Information Science

INFO 6940: Special Topics: Analytic Social Systems

Spring 2012
Tue/Thu 2:55-4:10 PM, Location: 301 College Ave (large conf room)
3 credits, S/U Optional

Professor: Paul Ginsparg (452 Phys.Sci.Bldg,
Office hours: Wed 1-2 PM (or by appointment)
Course website: (this page)

This course will cover recent research on the structure and analysis of social and technological networks, and some of the background material necessary to understand it. Topics are likely to include random graphs and small-world properties, cascades and critical phenomena, heavy-tailed distributions, game-theoretic models of behavior, decision-making, clustering and communities, social networks and the web, and reputation systems. The early part of the course will be more lecture-centric, using some of the more advanced sections of the Easley/Kleinberg text (Networks, Crowds, and Markets), and additional material from the research literature. It will then transition to a mixture of lecture and discussion of more recent selections from the research literature. There will also be some overlap with topics in Info 6850, albeit with less mathematical emphasis. The goal is to make the current literature accessible to students with less formal mathematical background.

Prerequisite: graduate standing. Introductory-level background in probability and statistics. Some basic programming experience might also be helpful for manipulating network datasets.
Note: Can be used for graduate core credit for formal analysis of social systems

Lecture 1 (Tue 24 Jan 12)

Course overview (Graph theory, Game theory, Strategic Interaction, Information Networks, Population Models, Structural Models, Aggregate Behavior)
Intro chapters 1,2,3 from E/K
Other refs: NYTimes: Separating You and Me? 4.74 Degrees re some recent local work arXiv:1111.4570, arXiv:1111.4503, arXiv:1112.1115

Lecture 2 (Thu 26 Jan 12)

Finished chpt 3 of E/K, including "advanced material" on Betweenness Measures and Graph Partitioning.
Plan is to complete chpt 4 (homophily, etc) next Tues, and begin discussion of articles mentioned in lecture 1.

(After class, I also mentioned these 2002 articles Uncloaking Terrorist Networks , Mapping Networks of Terrorist Cells; see also this blog entry (Updated 2007))

Another addendum: by coincidence Dunbar joined the networkers network over the weekend, arXiv:1201.5722 (Sex differences in intimate relationships, et al. + Barabási, Dunbar)

Lecture 3 (Tue 31 Jan 12)

Threaded into discussion of chpt 4, an early popular description of the "network contagion" mentioned on Thurs can be found here: Study Says Obesity Can Be Contagious (Jul 2007), and a more recent popular article questioning the results is this one: Disconnected? (Jul 2011: 'Critics now wonder how the "social contagion" studies ever passed peer review').
The latter mentions these technical articles (just for a flavor, not yet to be read in detail):

The first of the technical articles is described in this blog entry: Homophily, Contagion, Confounding: Pick Any Three (Apr 2010), explaining that equally good explanations can be found without invoking contagion or influence.

Even before the recent set of technical articles disputing the C/F network contagion media hype, there was already some popular backlash against the notion of causative network effects, e.g. from spring '10: Has a Plague of Social Illness Struck Mankind? (Apr, 2010) and Doubts About the Social Plague Stir in the Human Superorganism (Apr, 2010)

Note added, for a discussion of modularity and how it is used together with link betweenness centrality to find optimal community structure, see cond-mat/0308217 (Newman/Girvan, 2003)

Lecture 4 (Tue 2 Feb 12)

Finish up from last time (modularity, selection vs social influence in Wikipedia edits), then discussion of arXiv:1111.4570

Note: see Networks Journal Club for 7 Feb noon discussion of above arXiv:1004.4704

Lecture 5 (Tue 7 Feb 12)

Finish up discussion of arXiv:1111.4570 (see also this follow up arXiv:1205.5509 ) and also some of arXiv:1111.4503 (please have another look before class, specifically assortative mixing)

(Re the second, note also Why Your Friends Have More Friends Than You Do (Feld, 1991))

Then we'll finish Part I of the book with discussion of sections 5.4,5.5 of chapter 5 (at the end of the 3rd week, you're supposed to be familiar with all of chapts 1--5 of the course text)

Lecture 6 (Thu 9 Feb 12)

finish finishing chapter 5 (covered 5.4, 5.5, on weaker forms and generalizations of structural balance).

Also discussed The Role of Social Networks in Information Diffusion (see also blog post). In parallel with earlier discusions (which updated Milgram's small world to 2011 on-line social data), this one effectively updates Granovetter's strength of weak ties and their contribution to diffusion of novel information to modern electronic large-scale datasets.

Also mentioned that you might start perusing arXiv:cs.SI and arXiv:physics.soc-ph (or go back further, or check some recent conf proceedings) for student presentations of articles to class.

Lecture 7 (Tue 14 Feb 12)

For the next two weeks we'll be covering population models of network dynamics, focusing on the latter parts of chpts 17 and 18 of the text, and some related articles to be posted. You should be familiar with all of 16-18 by end lecture 10.

Lecture 8 (Thu 16 Feb 12)

Finish chpt 17, some of chpt 18 (popularity...)

Additional refs: Salganik/Dodds/Watts (2006), and popular version "Is Justin Timberlake a Product of Cumulative Advantage?"

