A Model of Opinion Dynamics Among Firms

1
A Model of Opinion Dynamics Among Firms
Potentials of Complexity Science for Business,
Governments, and the Media Collegium Budapest,
03-05.08.2006

• Volker Barth,
• GELENA research group,
• Oldenburg University

2
Outline

• Motivation/Background
• Basic model
• Opinion update based on external attribute
• Extended model
• Opinion update based on multiple attributes
• Conclusions

3
Motivation

• Model the opinion spread in agent networks
• Drivers, mechanisms,?
• Empirical case studies on learning of
  organisations with respect to sustainability
  (GELENA project)
• Imitation and copying (Consulting)
• Existence of networks
• Leader-follower dynamics within organisations
• Change agents, Time lags

4
Motivation

• Existing studies
• Binary opinions
• Continuous opinions
• Opinion exchange often based on thresholds
  (bounded confidence model)

Source Weisbuch et al., 2002
5
Basic Model (1) Agents

• Individual opinion (oi)

• Personal attribute (authenticity, credibility)
  ai 0.1, 0.2,1.0, Poisson
  distributed
• One meeting per time step
• Update opinion during meetingoB(t1) oB(t)
  (oA(t) - oB(t)) aA g, for aA gt aB

6
Basic Model (2) Social Networks

• Agents localised on lattice (torus)
• Social connections, no (direct) spatial meaning
• Limited environments (vonNeumann, Moore)
• Meet only agents within environment

s
7
Agent Interaction Sample Result

• 49 agents, 5000 meetings, Von Neumann, range 2

8
Agent Interaction Network effect

• 49 agents, 5000 meetings, average over 30 runs

9
Basic Model Summary

• Agent opinions fluctuate between visible opinion
  leaders
• Greater visibility ? smaller opinion range
• Local opinion leaders
• Eccentric, but unchanged opinionsno impact on
  mainstream
• Extend opinion range

10
Extended Model Firm Interaction

• Companies
• Firm opinion (follows CEO opinion)
• Firm attribute economic performance
• Agents now company executives
• Implicit Firm opinion and performance connected
• Executives update opinion upon
• average firm opinion of better-performing firms
  oi oi(t) (ltoF,e(t)gt - oi(t) ) d
• meetings with other executives

11
Firm Interaction Sample Results

• 49 executives, 5000 meetings, von Neumann (s1)

12
Firm Influence Parameter d

• 49 executives, 5000 meetings, avg. over 30 runs

13
Firm Interaction Sample Results 2

• 49 executives, 5000 meetings, von Neumann (s1)

14
Extended Model Summary

• Average firm opinion is attractor
• Convergence (Consensus) of mainstream depends
  on interaction speed
• Local opinion leaders still exist
• Small effects on mainstream in the short run
  (fuzzy fringe)
• Long-term attractor in the (very) long run ?
  Global opinion leader

15
Conclusions

• External attributes reflect actual opinion
  formation of humans
• Narrow consensus requires credible and widely
  known proponents ? otherwise localised opinions,
  wider range
• Orientation on uniform opinions leads to narrow
  consensus ? rules, values
• Fixed, unbalanced opinions can shift mainstreams
  in the long run ? Extremism

16

• Thank you!

17
Summary 1 Model

• Agents
• Individual property (authenticity) determines
  individual opinion updating behaviour
• Social network determines opinion exchange
• Companies
• Individual property (performance) important for
  perception by other executives
• Firm opinion follows executives opinion
• Executives opinion individual exchange
  superposed by averaged firm opinion

18
Summary 2 Findings

• Consensus as wide mainstreams depending on
• agents range of sight
• strength of firm opinion consideration
• Local opinion leaders
• primarily isolated eccentrics
• slight but ongoing impact on mainstream, when
  multiple criteria matter

19
Questions

• Can observed differences between sectors be
  understood by different
• dynamics of opinion formation of executives?
• dynamics of dissemination into organisations
  (i.e. firms) ltgt transaction costs ?
• network structures?
• Role of external influences?
• Consulting
• (Media)

20
Visibility of firms
21
Basic Model (2) Social Networks

• Agents localised on lattice
• Social connections, no (direct) spatial meaning
• Meet only known agents (visibility)
• Limited views possible (Moore, vonNeumann)

22
Agent Interaction Network effect 1

• 49 agents, 5000 meetings, average over 30 runs

23
Agent Interaction Network effect 2

• 49 agents, 5000 meetings, average over 30 runs

24
Firm Influence Parameter d

• 49 executives, 5000 meetings, avg. over 30 runs