August 2014

Sun Mon Tue Wed Thu Fri Sat
         

1

2
3 4 5 6 7 8 9
10 11

12

13 14 15 16
17 18 19 20 21 22 23
24 25 26 27 28 29 30
31            

Editor Login


Convener in chief:


David Lazer
(Methodology, Networked Governance)

Editors:


Stanley Wasserman
(Current Trends, Methodology, Social Networks)

David Gibson
(Social Networks, Interaction, Theory)

Yu-Ru Lin
(Networks, Visualization)

Ines Mergel
(Knowledge Sharing, Social Computing, Social Software, Government 20)

Maria Binz-Scharf
(Qualitative Methodology, Knowledge Sharing, eGovernment)

Alexander Schellong
(Admin, eGovernment, Government 20, Citizen Relationship Management)

Categories

Archives

Recent Entries

Recent Comments

Notification

Powered by
Movable Type 4.24-en




View Blog Stats

Blog Directory & Search engine
Academics Blog Top Sites

globe_blogs.gif
Blog Flux Local - Massachusetts
Blog Flux Directory

12 August 2014

Events/Announcements

Brian Granger on "Open, Reproducible and Exploratory Data Science" August 21

Please come by Northeastern next week for this talk by Brian Granger, especially if you are interested in IPython/Jupyter.

Open, Reproducible and Exploratory Data Science
Professor Brian Granger
Physics, Cal Poly State University
Lead Developer and Co-Founder, IPython and Jupyter Projects

Center for Complex Network Research (CCNR), Northeastern University
3:30 - 5:00pm, Thursday, August 21

Data Science involves the application of scientific methodologies to data driven computations across a wide range of fields. As Drew Conway has clarified, it sits at the intersection of hacking/programming, math/statistics and domain specific expertise. Because data science is data- and computing-centric it requires powerful software tools. In this talk I will describe open source software tools for data science that i) are built with open languages, architectures and standards, ii) promote reproducibility and iii) are optimized for exploratory data analysis and visualization.

In particular, I will describe the Jupyter Notebook (formerly named IPython), an open-source, web-based interactive computing environment for Python, R, Julia and other programming languages. The Notebook enables users to create documents that combine live code, narrative text, equations, images, video and other content. These notebook documents provide a complete and reproducible record of a computation, its results and accompanying material and can be shared over email, Dropbox, GitHub or converted to static PDF/LaTeX, HTML, Markdown, etc. Most importantly, the Jupyter Notebook is built on top of an open architecture for interactive computing that is completely language neutral, allowing it to serve as a foundation for other data science projects and products.

One of the most important aspects of data science is interacting with data. This involves iterative cycles of visualization, computation and human computer interaction to extract understanding and make predictions. Jupyter now provides an architecture for interactive JavaScript/HTML/CSS widgets that allows users to interact with their data in a direct and simple way by automatically creating appropriate user interfaces for Python objects and functions. This allows the power of modern JavaScript libraries (d3.js, leaflet.js, backbone.js, etc.) to be leveraged in Python/Julia/R driven computations.

Throughout the talk, I will provide examples of how IPython is being used across a wide range of fields including science, engineering, social sciences, finance, computer science, industry, publishing and journalism. Jupyter/IPython is funded through the Alfred P. Sloan Foundation, the Simons Foundation, the National Science Foundation, Microsoft and Rackspace.

Bio
Brian Granger is an Associate Professor of Physics at Cal Poly State University in San Luis Obispo, CA. He has a background in theoretical atomic, molecular and optical physics, with a Ph.D from the University of Colorado. His current research interests include quantum computing, symbolic computer algebra, parallel and distributed computing and interactive computing environments for scientific computing and data science. He is a lead developer on the IPython project, a co-founder of Project Jupyter, creator of PyZMQ and is an active contributor to a number of other open source projects focused on scientific computing in Python. He is @ellisonbg on Twitter and GitHub.

By David Lazer | 9:53 AM | Comments (0)

1 August 2014

Prediction

August 1 Executive Election Predictions

As with last month, there are two sets of predictions presented this month. The first is with our version 1.0 model - the same model as what we have used for previous predictions and that will serve as a reference point for version 2.0 (which also uses slightly updated data as input, hence the DM1 vs DM2):

Model Version 1.0 Predictions
Turkey - incumbent party win - 98.6%
Brazil - incumbent party win - 98.8%
Bosnia and Herzegovina - incumbent party win - 82.9%
Bosniak Election - incumbent party win - 96.8%
Croat Election - incumbent party win - 74.9%
Serbian Election - incumbent party win - 74.9%
Bolivia - incumbent party win - 99.7%
Mozambique - incumbent party win - 99.6%
Uruguay - incumbent party win - 98.7%
Namibia - incumbent party win - 56.0%
Romania - incumbent party lose - 80.9%
Tunisia - incumbent party lose - 100%
Ref: C1, DM1, and DP14

