Managing Complexity ?                     Stefan Thurnerwww.complex-systems.meduniwien.ac.at            www.santafe.edu   ...
To manage a system means that you are able to predict the possibleoutcomes of actions you takeIf you can not predict the o...
What kind of systems can you predict ?• Very small ones: physics of a few bodies: planet-sun, quark-quark,rocket-enemywhy ...
What are Complex Systems ?• CS are made from many elements• Elements are strongly interacting with each other• CS depend o...
Why is it so hard to predict complex systems?• Mathematical and statistical tools break down• Too big to handle• Not enoug...
Typical Complex Systems• Living systems• Social systems• Economic/financial systems• Firms + organisationsall these systems...
CS are co-evolving multiplex networks• States of individuals/firms change as a function of NW interaction• Network changes ...
The game changer: The computer + new math• Network theory: quantify causal processes on networks• Data availability: elect...
Networks• Network: connections of points by lines• Usually NW has structure + random components• Any data that can be stor...
Networks are dynamical: here sits the devil !• Networks change over time• Dynamical processes happen on networks• Dynamica...
Peter Drucker Forum, Vienna Feb 20, 2013 10
Example IWhat if we know everything? What can we predict?• then we have solved the data problem → good!• only problem: wha...
Massive Multiplayer Online Game – pardus.ata toy for the new data generationComplete knowledge on a human societyHave all ...
The properties of the subnetworks                                            Positive ties                     Negative ti...
Network–network interactions                                                                                              ...
How humans organize to stabilize their social NWs                                 +        +   +          +   +        -  ...
Gender differences in ’networking’                       Peter Drucker Forum, Vienna Feb 20, 2013 16
Predict: Where will you go next ?     (a)    Adjacency A            (b)   Distance D           (c)   Transition count M   ...
Alternative representation of humans – behavioral codeActions every (human) player can engage in life (game)C ... communic...
Action-streams→ Analyze the code as if it was the genomePLoS ONE (2012)                                             Peter ...
Predict: What is your next action ?                                (a)                          X                         ...
Example IIComplexity in financial networks and systemic-riskmanagement• If you lend something to somebody: want to know cre...
Interbank credit NW of AustriaAsset-liability NW of all Austrian banks, October 2003Quantitative Finance (2005)           ...
All financial flows between Austrian banks                                                                                  ...
Financial networks – management of systemic-risk• With this information → define a systemic risk measure: DebtRank• Can reg...
Example IIIWhat is complex in a firm or corporation ?• Corporation is an evolutionary network of communication-flows• Manage...
Why CS in corporations?Imagine new CEO entering a firm with say more than 1000 employees• Knows organigram. Knows what depa...
Communication: the nervous system of a firm• Telephone bill: every call = one lineID caller | dpt caller |ID called | dpt c...
emails of corporation, december 2011                                                                     Pajek• what do we...
emails: max. spanning tree – departments, dec 2011                                                                        ...
Destroy the core and you collapse !• Loss of core often leads to collapse of network                                      ...
Conclusion• Complexity arises through inter-connectedness: networks• Science: complexity catastrophe possible: quick & dra...
Conclusion II message for strategic management•   Communication-network monitoring–   know your corporation–   restructure...
Upcoming SlideShare
Loading in...5
×

Stefan Thurner at Drucker Forum Launch Event, Feb 20, 2013 "Complexity"

460

Published on

Presentation by Stephan Thurner, Professor at Santa Fe Institute and the Vienna Medical University on Complexity Science and its potential value for management.

Published in: Business
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
460
On Slideshare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
13
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Stefan Thurner at Drucker Forum Launch Event, Feb 20, 2013 "Complexity"

