modelling of marketing and business systems: agent -based
TRANSCRIPT
Modelling of Marketing and Business Systems: Agent-Based Simulation
Louise YoungSchool of Marketing, University of Western Sydney
Ian Wilkinson & Fabian HeldDiscipline of Marketing , The University of Sydney Business School
Bob Marks
University of Melbourne
Terry Bossomaier
Charles Sturt University
What academics want
• New focal phenomena –dynamics not (comparative) statics
• New contexts–ROW not USA (but continuing use of TLA)
• New theories–Mechanisms and process not variables
• New methodologies–Generative social science not surveys
Two ways of doing science
“Collect observational, survey or other forms of data and analyse them, possibly by estimating a model;
ORbegin from a theoretical understanding of certain
social behaviour, build a model of it, and then simulate its dynamics to gain a better understanding
of the complexity of a seemingly simple social system”
[Introduction, Nigel Gilbert Agent Based ModellingQuantitative Methods in the Social Sciences 2008]
Artificial Network Life
• ABM allow us to synthesise new forms of behaviour and organisation that have not existed in real life
• “[Synthesis] extends the empirical database upon which the theory of the discipline is built beyond the often highly accidental set of entities that nature happened to leave around for us to study” [Chris Langton].
From statics to dynamics
• From comparative static, variables-based thinking and theories
To:• Dynamic, Process, Evolutionary and
Mechanism based thinking and theories• Problematising non-change not change
–Reproducing the past is tough• Generative Social Science (Epstein 2009)
Generative explanations
“To ’explain’ an empirical regularity is to discover a set of simple mechanisms that would produce the former in any system
governed by the later” (Herbert Simon, 1968)
THE DYNAMICS AND EVOLUTION OF BUSINESS NETWORKS
What are we trying to explain?
Example of Business Networks: Coca-Cola (simplified)Source: Press releases; McKinsey analysis
Coca-Cola
Metique
Offline Purchasing consortium
Suppliers of • Sugar
Sweetener• Coloring• Aluminum
cans• Plastic bottles• Etc.
Wal-Mart
Albertson's
Private exchange for customer data
Data
Sales data
Worldwide Retail Exchange
AOL
Transora
Direct and indirect goods
Inventory and forecast information
Procter and Gamble
NewCo("healthy" snacks and beverages)
Nestlé
Product alliance
Brand venture
Brand venture
Kraft
Direct indirect goods
Inventory and forecast data
Coca-Cola bottling company/ CCE
Distribution
Incumbent
GoodsInformation
Startup
Consortium
Logistics marketplace
GEPolymerland
Source: Science 2007
World’s Export Production - Use Network
A different way of viewing
Studying what countries make-context matters! (MIT)
What is studied here
• Production requires capabilities that are difficult to acquire/slow to accumulate, are required by some industries but not others,
• Therefore present and future capabilities are constrained by network of product similarities, i.e. products requiring similar inputs
• Case study of South Korea versus Chile– Physical and human capital of each was determines
by adjacency in product space– This adjacency determines income and speed/nature
of development
Two exemplar problems
1. Emergence of Intermediaries
2. Emergence of Central Markets
Emergence of Intermediaries: How do we get from A to B?
Buyers
Suppliers
Intermediary
Suppliers
Buyers
?
Emergence of Markets:How do we get from C to D?
Specifics? Evolution, Equillibration and Balance
• Through time systems are remade and/or evolve.
• The events of the environment precipitate events within the system. – they create synergies (Holland 1998) – bring unintended consequences (Norton 2002).
• Conceptual underpinnings of evolution can be considered in terms balance and equilibration.
Balance and Process
• “By balanced state (or situation) is meant a harmonious state, one in which the entities comprising the situation and the feelings about them fit together without stress” (Heider 1958, 180).
• Equilibration is the process by which systems retain or move towards balance.
Why balance?
• It is argued that the seeking of balance is a fundamental motivator in human psychology (Appley 1990)
• Allows us – to organise and integrate our thoughts and ideas (Festinger
1957), – To organise our thoughts and sentiments in relations with
others (Heider 1958), – is the basic means by which learning and knowledge
formation occurs (Feldman 1995). • Balance is so fundamental that humans are biologically
driven to seek it (Parkins 1990).
Moving to balance
• The process of moving towards a balanced state is equilibration.
• Equilibration has been widely considered in the social sciences
• Equilibration is also considered in the study of complex adaptive systems (eg. Prigogine 1990, Kauffman 1995, see Heylighen 1997 for an overview).
