antonio a. casilli - networks, complexity, and privacy

66
Institut Mines-Télécom Networks, complexity, and privacy Antonio A. Casilli (Telecom ParisTech SES)

Post on 18-Oct-2014

883 views

Category:

Education


4 download

DESCRIPTION

ATHENS programme seminar by Antonio A. Casilli (Telecom ParisTech, Nov. 19, 2013).

TRANSCRIPT

Page 1: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom

Networks, complexity, and privacy

Antonio A. Casilli

(Telecom ParisTech SES)

Page 2: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/20232

Social networks

Télécom ParisTech

Page 3: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/20233

 

If I say «social network»

Télécom ParisTech

Page 4: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/20234

 

If I say «social network»

Télécom ParisTech

Page 5: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/20235

 

If I say «social network»

Télécom ParisTech

Page 6: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/20236

If I say «social network»

Télécom ParisTech

Page 7: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/20237

If I say «social network»

Télécom ParisTech

Page 8: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/20238

If I say «social network»

Télécom ParisTech

Page 9: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/20239

If I say «social network»

Télécom ParisTech

Page 10: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202310

 

If I say «social network»

Télécom ParisTech

Page 11: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202311

 

If I say «social network»

Télécom ParisTech

Page 12: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202312

Social network: a way of describing human groups as a set of social actors (nodes) and relationships existing among them (ties)

Human groups as networks

Télécom ParisTech

Page 13: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202313

 

Human groups as networks

Télécom ParisTech

Page 14: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202314

Human groups as networks

Télécom ParisTech

Page 15: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202315

Human groups as networks

Télécom ParisTech

Page 16: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202316

Human groups as networks

Télécom ParisTech

Page 17: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202317

Central Members

Bridges

Group Members

Isolate

Peripherals

Télécom ParisTech

Human groups as networks

Page 18: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202318

Is computer-mediated interaction changing the overall structure of human networks? Comparing computer-mediated and face-to-face

relationships: which networks are larger? Further refinements: are personal networks mainly

composed of "strong" or "weak" ties? Are there more weak ties in online personal networks?

Are personal networks densely knitted, or sparse? Are online personal networks sparser?

Computer-mediated interactions

Télécom ParisTech

Page 19: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202319

1992 Robin Dunbar

148

Télécom ParisTech

Computer-mediated interactions

Page 20: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202320

290

2000 Peter Killworth

Télécom ParisTech

Computer-mediated interactions

Page 21: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202321

610

2010 Matthew Salganik

Télécom ParisTech

Computer-mediated interactions

Page 22: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202322

2012

Télécom ParisTech

Page 23: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202323

1969: six degrees of separation

Télécom ParisTech

Computer-mediated interactions

Page 24: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202324

2012: four degrees of separation

Télécom ParisTech

Computer-mediated interactions

Page 25: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom

Two possible explanations

Higher transitivity of online networks

Presence of big hubs

25 Télécom ParisTech

Computer-mediated interactions

04/07/2023

Page 26: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202326

Different types of online «social capital»

Bonding : homogenous groups and cohesion Bridging : information circulating among heterogenous groups

Bonding

Bridging

Télécom ParisTech

Computer-mediated interactions

Page 27: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202327

From a “little boxes” society

Télécom ParisTech

Computer-mediated interactions

Page 28: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202328

…to “networked individualism”?

Télécom ParisTech

Computer-mediated interactions

Page 29: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202329 Télécom ParisTech

Computer-mediated interactions

Page 30: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202330

“Glocal” networks

Télécom ParisTech

Computer-mediated interactions

Page 31: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202331

Experiment: create two accounts The fomer (actual profile) discloses

more personal details, the latter (control profile) discloses less

Invite 100 users to friend them (50 each)

Friends provide feedback on how to enrich profiles (Comments, Messages, Likes, Shares)

Compare two accounts over 50 days

A social media experiment

Télécom ParisTech

Page 32: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202332

Observation notes:

– « Jusqu’à aujourd’hui, les retours sur les deux profils sont assez négatifs. Les connaissances de sexe féminin surtout ne se gênent pas pour exprimer leur aversion. Une amie définit le profil 1 comme ‘effrayant’, une autre qualifie la photo du profil 2 de ‘monstrueuse’ ».

–« Indication : utilisateur du profil 1 apprécie la cuisine japonaise et écoute de la musique punk. Il lit des bandes dessinées et des poètes de la beat generation ».

–« Profil 1 constamment ouvert dans mon navigateur. En automatique des petites fenêtres contenant des suggestions ou des ‘morceaux choisis’ par ses amis. ‘L’utilisatrice X est fan de l’artiste peintre Tel’ ; ‘L’utilisateur Y a aimé le dernier livre de l’écrivain Telautre’ ».