Model Version 2.0 Predictions
Turkey - incumbent party win - 83.4%
Brazil - incumbent party win - 91.5%
Bosnia and Herzegovina - incumbent party lose - 58.2%
Bosniak Election - incumbent party win - 66.6%
Croat Election - incumbent party lose - 60.8%
Serbian Election - incumbent party lose - 60.8%
Bolivia - incumbent party win - 93.0%
Mozambique - incumbent party win - 89.4%
Uruguay - incumbent party win - 81.3%
Namibia - incumbent party win - 53.7%
Romania - incumbent party lose - 75.8%
Tunisia - incumbent party lose - 100%
Ref: C1, DM2, and DP14

Here are the notes for this month's predictions. First, Tunisia is set to 100% incumbent party loss probability because the incumbent's party was disbanded. Second, results for Bosnia and Herzegovina are difficult to interpret because there are three presidents being elected. We have attempted to break it down by individual presidential election, but all of these results should be taken as tentative until we have public opinion data for the elections. We expect to receive better data in the next few weeks that should help us better discriminate these elections in both models. Third, the prediction for Turkey has shifted because of the release of public opinion polls showing the incumbent party's candidate with a strong lead.

By David Lazer | 9:00 AM | Comments (0)

2 July 2014

Big data
Commentary

A modest proposal to Facebook

Many/most/all of the readers of this blog have now heard of the Facebook emotion contagion study published in PNAS last week. Briefly: Facebook researchers, in collaboration with scholars at Cornell and UCSF, experimentally manipulated the algorithm that determines the subset of posts you see on Facebook, such that some people saw more positive posts, and others more negative posts. Their finding was, roughly, that negativity begets negativity, and positivity positivity. This paper has gone through a remarkably fast cycle of "isn't that interesting but a bit creepy" to methodological critiques ("they're not really measuring emotion") to a vast "Facebook is unethically manipulating our emotions!"

This post is not a commentary on the science, the ethics of this study, when is consent required, the structure of ethical self regulation (via IRB) in the US vs other countries (usually with no equivalents of IRBs), or the generally important question of the implications of our increasingly algorithmically organized societies. These will be subjects for future posts, and of many future classroom discussions I will have with doctoral students about research ethics. Rather, my concern right now is that this event has the potential to damage our collective capacity to create knowledge regarding human society because of the potential for public relations fiascos for companies. Of course--knowledge production will continue regardless, but perhaps all be safely proprietary, within the research departments of companies. Such an outcome would be terrible, not only for our collective understanding of human society, but also for these companies, because, paradoxically, the participation in vigorous public intellectual debates is important for the capacity of developing proprietary knowledge. Knowledge does not grow in hermetically sealed silos, and it is not coincidence that our creative industries have grown up in near proximity to universities, which at their best are highly permeable intellectual hot houses.

I'd therefore like to make a modest proposal about academic-industry cooperation, which is that companies like Facebook should create opt-in experimental panels, with an initial clear, short, transparent and in your face, fairly general and flexible consent about the types of ways their sociotechnical environment would be (modestly) experimentally varied. (And if certain experiments exceeded those parameters, there could be an additional consent required for specific experiments.) Subsequent to the completion of a study, study participants would be informed of the study, with a plain English explanation of the findings, as well as access to subsequent publications. Indeed, I'd note that my team has created a platform along this model, Volunteer Science, which is partially built on top of the Facebook API. Our challenge is building a user base. Facebook would not have a problem building a volunteer army to help out science--they could have a million recruits tomorrow.

I don't claim this is a cure all, but it would cure a lot--indeed, I think the entire current mess would have been avoided if the research had been done on such a volunteer base.

I'd note that Facebook and the like would (and will) continue to do A/B testing, and generally experimentally tweaking their algorithms in ways that (1) create variations in individual experience, and (2) have potentially important consequences, individually and collectively. This should be vigorously studied by scholars, and debated and scrutinized in the broader society. But the issue of whether and how a company like Facebook can participate in academic research, and in particular conduct field experiments, is actually solvable.

By David Lazer | 10:25 PM | Comments (0)

1 July 2014

Prediction

July 1 Executive election predictions

As with last month, there are two sets of predictions presented this month. The first is with our version 1.0 model - the same model as what we have used for previous predictions and that will serve as a reference point for our updated version. The second is the beta version 2.0 that also uses updated data.