Lecture 9 (Tue 21 Feb 12)

Finished chapter 18 on network effects, specifically section 18.7 (and was planning to discuss some articles involving recommender systems and their possible benefits and drawbacks, continuing the discussion also of the Watts et al experiment, deferred to Thurs).

Here are the older (INFO 2950) problems I mentioned re the Polya urn and related.
I also used this slide (from INFO 4300) re Amazon/Wikipedia power laws, based on this (INFO 2950) problem on the "Long Tail".

Additional popular refs re power laws for cities: tale of cities (Glaeser), math/city (Strogatz)

Lecture 10 (Tue 23 Feb 12)

The next focus will be the structural network models of chapters 19-21, covered most of 19.1-19.4 .

Discussed a bit of this article posted last week arXiv:1202.2461 (How the Scientific Community Reacts to Newly Submitted Preprints: Article Downloads, Twitter Mentions, and Citations), but not in extensive detail for reasons mentioned. Spent more time discussing arXiv:0907.4740.

We will also start on the logistics for student presentations.

Week of 27 Feb - 2 Mar

Lecture 11 (Tue 6 Mar 12)

We finished cascading behavior, role of weak ties, collective action, and "bilingual" extensions (sections 19.3-19.6 and 19.7C).

We will also discuss some of the choices for presentation, to help guide 2nd half syllabus. Please come prepared with your selections.

Lecture 12 (Thu 8 Mar 12)

We will cover epidemics, specifically sections 21.7-21.8.

We will also continue discussing the choices for presentation, to help guide 2nd half syllabus. Please come prepared with your selections.

Lecture 13 (Tue 13 Mar 12)

Finished the discusion of coalescent phenomena (including discussion of geometric series, see older (INFO 2950) problem, and world population).
Tied up some lingering loose threads re info cascades from sections 19.7A,B, and mention of SIR class models (21.3, 21.4).

Lecture 14 (Thu 15 Mar 12)

Will complete the technical discussion of small-world phenomena, sections 20.3-20.7, and finalize schedule for presentations after the break. (so far: Leo: 3 Apr (visualization), Alistair: 5 Apr (Digg); Elizabeth: 10 Apr (effect of exposure to preferences), Shion: 12 Apr (mix at mixers); Wei: 17 Apr, Matt: 19 Apr (assortative happiness) Xiying: 24 Apr, Jaeyoon: 26 Apr );

Spring Break (19-23 Mar 12)
Note: so far in the book we've covered Part I (chpts 1-5), Part V (chpts 17,18, will return to 16), Part VI (chpts 19-21)

Lecture 15 (Tue 27 Mar 12)

After completing some final details on sect 20.7, will review chpt 16 (Bayes Theorem) in preparation for part VII on aggregate behavior, probably covering chpts 22-24 in next few lectures.

Lecture 16 (Thu 29 Mar 12)

Completed review of chpt 16 (Bayes etc) and then chpt 22 (covered only to 22.2, so 22.3- next week).
See also notes on expected number of steps to avalanche, as discussed in class.

Lecture 17 (Tue 3 Apr 12)

Covered 22.3 (Aggregate beliefs and the wisdom of crowds).

Lecture 18 (Thu 5 Apr 12)

Finish Chpt 22.4-22.10 (Markets and Information), and likely cover 23.7-23.10 (voting as information aggregation) next week

Iowa Electronic Markets (IEM (Univ Iowa)), e.g., 2012 presidential election market graph

Some Intrade links: i) How it Works (note contracts are for $10 shares). ii) Republican Presidential Nominee in 2012: Romney at $9.52/share (95.2% chance), Santorum at $0.08/share (.8% chance, 103 shares available to buy [think even will occur], 70 shares available to sell [think event won't occur]). iii) President in 2012: Obama at $6.08/share (60.8% chance)

Lecture 19 (Tue 10 Apr 12)

We'll also finish some last details of chpt.22 (Markets and Information; though 22.6-22.9 still pending), and start covering 23.7-23.10 (voting as information aggregation).

Lecture 20 (Thu 12 Apr 12)

finished 23.7-23.10

Lecture 19 (Tue 17 Apr 12)

lecture: 24.1-24.2

Lecture 20 (Thu 19 Apr 12)

lecture: 24.2-24.3

Lecture 21 (Tue 24 Apr 12)

Covered 6.1-6.4. (So far we've covered roughly Parts I,V,VI,VII of the text, focusing on the more advanced material. The lecture plan for the remaining two weeks is to cover the first half of Part II (game theory), emphasizing the advanced material in chpts 6,7.)

Lecture 22 (Thu 26 Apr 12)

Here is a brief description of Boczkowski's work I mentioned re the "mismatch between supply and demand of news in the elite media" (see also his IS colloquium here last year)

Covered 6.5-6.8 (more on Nash equilibria)

Lecture 23 (Thu 1 May 12)

Lecture: finished 6.8 - 6.10 (A,B) finish chpts 6, 7

(Re George Price, see Death of an Altruist [also this radiolab audio link]
Re "Predicting X from Twitter is a popular fad within the Twitter research subculture", see this critical bibliography of election prediction using twitter data)

Lecture 24 (Thu 3 May 12)

Lecture: evolutionary game theory: chpt. 7

(Note: a recent observation about prediction markets)