Dirk Helbing (ETH Zurich)! Core Principles of an Information System to Anticipate Extreme Events!

Transcription

1 Dirk Helbing (ETH Zurich)! Core Principles of an Information System to Anticipate Extreme Events! Follow on!

2 Extreme Events!

3 We must realize that we are living in a complex, global, hyper-connected world.! New Problems: Systemic Instabilities! Examples:! 1. Climate change! 2. Global resource shortages (war, migration, terrorism)! 3. Financial, economic and debt crisis!!! 2010 Paresh Nath!

4 How the Interplay of Risk and Complexity! Creates Uncertainty!

5 Loss of Control through Cascade Effects! Mousetrap fission, by Gerhard G. Paulus, University of Jena,

6 Cascading Effects During Financial Crises! Video by Frank Schweitzer et al.!

7 Would Big Data fix the world?!

8

9 ! CHRIS ANDERSON S DREAM:!! The end of theory: the data deluge makes the scientific method obsolete.!! Wired Magazine 16(7), 2008!

10 Can we know everything?! A. Pentland et al.!

11 Can We Build A Crystal Ball?!

12 Would! so much information enable a wise king?!

13 The Magic Formula?! more! data! = more knowledge! = more power! = more success!

14

15

16

17 Data is the new oil Clive Humby!! but The models Digital are Revolution the new gold! Is on Its Way!! and participation means social capital!! Exponential growth of computing. 20th through 21st century.! Coutesy of Ray Kurzweil and Kurzweil Technologies, Inc. License: CC-BY 1.0.!

18

19 Classification Errors Will Happen! Pseudo-objectivity is a serious problem!

20 Maurits Cornelis Escher, thanks to Friedemann Mattern!

21 Remove layers! of deception!

22 Barack Obama s Warning! this is also a time around the world when some of the fundamental ideals of liberal democracies are under attack, and when notions of objectivity, and of a free press, and of facts, and of evidence are trying to be undermined. Or, in some cases, ignored entirely.! And in such a climate, it s not enough just to give people a megaphone. And that s why your power and your responsibility to dig and to question and to counter distortions and untruths is more important than ever.! Obama s Correspondents Dinner Speech (May 1, 2016)!!

23

24

25

26

27

28 Manipulation of the Choices of People by Personalized Information!

29 Limits of Conventional Big Data Analytics (here: Google Flu Trends)! POLICYFORUM BIG DATA The Parable of Google Flu: Traps in Big Data Analysis David Lazer, 1, 2 * Ryan Kennedy, 1, 3, 4 Gary King, 3 Alessandro Vespignani 5,6,3 I n February 2013, Google Flu Trends (GFT) made headlines but not for a reason that Google executives or the creators of the flu tracking system would have hoped. Nature reported that GFT was predicting more than double the proportion of doctor visits for influenza-like illness (ILI) than the Centers for Disease Control and Prevention (CDC), which bases its estimates on surveillance reports from laboratories across the United States ( 1, 2). This happened despite the fact that GFT was built to predict CDC Large errors in flu prediction were largely avoidable, which offers lessons for the use of big data. run ever since, with a few changes announced in October 2013 ( 10, 15). Although not widely reported until 2013, the new GFT has been persistently overestimating flu prevalence for a much longer time. GFT also missed by a very large margin in the flu season and has missed high for 100 out of 108 weeks starting with August 2011 (see the graph ). These errors are not randomly distributed. For example, last week s errors predict this week s errors (temporal auto- n March 21, 2014

30 See What Is Wrong! fmri to see diseases!

31 A Planetary Nervous System would enable tailored measurements, if well designed.! The Nervousnet project (nervousnet.info) has started to work on such a project. It is currently intended to run similarly to Linux, Wikipedia, OpenStreetMap or the World Wide Web consortium, i.e. it is open for international collaboration with various stakeholders.!

32

33 Each Smartphone Contains >10 Sensors! Today s Universal Tool!

34 We Can Connect Smartphones + Sensors! to Build a Global Measurement System! Photo: apple ios and android phones!

35 Most extreme events are rare, i.e. their probability of occurrence is small.! This makes it difficult to estimate the occurrence probabilities and expected damage sizes from data, while both the under- and over-estimation of extreme events is very costly for society.!

36 SENSITIVITY IMRE KONDOR

37 A single data point can dramatically change the predictions.! One cannot remove outliers from the data set. In a sense, all relevant data are outliers!!

38 Estimation Error over Number of Parameters! Source: Imre Kondor et al.

39 CHAOS

40 To reach an increased accuracy of estimations, biases in the data and information filters used must be reduced to a minimum. This disqualifies the current Big Data approach.! Representative data combined with suitable models and tailored measurements ("Smart Data") can outperform Big Data analytics.!!