  1. 1. Managing Complexity ? Stefan Thurnerwww.complex-systems.meduniwien.ac.at www.santafe.edu Peter Drucker Forum, Vienna Feb 20, 2013
  2. 2. To manage a system means that you are able to predict the possibleoutcomes of actions you takeIf you can not predict the outcome of management actions - you do notmanage or control the system: you are subject to luck, fate, external events Peter Drucker Forum, Vienna Feb 20, 2013 1
  3. 3. What kind of systems can you predict ?• Very small ones: physics of a few bodies: planet-sun, quark-quark,rocket-enemywhy ? can manage to handle laws of physics• Very large ones: gases, solid bodieswhy ? do not interact strongly → describe laws of physics with laws ofstatistics• Managing CS has been out of scope since beginning of mankindwhy ? too large to handle governing ’laws’, too small for statistics to work,too interconnected Peter Drucker Forum, Vienna Feb 20, 2013 2
  4. 4. What are Complex Systems ?• CS are made from many elements• Elements are strongly interacting with each other• CS depend on details of the system: who interacts with whom• CS often pose chicken-egg problems→ So far: very hard to handle in predictive ways Peter Drucker Forum, Vienna Feb 20, 2013 3
  5. 5. Why is it so hard to predict complex systems?• Mathematical and statistical tools break down• Too big to handle• Not enough data available• No concept of the backbone of CS – networks Peter Drucker Forum, Vienna Feb 20, 2013 4
  6. 6. Typical Complex Systems• Living systems• Social systems• Economic/financial systems• Firms + organisationsall these systems are evolutionary and are based on networks Peter Drucker Forum, Vienna Feb 20, 2013 5
  7. 7. CS are co-evolving multiplex networks• States of individuals/firms change as a function of NW interaction• Network changes as a function of the states of the agentsfirst point only: physicssecond point: makes it a complex system: society, eco-system, market, ... Peter Drucker Forum, Vienna Feb 20, 2013 6
  8. 8. The game changer: The computer + new math• Network theory: quantify causal processes on networks• Data availability: electronic fingerprints everywhere• New statistics: new inference methods and superstatistics• Storage and computation costs practically nothingIf you do not have all of the above → back to start: can NOT manage CSIf you can map a CS into a dynamical NW → can manage CSTheory of CS: combination of dynamical systems and NWs Peter Drucker Forum, Vienna Feb 20, 2013 7
  9. 9. Networks• Network: connections of points by lines• Usually NW has structure + random components• Any data that can be stored in database can be represented as network• Network theory: quantify stability, efficiency, hierarchy, clustering, ...• Novel mathematics: networks become technically manageable Peter Drucker Forum, Vienna Feb 20, 2013 8
  10. 10. Networks are dynamical: here sits the devil !• Networks change over time• Dynamical processes happen on networks• Dynamical processes happen on networks which change (chicken–egg)• Networks are not independent: they influence each other → Multiplex Peter Drucker Forum, Vienna Feb 20, 2013 9
  11. 11. Peter Drucker Forum, Vienna Feb 20, 2013 10
  12. 12. Example IWhat if we know everything? What can we predict?• then we have solved the data problem → good!• only problem: what to do with all that information? Peter Drucker Forum, Vienna Feb 20, 2013 11
  13. 13. Massive Multiplayer Online Game – pardus.ata toy for the new data generationComplete knowledge on a human societyHave all data of all 500.000 players of all of their actions and interactionsperformed in a ’second life’ like virtual world www.pardus.at• have complete knowledge of an entire human society• have time evolution of the co-evolving multiplex and states Peter Drucker Forum, Vienna Feb 20, 2013 12
  14. 14. The properties of the subnetworks Positive ties Negative ties 0 10 −1 kin −1.0 10 (a) (d) kout P(≥ k) −2 10 −3 10 Friendship Enmity −4 10 0 10 −1 −1.7 10 (b) (e) P(≥ k) −2 10 −3 10 Communication Attack −4 10 0 10 −1 −1.5 10 (c) (f) P(≥ k) −2 10 −3 10 Trade Bounty −4 10 0 1 2 3 0 1 2 3 10 10 10 10 10 10 10 10 k k• Positive links are highly reciprocal, negative links are not• Power-law degree distributions indicate aggressive actions• Positive links cluster Peter Drucker Forum, Vienna Feb 20, 2013 13