Equilibration is:
“not a simple balance of forces as in mechanics ….but in the sense of self-regulation; that is, a series of active compensations in response to external disturbances and an adjustment that is both retroactive (loop systems or feedbacks) and anticipatory, consisting of a permanent system of compensations.” (Piaget 1976, p. 74)
EXISTING RESEARCH
General Theories and Models
• IMP Interaction model- AARS e.g. Hakansson et al 1982, Hakansson and Snehota 1997, Welch and Wilkinson 2002
• Stage models e.g. Ford 1980; Dwyer, Schurr & Oh 1987
• Evolution, events, process and flow theories e.g. Bairstow 2011, Easton and Lundgren 1992, Gadde & Håkansson 1992, Hakansson 1992, Hakansson and Lundgren 1997, Hakansson and Snehota 1997, Hibbert and Wilkinson 1991, 1994, Huang 2010, Wilkinson 1990, Young 2006
• Organisational change theories e.g. Aldrich 1999, Buttriss and Wilkinson 2005, Van de Ven and Engleman 2004
• Patterns of change theories e.g.Hollander (Wheel and Accordians), Gadde and Mattsson 1987, Johanson and Hakansson 1991, Hakansson and Snehota 1997
• NK Modelse.g. Easton and Wilkinson 1996, 1997, Easton et al 1997, 2007, Wiley et al 2001
Research Methods
• Case studies and descriptive historiese.g. Ford & Redwood 2005, Fu et al 1999, Hertz 1998, Kinch 1993, Narayandas and Rangan 2004, Roy and Wilkinson 2004
• Analyses of specific mechanisms and processes e.g. Andersson and Mattsson 2009, Bairstow 2010, Bairstow and Young 2011, Easton and Araujo 1994, Halinen and
Törnroos 1995, Halinen 1998, Huang and Wilkinson 2006, Huang 2010, Medlin 2002, Wilkinson and Young 2004
• Computer simulations
– Early attempts (systems dynamics, fixed structure)e.g. Forrestor 1961, Balderston and Hoggatt 1962, Bowersox et al 1972, Wilkinson 1986
– Recent resurgence (evolving structure and dynamics)e.g. Debenham and Wilkinson 2006, Easton and Wilkinson 1997, Easton et al 1999, Easton et al 2007, Følgesvold
and Prenkert 2009, Ladley et al 2007, Rand and Rust 2011, Wilkinson and Wiley 2001
MODELING DYNAMICS AND MECHANISMS
OUR Focus: Generative explanation
Defining Mechanisms - the Verbs of Explanation
Cam
pbel
l (20
05)
the processes that account for causal relationships among variables
Hed
strö
m(2
005)
a constellation of entities and activities that are organized such that they regularly bring about a particular type of outcome
Elst
er(1
989)
the nuts, bolts, cogs, and wheels that link causes with effects
Types of Mechanisms in Business Relations and Networks
Mechanisms of specializatione.g. Acting, producing, consuming, buying, selling, learning, specialising, in-sourcing, outsourcing
Business mating mechanisms e.g. Finding, being found, attracting, repelling, choosing, being chosen
Business dancing mechanismse.g. Interacting, exchanging, responding, trusting, committing, learning and adapting in relations
Mechanisms connecting relationse.g. Enabling and constraining effects of other relations, comparing, accessing, prioritising
Environmental mechanismse.g. Enabling and constraining effects of environment and starting conditions
BUSINESS RELATIONS AND NETWORKS AS COMPLEX ADAPTIVE SYSTEMS
Complex Adaptive Systems
Locally Interacting Agents
AGENT-BASED MODELLING
ABM complements and extends existing research methods not replaces them
• Systematic case histories to identify processes and mechanisms–Qualitative historical case studies
• Validating and testing ABM against cases studies and emprical results (even SEMs!)–But agents in a model can give 100%
response rates and don’t lie!!–Need mechanisms of non response and self
serving biases in our ABM too!
Systematic Case Histories
Source: Huang 2011
Source: Huang 2011
Source: Huang 2011
Time is represented in different ways (this also emerges from using Ethno2)
Event analysis: Australian IT
Process, mechanism and evolution
The Meaning of the Lines
• How do we get from situation A (at time t) to Situation B at time t+n?
• Identifying mechanisms and processes
• Not just about where but the journey (equillibration) and the “reasons” for it.
• Attribution (direct and indirect) of reasons
Methods for identifying mechanisms in cases (methods for study of discourse)
What Business Network Mechanisms have been previously modeled?
In sum – Lots!
Computer models exist for each category of mechanisms
Many simulations use search / mating mechanisms
Strategy and strategy change mostly studied through evolutionary iterated game theory
Few examples of specializing and dividing up of labour in simulations
Limited modelling of relationship evolution processes except for evolution of cooperation and trust
BUILDING AGENT-BASED MODELS OF RELATIONS AND NETWORKS
Modules for Simulation Models ofDynamic Business Relations and Networks
Acting and specializing
Dancing
Mating
Interconnecting
Environmental impacts
Step 1: Specialization- Learning and Exchange
Acting and specializing:
• Heterogenous capabilities
• Production• Resource allocation• Specialization
through learning
Dancing:• n/a
Mating: • Exchange with random
partners
Connecting:• Scarcity of resources
Environment:• Neighborhoods• Distances• Random distribution
Simulating the Effects of Specialization and Exchange
Observed Statistics of Simulation Development
• Continuously monitoring– Deficits in supply– Progress of specialization– Energy and successful agents– Average number of potential exchange partners
Step 2: Relations Change Over Time- Stickiness
Acting and specializing:
• Homogeneous
Dancing:• Stickiness
(Age reduces chanceof switching)
Mating: • Random• Preferential
attachment• Friends-of-friends
Connecting:• Implied in Friends-of-
friends mating
Environment:• Initial network
structures differ
Simulating the Effects of Developments in a Relationship
Observed Statistics of Network Structure Development
Degree Distribution• Histogram of the
current numbers of relations for all agents
• Distribution of relationship ages
• Size of giant component
• Clustering coefficient
What stops more researchers doing this type of research?
• Ignorance and resistance (1/3, 1/3, 1/3?)
• Lack of training in high level programming–We teach maths and stats but not
programming
–Programming is different
–Programming requires algorithmic thinking (i.e. mechanisms and processes)
What can we do about it?
1. Build more user friendly programming languages
2. Provide more training in programming in schools and undergrad and postgrad programs
3. Build Flight Simulators for ABM design, development and testing
BNAS: Business Network ABM SystemA Flight Simulator for Researchers, Educators, Managers and Policymakers
Activities and specialization:
• Allocation• Production • Exchange• Learning
Dancing:• Improvements in
relationship efficiency
Mating: • Various search
mechanisms• Reputation and
expectations
Connecting:• Comparison of
alternatives and optimization
Environment:• Cost functions• Heterogeneity• Spacial distance