A social media experiment

Télécom ParisTech

Page 33: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202333

1. Two Facebook profiles initial state

A social media experiment

Télécom ParisTech

Page 34: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202334

2. Profile 1 discloses personal preferences

A social media experiment

Télécom ParisTech

Page 35: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202335

3. Profile 1 discloses bio

A social media experiment

Télécom ParisTech

Page 36: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202336

4. Profile 1 uploads a photo album

A social media experiment

Télécom ParisTech

Page 37: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202337

Compare social graphs

Disclosing profile has a larger, more varied network

Better management of social capital: balance bw bonding (social cohesion) and bridging (social connectivity)

A social media experiment

Télécom ParisTech

Page 38: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202338

Bridging

Bonding

A social media experiment

Télécom ParisTech

Page 39: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202339

A social media experiment

Télécom ParisTech

Page 40: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202340

Studying complexity

Télécom ParisTech

Page 41: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202341

Complexity and social science

Télécom ParisTech

Chaos, social dynamics, emergent behaviours

Page 42: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202342

Complexity and social science

Télécom ParisTech

Social systems, self-organization, autopoiesis, complex adaptive systems

Page 43: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202343

Agent-based modelling

Agent-based computer simulations

Generate socially consistent scenarios on a computer;

Analyse the resulting scenario outcomes to:

Identify sufficient conditions under which different outcomes emerge;

Assess their sensitivity to parameter changes.

An aid to perform a thought experiment.

Télécom ParisTech

Page 44: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202344

Agent-based modelling

The logic of an agent-based model

Generate an artificial population of agents in an environment;

Endow them with basic rules of behaviour;

Let them interact for a certain time and step aside;

Observe outcomes at the system level at the end.

Télécom ParisTech

Page 45: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202345

Agent-based modelling

KISS (Keep It Simple and Stupid)

Schelling’s segregation model (1973)

How tolerant individuals have to be in order to avoid collective segregation (the creation of ghettoes) in a given social space?

Some surprising results…

Télécom ParisTech

Page 46: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202346

Agent-based modelling

Télécom ParisTech

‘‘Pure’’ models ‘‘Empirical’’ models

. Built by abstraction from a target system (a social phenomenon or context).

. Mainly regarded as tools for generating, expressing and testing theories.

. Not always realistically representing choices and behaviors at the micro level.

. Enable in-depth reflection on the possible unintended social consequences of purposeful individual actions.

. Open to estimation and validation via qualitative and quantitative data.

. Quantitative data can be used to assess the probability that a certain event takes place within a given population of agents (either predictively or retrodictively).

. Use of qualitative data to inform simulation rules and parameters is also attested since the late 1990s (structural validation).

Page 47: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202347

Agent-based modelling

Télécom ParisTech

Page 48: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202348

Agent-based modelling

Télécom ParisTech

Page 49: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202349

Agent-based modelling

Télécom ParisTech

Page 50: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202350

Privacy

Télécom ParisTech

Page 51: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech51

The privacy challenge in social media

Periodic privacy incidents on FB

Mark Zuckerberg: ”Public is the new social norm”

Are we approaching the “End of Privacy” as we know it?

Alleged tendency to "renounce privacy" for an open, connected existence (publicness)?

The end of privacy online?

Page 52: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech52

The end of privacy online?

Page 53: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech53

The end of privacy online?

Page 54: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202354 Télécom ParisTech

Date Privacy-related incident Users’ reaction

05/09/2006 Introduction of News Feed (content and user updates aggregator).

Users’ uproar over the default opt-in policy. Creation of the advocacy group “Students against Facebook News Feed” to protest the new feature. The group attracts almost 300,000 members, leading to apologies by Mark Zuckerberg, Facebook’s funder and CEO.

26/09/2006 Facebook reinforces privacy options for users (to limit searchability and tie formation) to anticipate the gradual opening of its membership to any US and Canada college students with a valid email address and over the age of 13.

06/11/2007 Introduction of Beacon (advertising system aggregating purchase data over several platforms, most prominently Amazon).

Prominent political activist group MoveOn.org creates an online petition against Beacon. Their Facebook group reaches 50,000 members, which leads Mr Zuckerberg to issue an official apology. Beacon ultimately shut down in September 2009.

09/12/2009 Facebook changes its privacy settings, making sharing with everyone compulsory: legal names, profile pictures, and gender are now public by default.

An alliance of privacy organisations files a complaint with America’s Federal Trade Commission (FTC).

21/04/2010 Facebook introduces the Like button social plugin for external websites. Users can now log in, like and share contents (“frictionless sharing”) on other services through their Facebook account.