Model Version 1.0 Predictions
Indonesia - incumbent party lose - 75.5%
Turkey - incumbent party lose - 64.3%
Brazil - incumbent party win - 98.8%
Bosnia and Herzegovina - incumbent party win - 82.9%
Bolivia - incumbent party win - 99.7%
Mozambique - incumbent party win - 99.6%
Uruguay - incumbent party win - 98.7%
Namibia - incumbent party win - 56.0%
Romania - incumbent party lose - 80.9%
Tunisia - incumbent party lose - 69.0%
Ref: C1, DM1, and DP13

Model Version 2.0 Predictions
Indonesia - incumbent party lose - 70.7%
Turkey - incumbent party lose - 55.1%
Brazil - incumbent party win - 91.9%
Bosnia and Herzegovina - incumbent party lose - 59.1%
Bolivia - incumbent party win - 92.9%
Mozambique - incumbent party win - 89.8%
Uruguay - incumbent party win - 83.7%
Namibia - incumbent party win - 54.9%
Romania - incumbent party lose - 75.0%
Tunisia - incumbent party lose - 55.1%
Ref: C1, DM2, and DP13

Here are the caveats for this month's predictions. The results for Turkey are still early, since campaigning will not begin until July 11 and candidates are still not certain. Bosnia and Herzegovina has a three president system that we are still figuring out how to model. Currently we are counting a loss by any incumbent party as a loss. In Bolivia there is no official candidate list yet. Similarly, in Romania, most candidates have not yet been announced. Finally, Tunisia is a transition state and, while there are plenty of polls, the election has been delayed several times already and there are a lot of potential candidates.

By David Lazer | 9:00 AM | Comments (0)

13 June 2014

WIRE Workshop @ Harvard on June 17: Working with Internet Archives for Research

Many of the readers of this blog will be interested in this event:

WIRE Workshop: Working with Internet Archives for Research
Institute for Quantitative Social Science (IQSS)
Harvard University
1737 Cambridge Street
Cambridge, MA

Tuesday June 17, 2014
Room S010
1pm - 5pm

Please join us on Tuesday for a series of public presentations highlight ongoing research at the intersection of network analysis, large-scale data and archival Internet studies. This workshop is hosted by a team of scholars from Rutgers University, Northeastern University, and the Internet Archive. The aim of the workshop is twofold. The workshop will provide a forum for presentations and discussions of ongoing research involving community development and historical Internet data. Presentation sessions will focus on a variety of themes, derived from ongoing research about online community emergence and evolution. A closing session will be devoted to discussing future research needs and unanswered research questions with regard to data and access to historical Internet records. The workshop will provide a mechanism for discussing the functions that should be incorporated into a prototype historical Web extractor, and for outlining potential research questions to be addressed with a prototype tool and databases. In addition, key questions gathered during the workshop will serve as initial discussion points for the online community that will support ongoing interaction between researchers.

This event is cosponsored by the NetSCI Lab at Rutgers, NULab for texts, maps, and networks at Northeastern, IQSS at Harvard, and the Internet Archive.

For more information see:

AGENDA

1:00pm - 1:45pm: Opening Session
Welcoming Remarks
David Lazer, Northeastern University
Overview of the Archive Hub project and Internet Archive Research
Matthew Weber, Rutgers University

1:45pm - 2:30pm: Internet Archives and Research Potential
Insight into the Internet Archives
Kris Carpenter, Director, Web Archive, Internet Archive
Web Wide Crawls
Vinay Goel, Senior Data Engineer, Internet Archive

2:30pm - 3:00pm: Research Highlights
Ancient History of the UK Web
Eric Meyer and Scott Hale, Oxford Internet Institute

3:00pm - 3:30pm: Coffee Break

3:30pm - 5:00pm: Research Highlights
Research Infrastructure for the Study of Archived Web Materials
Neils Brugger, Associate Professor, Head of the Centre for Internet Studies and of NetLab, Aarhus University
ALEXANDRIA: Temporal Retrieval, Exploration and Analytics in Web Archives
Wolfgang Niejdl, Director, LS3 Research Center
WebScience and Archival Internet Research
Thanassis Tiropanis, Senior Lecturer, University of Southampton

5:00pm - 5:30pm: Challenges for Future Research

By David Lazer | 11:19 AM | Comments (0)

6 June 2014

Events/Announcements

PhD program in Complex Networks at IMT

Readers of this blog might find this doctoral program to be of interest:

2014/15 PhD program in Complex Networks at IMT Institute for Advanced Studies Lucca
Deadline for applications - July 14th 2014, 6 pm Italian time
To apply: http://phd.imtlucca.it

Applications are now being accepted for PhD students in Complex Networks for the 2014/15 PhD program at IMT Institute for Advanced Studies Lucca (www.imtlucca.it). The three year doctoral program is articulated in curricula. The 8 curricula currently offered are field-specific, although in many instances they share a common scientific background. The Complex Networks curriculum introduces a novel way to look at several natural and technological phenomena. In this curriculum we use this framework derived from the mathematics of Graph Theory, the database analysis of Computer Science and the modelling skills of Statistical Physics to describe some specific natural phenomena. Among the multidisciplinary research units at IMT, the research unit NETWORKS (Complex Networks, http://networks.imtlucca.it/) will be the primary contributor to the curriculum.