  15. 15. Network–network interactions Degree / Rank correlation Link overlap 0.2 Degree correlation 1 Degree rank correlation Link overlap 0.16 0.8 0.12 0.6 0.08 0.4 0.04 0.2 0 0 C:F T:C E:A C:A E:B A:B C:E T:F F:A T:A C:B T:E F:B F:E T:BHow does one network shape and influence the other?Proceedings of the National Academy of Science USA (2010) Peter Drucker Forum, Vienna Feb 20, 2013 14
  16. 16. How humans organize to stabilize their social NWs + + + + + - - - + - - - Strong formulation of balance B U B U Weak formulation of balance B U B B 26,329 4,428 39,519 8,032 10,608 30,145 28,545 9,009 71 -112 47 -5Social balance theory – test empirically with natural science quality Peter Drucker Forum, Vienna Feb 20, 2013 15
  17. 17. Gender differences in ’networking’ Peter Drucker Forum, Vienna Feb 20, 2013 16
  18. 18. Predict: Where will you go next ? (a) Adjacency A (b) Distance D (c) Transition count M 106 1 20 104 10 102 0 0 100 (d) (e) (f)Scientific Reports (2012) Peter Drucker Forum, Vienna Feb 20, 2013 17
  19. 19. Alternative representation of humans – behavioral codeActions every (human) player can engage in life (game)C ... communicateT ... tradeF ... establish a friendship linkX ... remove an enemy linkA ... attackB ... place a bounty on someoneD ... delete a friendship linkE ... establish an enemy link... Peter Drucker Forum, Vienna Feb 20, 2013 18
  20. 20. Action-streams→ Analyze the code as if it was the genomePLoS ONE (2012) Peter Drucker Forum, Vienna Feb 20, 2013 19
  21. 21. Predict: What is your next action ? (a) X r Tr Cr 0.7 F E r 0.6 r A Br r 0.5 Dr 0.4 t D B A 0.3 E F 0.2 C T 0.1 X Xr Tr Cr Fr Er Ar Br Dr D B A E F C T X t+1 (b) + 0.81 (−399) 0.26 (−437) 1.34 (714) 0.23 (−349) − 0.21 (−482) 14.67 (1875) 0.29 (−389) 0.99 (−0.3) t +r 1.34 (661) 0.24 (−432) 0.78 (−410) 0.35 (−295) −r 0.25 (−341) 0.87 (−19) 0.34 (−301) 19.93 (1837) + − +r −r t+1PLoS ONE (2012) Peter Drucker Forum, Vienna Feb 20, 2013 20
  22. 22. Example IIComplexity in financial networks and systemic-riskmanagement• If you lend something to somebody: want to know creditworthiness• In a network can not do that, unless know riskiness of EVERYBODY• No rational decision making on lending possible without transparency• Central Banks have almost complete information Peter Drucker Forum, Vienna Feb 20, 2013 21
  23. 23. Interbank credit NW of AustriaAsset-liability NW of all Austrian banks, October 2003Quantitative Finance (2005) Peter Drucker Forum, Vienna Feb 20, 2013 22
  24. 24. All financial flows between Austrian banks Pajekaggregated 2005-2006, 211 Banks, 4.187.943 transactions, volume 11.07 trillionEUREuropean Physics Journal B (2009) Peter Drucker Forum, Vienna Feb 20, 2013 23
  25. 25. Financial networks – management of systemic-risk• With this information → define a systemic risk measure: DebtRank• Can regulate the interbank system that it becomes systemically risk free• Idea: forbid borrowing from systemically risky banks 0.07 normal mode transparent mode 0.06 fast mode 0.05 frequency 0.04 0.03 0.02 0.01 (a) 0 0 50 100 150 total losses to banks (L)http://www.bloomberg.com/news/2013-02-10/fix-finance-by-shedding-light-on-its-complexities.html Peter Drucker Forum, Vienna Feb 20, 2013 24
  26. 26. Example IIIWhat is complex in a firm or corporation ?• Corporation is an evolutionary network of communication-flows• Management controls these flows and tries to re-arrange the NW• With and without management: the network evolves = re-arranges→ CS tools assist to:• know the network (reality check)• find ways to re-arrange it (optimize)• monitor its change after managerial interventionNote: it is not easy to tell a network how to re-arrange ! Peter Drucker Forum, Vienna Feb 20, 2013 25
  27. 27. Why CS in corporations?Imagine new CEO entering a firm with say more than 1000 employees• Knows organigram. Knows what departments should do + hierarchy• Does not know: what departments are in reality• Are departments structured in reality as the CEO thinks they are?• How hierarchical is the firm in reality?• Do re-structuring interventions work?• Do optimal solutions to organization structure exist, given real structure→ Answer to this: communication flows Peter Drucker Forum, Vienna Feb 20, 2013 26