Prompted by their constituents, a group of American senators asks the FTC to establish privacy guidelines for Facebook. Privacy groups file a formal complaint to the FTC against Facebook’s “unfair and deceptive trade practice of sharing user information with the public and with third-party application developers”. At the end of May 2010, Mr Zuckerberg announces new and simplified privacy settings.

14/01/2011 Facebook makes users’ addresses and phone numbers available to external websites.

After negative feedback from users, Facebook disables the feature. At the end of the month, the fan page of Mr Zuckerberg is hacked and compromised. The following day, Facebook starts implementing https secure pages.

08/2011 Following a series of complaints filed by Austrian student association Europe v. Facebook. org, it emerges that Facebook fails to comply with the rule of allowing its users to download their own personal data: it provides only 39 over 84 personal data categories.

Negative media attention and creation of several campaigns requiring Facebook to give users full access to their data.

05/2012 Facebook proposes a new and more complex privacy policy while asking for generic “users’ feedback”.

40,000 user comments force vote on proposed alternatives to privacy policies.

20/06/2012 Facebook announces acquisition of facial recognition technology company Face.com (creates database of users’ biometric information through photo-tagging).

Privacy advocacy groups file complaint to the FTC recommending suspension of facial recognition technology and protesting creation of biometric profiles of users without their explicit consent.

Page 55: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech55

To find an answer to this question let’s try and build an agent-based model that represent the possible equilibriums for a system of agents disclosing personal informations online

Phase 1: empirical observation

Phase 2: modelling

1

2

Modelling privacy

Page 56: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech56

Remember our experiment on disclosure

Personal network of actual profile continues to grow in size and displays a distinctive balance between social cohesion (bonding) and social connectedness (bridging)

Disclosure is crucial: does this necessarily validate the ‘End-of-privacy’ hypothesis?

Modelling privacy

Page 57: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech57

Problematizing privacy In fact, online interactions

complexify the very notion of privacy

Traditional notion based on metaphor of concentric circles of intimacy

Mono-directional notion: a core of sensitive data to be protected.

This notion no longer seems adapted to interactions in a networked society.

Modelling privacy

Page 58: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech58

Privacy as a multi-directional, dynamic process

Online privacy better described through multi-directional negotiation

Individuals send signals to, and receive feedback from, their social environment.

Self-disclosure accompanies adaptation to signals from the (social) environment over time.

Modelling privacy

Page 59: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech59

We need to design a social system with: Formation of personal networks through bonding

and bridging ;• Disclosure needed to form ties;• Adaptation to signals from the environment through a

feedback process;

What will be the final configuration of the system, in terms of degree of disclosure?

Modelling privacy

Page 60: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech60

Behavioral rules:• Tie formation allowing for both

bonding and bridging social capital;• Binary on/off visibility settings;• Homophilous choice of network

contacts.

Parameters:• Tendency to value bonding /

bridging social capital;• Openness to cultural diversity.

Indicators:• Mean privacy level;• Number and size of components.

Our simulation model

Page 61: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech61

Resulting system configurations

Our simulation model

(1) Echo-chambers (2) Large components (3) Generalized connectedness.

Page 62: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech62

How parameter values affect results

Our simulation model

Treemap: varying modes of valuing bonding/bridging ties and levels of cultural openness. Size of rectangles is proportional to size of largest network component, colour represents differences in number of components.

Page 63: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech63

Effects on social division• When bonding prevails, echo-

chambers always emerge regardless of the cultural openness of agents;

• When bridging prevails, the degree of cultural openness determines whether the result is one or few large components.

Effects on privacy choices• When bonding prevails, average

privacy changes little regardless of the cultural openness of agents;

• When bridging prevails, high cultural openness prompts increased privacy protection.

Our simulation model

Evolution of mean privacy over time, with high bridging social capital and high cultural openness.

Page 64: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech64

Network structure matters• Relative value of

bonding/bridging ties affects final outcomes;

• Homophily need not be socially divisive;

Important to focus on motivations on people to form social capital online;

Networking service architecture likely to play a key role.

Results

Evolution of mean privacy over time, with high bridging social capital and high cultural openness.

Page 65: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/2023 Télécom ParisTech65

No “End of Privacy” in sight

Social media usage is not bound to destroy privacy

It is when connectedness is at its highest that privacy re-surfaces;

It becomes important to consider users’ attitudes in discussions of providers’ privacy policies.

Results

Privacy cycles in the presence of service provider interventions to unlock privacy setting by default

Page 66: Antonio A. Casilli - Networks, complexity, and privacy

Institut Mines-Télécom04/07/202366

Thank you!

Email : [email protected] Blog : http://www.bodyspacesociety.eu Twitter : @bodyspacesoc

Télécom ParisTech