Idea propagation in organizations. Christopher A White June 10, 2009

Transcription

1 Idea propagation in organizations Christopher A White June 10, 2009 All Rights Reserved Alcatel-Lucent 2008

2 Why Ideas? Ideas are the raw material, and crucial starting point necessary for generating and introducing disruptive innovations to the market. 2 White 10 June 2009

3 Idea capture Although idea generation may be easy, the creation of disruptive ideas which can emerge as disruptive innovations is very difficult. Idea Generation Disruptive Innovation Infeasible Wrong Team Inadequate Resources IP Issues Late To Market 3 White 10 June 2009

4 Why Idea propagation? Ideas enter an organization from many sources. We seek to model the organizational structure and processes which encourage (discourage) idea propagation. Seminars Workshops Criticism Organizational Knowledge Meetings Conferences 4 White 10 June 2009

5 Simulation goals We strive to develop a quantitative model of the impact of organizational structures, social networks and processes on the propagation and capture of ideas. 5 White 10 June 2009

6 Bell Labs global presence Bell Labs has a global footprint, yet functions as one organization. Ideas originating in one region must be reinforced through exposure and introductions to other regions. CANADA USA UK IRELAND GERMANY FRANCE BELGIUM CHINA INDIA 6 White 10 June 2009

7 Bell Labs global partnerships In addition to the global Bell Labs locations, we have a variety of academic partnerships around the world. Bell labs aspires to bring together problems and solutions around the world. 7 White 10 June 2009

8 Bell Labs CSO Partnering Efforts The Chief Scientist s Office is responsible for academic research partnerships, long term research strategy, and for fostering technical excellence through both internal and external activities. Princeton Supelec INRIA Portugal Egypt Korea Singapore Australia New Zealand 8 White 10 June 2009

9 Outline Hierarchical Simulation tools and examples The GYRUS Simulator for idea propagation The Bell Lab s Grand Challenge Challenge Future work 9 White 10 June 2009

10 Hierarchical Simulation tools Bell Labs has a long history of utilizing modeling and simulation to understand complicated systems. Recently, we have applied similar simulation strategies to several very different applications. Optical Networks IPTV Delivery Idea propagation Lambda Xtreme TPSDA Grand Challenge 10 White 10 June 2009

11 6 orders of magnitude 11 White 10 June 2009

12 A Transparent Optical Mesh (ATOM) Simulator Topology Traffic Physics 12 White 10 June 2009

13 Power Transients in Optical Mesh Networks One primary application of the ATOM tool was to understand power transients in optical networks. Here we trace the detailed time dependence arising from the loss of power originating at node Network 13 White 10 June 2009

14 Control Scheduling We predict both behaviour and the control algorithms which modify that behaviour. The ATOM tool has been used to develop and optimize both amplifier and network level control schemes. Unscheduled Control Scheduled Control 14 White 10 June 2009

15 Simulation Strategy Our primary computational entity (or Agent) takes inputs and maps a processed input to an output port. Arbitrary Processing Input Ports Output Ports 15 White 10 June 2009

16 Simulation Strategy Entities can be composed by cascading other entities enabling the creation of arbitrarily complex entities. An entity might be an individual or a group of individuals Input Ports Output Ports 16 White 10 June 2009

17 Simulation Strategy All entities contain one or more finite state machines to simulate processes. All finite state machines execute in parallel. Waiting Read Input Network Run Msg 10 ms Read Output 10 ms Compare with Target Non Match Target 0.2 db / 100 ms Adjust WSS Match Target Send Amp Finish msg 50 iterations 17 White 10 June 2009

18 Simple to Complex Complicated behavior arises from many independent entities independently performing simple tasks and exchanging messages. 18 White 10 June 2009

19 GYRUS: Idea propagation simulation tool Topology Waiting Ideas Attend Workshop 10 ms Attend Talk 10 ms Read Abstracts Reinforce Expose to Idea Knowledge Model Process 19 White 10 June 2009

20 Topology: Connections between people When compared to networks, it is a critical difference that an organization might have multiple effective topologies. World World World 20 White 10 June 2009

21 Process: Introduce and Reinforce Ideas Processes represent the events by which ideas get introduced and reinforced within an organization. Waiting Attend Workshop 10 ms Attend Talk 10 ms Read Abstracts Expose to Idea Reinforce Attend Seminars Attend Conferences Dept. Meetings Write Papers Review Papers Mentor Students 21 White 10 June 2009

22 Knowledge Model Entities retain a subset of all ideas experienced. Learned ideas can be passed to other entities. Idea Idea Learn Forget Propagate ideas Query 22 White 10 June 2009

23 Knowledge model vectors We have defined an abstract knowledge model based upon vectors in a high dimensional (N > 10000) space. Similar ideas point in the same direction, different ideas point in an orthogonal direction. Different Idea Similar Ideas New Idea Residual Stored Idea 23 White 10 June 2009

24 Learning We have defined an abstract learning model based upon vectors in a high dimensional space. Similar ideas point in the same direction, different ideas point in an orthogonal direction. Expose to an Idea Reinforce Compute Residual Store? idea R Lnew L i idea α = α i i + L i idea R = L i i idea L i L = new R R R 24 White 10 June 2009

25 Bell Labs publication release database We counted the frequency of words in the titles contained within the Bell Labs publication release database. Similar archives of the evolution of ideas exist. Data from Allison Walker 25 White 10 June 2009

26 Results for an Individual Simple test cases for a single individual connection the knowledge model to the real world. Forgetting Reinforce Random Reinforce Random Reinforce 26 White 10 June 2009

27 Mutual reinforcement of an idea This system contains two entities which mutually reinforce the injected idea. When the rate of reinforcement balances the rate of forgetting, an idea will persist. Ideas compete for attention of entities: more process isn t always positive. 27 White 10 June 2009

28 Topology Even simple topologies can lead to complex behavior. Here a simple idea injected at individual one propagates to large group. Dept Group Ind 28 White 10 June 2009

29 Organizational coupling constants It is possible to track the flow of ideas through an organization. Specifically, one can find the chance of two individuals exchanging an idea White 10 June 2009

30 Bell Labs Grand Challenge Challenge The Bell Labs Grand Challenge Challenge was a collaborative brainstorming exercise across the global Bell Labs research community. The goal was to select a single Bell Labs Grand Challenge. 30 White 10 June 2009

31 Grand Challenge Process During the execution, we documented and followed the propagation of ideas. The Grand Challenge Challenge represents an ideal data set for benchmarking our models. 31 White 10 June 2009

32 Grand Challenge Process What would happen if we introduced an additional team formation phase to the GCC process? 32 White 10 June 2009 Form Teams

33 Grand Challenge Challenge simulation What would happen if we introduced an additional team formation phase to the GCC process? 33 White 10 June 2009

34 Why Form teams? The formation of teams adds a new topological network. In addition, processes at the group level enable the reinforcement of ideas between individuals. 34 White 10 June 2009

35 10 ms 10 ms Ongoing / Future work Knowledge model extension More sophisticated representations Non-linear models Feedback based models Organizational optimization Topology and Process Waiting Attend Workshop Attend Talk Read Abstracts Reinforce Expose to Idea Metrics Feedback mechanisms 35 White 10 June 2009

36 Innovation Modeling Market Feedback Resource s Ideas Invention Innovation 36 White 10 June 2009