  28. 28. Communication: the nervous system of a firm• Telephone bill: every call = one lineID caller | dpt caller |ID called | dpt called | time | duration10000 50 50404 7 16 10 2009 14 01 00 00 1410003 50 50091 2 04 11 2009 10 28 00 00 0410003 50 50032 2 04 11 2009 16 27 00 01 1010004 50 50404 7 30 09 2009 08 27 00 02 58.........• emails: log-files on server: every mail = one lineID Sender | dpt sender | ID recipient | dpt recipient | kB | time sent100001 1 508193 50 9486 2009 11 03 07 13 15100001 1 508193 50 3444 2009 11 03 07 13 15100001 1 509485 50 2996 2009 11 03 13 27 11100001 1 509485 50 3122 2009 11 03 13 27 11......... Peter Drucker Forum, Vienna Feb 20, 2013 27
  29. 29. emails of corporation, december 2011 Pajek• what do we see? nothing ! → network theorywith kind permission from G¨nther Weiss u Peter Drucker Forum, Vienna Feb 20, 2013 28
  30. 30. emails: max. spanning tree – departments, dec 2011 177171 350 342 176 32 101 44 208 124 142 321 362 102 303 125 135 310 100 360 106 131 22 126 345 359 18 272 160 27 90 92 265 306 318 86 183 120 151 122 123 24 16 69 209 31 87 93 84 211 172 30 89 88 296 78 85 59 158 82 72 264 83 253 364 71 127 312 219 139 134 185 293 221 361 218 184 25 57 195 313 109 216 56 215 234 232 239238 227 235 338 237 231 230 132 337 228 229 339 128 305 130 168 217 189 223 300 27639 2 42 377 225 97 98 292 47 336 304 368 37 96 62 376 375 320 10 236 289 263 291 95 19 268 46 202204 68 6 7 45 33 58 5 220 275 20 250 65 107 8 233 328 52 53 199 308 203 61 334 314 35 366 43 222 50 64 94 104 341 226 374 369 55 54 11 3 1 4 196 192 224 188 243 240 242 66 145 257 169 271 259 213 324 279 210 258 9 105 307 51 34 181 335 269 173 80 252 12 167 206 319 48 248 198 15 137 365 323 302 38 70 74 255 280 212 290 367 81 166 281 343 207 161 325 75 13 249 286 191 79 256 155 267246 17 29 118 76 241 147 162 331 283 352 186 200 197 309 67 333 190 77 357 143 349 363 261 288 205 112 193 40 119 282 150 14 36 372 164 28 330 254 332 244 187 133 348 182 149 156 146 214 274 329 340 262 294 174 194 301 311 326 347 179 355 154 121 370 26 23 373 247245 315 49 327 201 298 152 63 175 41 153 129 144159 99 165 353 287 110 180 163 103 140 113 371 270 178 60 354378 260 277 117 295 317 136 138 21 148356 316 284 285 266 170 141 299 278 108 115 157 358 322 297 351 34673 91 273 114 116 111 251 344 Pajek• Support to questions: what is state of the firm – is restructuring needed?• Degree of clustering → rethink dept. structure? monitor restructuring?with kind permission from G¨nther Weiss u Peter Drucker Forum, Vienna Feb 20, 2013 29
  31. 31. Destroy the core and you collapse !• Loss of core often leads to collapse of network 252 301 261 213 365 123 133 121 170 86 84 298 153 151 129 349 306 212 139 136 141 38 268 57 156 9092164 211 81 83 142 128 147 303 266 280 210 89 27 120 160 122 99 125 98 140 127 152 124 183 102 347 180 103 93 76 264 80 75 162 105 265 95 256 234 325 29 119 13222 88 71 87 77 70 78 131 135 134148 97 47 67 106 179 62 82 44 31 61 79 85 100 68 104 208 155 144 149 63 28 14374 19 154 59 137 172 130 343 115 158 91 146 60 163 126 32 228 96 176 341 150 159 30 72 11894 173 35 161 182 207 206 316 73 241 145 198 272 58 167 230 174 171 113 66 283 315 317 69 14 190 177 339 188 229 227 305 373 114 64 18 16 26 108 243 21 203 15 48 192 201 186 240 112 17 200 34 233232 237 236 238 235 239 42 307 249 194 41 111 109 204 168 189 45 302 197 199 270 2325 246165 248 20 110 65 224 39 287 344 166 205 214 187 49 337 193 191 184 313 335 289 276 5169 9 3 51 175 222 218 196 195 216 309 6 46 271 269 24 12 4 22154 223 215 231 312 55 8 181 217 11 185 220 219 202 36 294 10 374 314 71 244 50 53 52 43 2 304 40 320 56 262 242 37 245 377 13 33 0 288 285 138 157 286284 281 178 282 279 278 275 226 209 225 273 263 259 257255 253 251 247 277 274 267 260 258 254 250 Pajekwith kind permission from G¨nther Weiss u Peter Drucker Forum, Vienna Feb 20, 2013 30
  32. 32. Conclusion• Complexity arises through inter-connectedness: networks• Science: complexity catastrophe possible: quick & drastic changes• If can map CS to dynamical co-evolving NW → can manage it• Can manage systemic-risk (risk of drastic change - collapse) Peter Drucker Forum, Vienna Feb 20, 2013 31
  33. 33. Conclusion II message for strategic management• Communication-network monitoring– know your corporation– restructure departments– monitor restructuring events• Core-detection of a corporation– verify keystone-emlpoyees– identify potentials easy to realize: AC cycle analysis– identify cores• Problems with these techniques:– Legal issues, privacy rights, anonnymization, and data storage Peter Drucker Forum, Vienna Feb 20, 2013 32
  1. Gostou de algum slide específico?

    Recortar slides é uma maneira fácil de colecionar informações para acessar mais tarde.

×