41 Complexity of Epidemic Spreading! Dirk Brockmann and DH, Science (2013)!

42 Combining Little Data with A Model Reveals the Laws of Epidemic Spreading! Dirk Brockmann and DH, Science (2013)!

43 In complex systems, a pluralistic modeling approach (supporting a "collective intelligence" effect) is superior to the use of a single model, even if it is the best model on the market.! The requirement of pluralism calls for a sufficiently open and participatory approach. To get the "big picture", it is crucial to have sufficiently many and diverse perspectives on a complex problem.!

44 Multiple Perspectives Needed to Grasp Full Complexity MULTI-PERSPECTIVE APPROACH PLURALISTIC MODELING

45

46 Diversity Wins, Not the Best! Top-down and majority decisions obstruct collective intelligence!!! Wisdom of crowds requires independent exploration and then integration!

47 Citizen Science could play an important role. By means of a crowd-sourced approach, measurement data can be collected, curated, annotated, augmented and evaluated.!

48

49 Crowd Sensing, Citizen Science!

50 Visualization of Acceleration Data!

51 Detect Earthquakes and Warn Our Friends!

52 Reciprocity!

53 Reciprocity is a principle that allows to reach co-opetition (a combination of competition with cooperation) in an open, participatory system! With a suitable micropayment or incentive system, users can be motivated to fill measurement gaps as needed.!

54

55 Optimal Incentives: Minorities Report!! Richard Mann, DH (2016)

56 The reliability of information is key. Data sets and data sources must be assessed for reliability. This requires the use of suitable reputation systems.! Data sets should have multiple reputation values, data sources multiple qualification values.!

57 Reputation Systems to Increase Quality!

58 What is a signal and what is noise depends on the task to be solved. Therefore, reputation and qualification values are not absolute, but they depend on the problem to be addressed. This requires again a pluralistic filtering approach and, as a precondition, a pluralistic reputation system.!!!

59 Complementary to the reputation system, a qualification system (merit-based mechanisms) will be useful to determine who has access to what data (and how much), and who is entitled to perform certain kinds of operations or not.!!!

60 Qualification! Qualification System

61 Extreme events are often a result of cascading failures, which result from network interdependencies.! Some extreme events can be predicted by advance warning signs, such as unusually large variations of key variables, which may indicate a transition from stable to unstable system behavior.!

62

63 To assess the systemic risk related with network interdependencies, one should measure the vulnerability of the network nodes (system components) and the likelihood of transmission of a functional failure to neighboring nodes via network links (interactions).! Some of this can be done by means of computer simulations or interactive multi-player games in virtual worlds.!

64

65 Failure and transmission rates may be measured experimentally, and also by online monitoring and statistical analysis.! Nodes with a high in-degree are particularly vulnerable to cascade effects, nodes with high out-degrees are particularly likely to promote them. They deserve particular attention.! Based on failure and transmission rates, it is possible to estimate the probability distribution of possible damages, and the system components that are most frequently involved in cascade failures.!

66

67 New design principles for complex dynamical systems are at least as important as measurements and predictions. This requires knowledge from the areas of complexity science and mechanism design.! It is particularly important to design links in ways that will interrupt the transmission of failures and stop undesirable cascade effects as needed.! Distributed data storage and control principles can increase the security and resilience of the information system.!

68 Decentralization! Modular Design! Distributed Control! Subsidiarity Principle!

69 Exponential vs.factorial Growth Implications for the Governance of Complex Systems! Systemic! complexity! Loss of top-! down control:! Distributed! control! Big Data: Evidence-based! decision-making! Data! volume! Processing power! Not enough data to take good decisions!

70 Connecting the Dots!

71

72 The Digital Revolution on Its Way! : Thirty years war!!agricultural society ( anarchy )! : Ludwig XIV!!!feudal society (top-down regulation/! : Voltaire!!!!!! centralized control)! 1712, 1769 (James Watt): Invention of the steam engine! : Adam Smith!!!industrial society (bottom-up self-organization/! 1759: Theory of Moral Sentiments!!! distributed control)! 1776: Wealth of Nations! : French revolution! 1868: Public school established! : Max Weber!!!service society (administration/optimization)! 1905: on protestant work ethics and the spirit of capitalism! 1905, 1917: Russian revolutions! , : World Wars I+II! 1941: Zuse computer! 1989: World Wide Web! 1996: IBM Deep Blue!!!! 2004: Facebook!!!!digital society (social self-regulation/! 2010: Twitter revolutions, Arab Spring begins collective intelligence)! 2007-?: World financial, economic, and debt crisis, Ukrainian crisis, IS,!!

73 The Digital Revolution Is Transforming! Our Society!

74 Planetary Boundaries! W. Steffen et al. Science 347 (2015)! Our main problem is the lack of sustainability!

75 Circular Economy + Sharing Economy!

76 Finance 4.0:!! The need of a multidimensional incentive/ reward system!

77 Smart Technologies + Smart Citizens = Smarter Cities (Man Machine Symbiosis)! Image Source: