.

Magee. Engineering systems: meeting human needs in a complex

technological world. MIT Press, 2011.

• Great reference:
• http://strategic.mit.edu/docs/es_book_004_proof.pdf

Illities special focus: flexibility

Emergence

• The phenomena where larger entities, patterns, regularities arise

through interactions among smaller and simpler entities that

themselves do not exhibit such properties

• The idea that in many complex environments, structures, behaviours

and properties seem to emerge even though there is no blueprint for

individual agents to follow

• Individual agents self-organizing in an environment to form complex

behaviour

Emergence continued

• Inspiring various management/academic approaches:
  • Genetic algorithms
  • simulated annealing
  • ant crawl optimization
  • Ensemble methods
  • Delphi Method
  • Prediction markets
  • Agent based modelling and problem solving
• Reference:
• Miller, Peter. "Swarm theory." National Geographic 212.1 (2007): 127
• Surowiecki, James. The wisdom of crowds. Random House LLC, 2005.

System dynamics

• Properties of a system the Systems Dynamics tries to model
  • Systems are dynamic not static. To understand them we need to model them over time not at one snapshot.
  • They are primarily governed by feedback mechanisms. Shower example.
  • They are typically non-linear
  • History dependent (A particular decision could have a permanent impact on the system for a long time to come)
  • Adaptive (decisions and behaviour are context specific and change over time)
  • Counterintuitive (cause and effect are too far apart in time and space, simple solutions rarely have the intended effect (minimum wage example))
  • Trade-offs between long run and short run response (time delays, nonlinearity, etc.)

Networks

• Network theory basics
• Understanding of node (vertices) connected by edges (links or arcs)
• Adjacency matrix and incidence matrix
• Node Metrics: Degree of a node, clustering co-efficient of a node,

Betweeness of a node,

• Graph metrics: K-regular graph, Distance, Diameter, Degree distributions,

Degree correlations, and Network resilience

• References: (Look at the presentations for lectures 7 and 8)
• http://ocw.mit.edu/courses/engineering-systems-division/esd-00-

introduction-to-engineering-systems-spring-2011/lecture-notes/

Simulation

• Introduction: What is it? Why and where is it useful?
  • Advantages, Disadvantages and differences between
    1. Experimentation
    2. Modelling
       1. Mathematical modelling
       2. Simulation modelling
• Where do we simulate
  • Systems Dynamic models
  • Agent Based Modelling (important to understand the difference between systems dynamics and agent based modelling)
  • Anywhere there is stochasticity
  • Even when there is no stochasticity
• References: Excel sheet shared with the class

Optimisation, data analysis and

experimentation

Experimentation

• Create the data
• Build a model from data
• Optimize the model to find the optimal inputs

Data Analysis

• Supervised learning
• Mostly converting data to model

Optimisation

• Given a model what are the inputs that give me the best output

Optimisation, data analysis and

experimentation

• Multiobjective optimisation: The concepts of pareto/efficient frontiers (Dominated versus Non-dominated solutions)
• Data analysis: The problem of multicollinearity (correlated input variables in a regression problem). Correlation versus causation the ice-cream story
• Experimentation: Formal experimentation involves systematic, purposeful changes to input variables in an attempt to gain knowledge about the system and/or find the ideal settings that result in the best output.
• Experimentation: The only real form of causal research

http://en.wikipedia.org/wiki/Design_of_experiments

• Esther du Flo talk: http://www.ted.com/talks/esther_duflo_social_experiments_to_fight_poverty?la nguage=en