In particular, during the program students will carry out research in areas such as the analysis of communities in brain networks; the reconstruction of network of correlation from time series analysis; the use of bipartite networks spectral properties for the clustering of patients with similar diseases definition of chemical networks; the study of the root-apices interaction in plants and the study of channel networks in plants; the study of financial and economic networks from real data. Graduates from the curriculum will be qualified to work in universities, public and industrial research centers, and to take on professional roles and high-profile tasks and responsibilities in both private companies and public institutions.

All students are based in the newly restored San Francesco complex, a fully integrated campus in the historical center of the beautiful Tuscan city of Lucca. The campus includes residential facilities, including an on-site canteen, study and living rooms, and outdoor recreational spaces, all serving to add a cultural and social dimension to the professional experience of the doctoral program. Eligible students will also receive a research scholarship which amounts to approximately €13,600/year.

The PhD program at IMT attracts students from around the world, providing a truly international environment. English is the official language of the Institute. To further enhance the internationality of the program, all students have the opportunity to spend anywhere from 2-9 months abroad at a research institute or university, with the possibility of receiving additional financing through the Erasmus+ program.

To find out more about the institute, the requirements for admission and how to apply, please see http://phd.imtlucca.it.

The deadline for applications is July 14th 2014, 6 pm Italian time.

Find IMT on Facebook, Twitter, LinkedIn and YouTube for the latest news.

By David Lazer | 5:09 AM | Comments (0)

1 June 2014

Prediction

June 1 Executive election predictions

There are two models for this month. The first is our standard model. The second is the experimental version. We are still hashing out their relative merits, but for now we are posting predictions from both.

Model 1

Syria - incumbent party win - 99.7%
Mauritania - incumbent party win - 98.6%
Indonesia - incumbent party lose - 75.5%
Turkey - incumbent party lose - 64.3%
Brazil - incumbent party win - 98.8%
Bosnia & Herzegovina - incumbent parties win - 82.9%
Bolivia - incumbent party win - 94.0%
Mozambique - incumbent party win - 82.4%
Uruguay - incumbent party lose - 61.5%
Namibia - incumbent party win - 56.0%
Romania - incumbent party win - 96.8%
Comoros - incumbent party lose - 60.0%
Colombia - incumbent party lose - 57.8%

Ref: C1, DM1, and DP12


Model 2

Syria - incumbent party win - 86.4%
Mauritania - incumbent party win - 87.6%
Indonesia - incumbent party lose - 62.3%
Turkey - incumbent party lose - 57.1%
Brazil - incumbent party win - 90.7%
Bosnia & Herzegovina - incumbent party lose - 59.2%
Bolivia - incumbent party win - 77.7%
Mozambique - incumbent party win - 54.2%
Uruguay - incumbent party lose - 60.2%
Namibia - incumbent party win - 53.8%
Romania - incumbent party win - 81.2%
Comoros - incumbent party lose - 64.2%
Colombia - incumbent party win - 50.7%

Ref: C2, DM1, and DP12


Caveats:

1. Syria is, of course, a transition state. There is really no doubt that Assad will win. (Indeed, it may not pass the basic threshold of an election that can be lost, which is our baseline for inclusion in the training data.)

2. There are three presidents in Bosnia-Herzegovina. NELDA codes it as incumbent party losing unless all three are re-elected.

3. We are still waiting for official candidate registration (or completion of primaries) in Turkey, Bolivia, Uruguay, and Namibia.

By David Lazer | 9:45 PM | Comments (0)

12 May 2014

Social Networks and Webs of Spiders ----- Webs of Webs!


The social lives of spiders ..... who builds which webs?

This is a great piece (from James Gorman, one of the lead Science guys at The New York Times, on the structure and composition of webs built by spiders. Very cool.

By Stan Wasserman | 3:11 PM | Comments (0)

11 May 2014

Government Innovator Blog

For those in the community that come from government (US federal, state or from abroad) or have an interest in innovations in the public sector, I would like to recommend for you to visit Andrew Feldman's gov innnovator blog.

The blog covers a range of topics and trends relevant for public managers and policy makers based on audio or video interviews with active duty leaders in government. If you you have an important topic to share with the community, make sure you reach out to Andy so that he can interview you.

This is a side project for Andy. As a professional, Andy (Twitter: @andyfeldman) is a Special Advisor on the Evidence and Innovation Team at the White House Office of Management and Budget. He received an M.P.P. from the Harvard Kennedy School and a Ph.D. in Public Policy from Harvard University. I encourage you to check it out.

By Alexander Schellong | 8:59 AM | Comments (0)

10 May 2014

Events/Announcements

KDD Workshop on Learning about Emergencies from Social Information

The 2014 KDD Workshop on Learning about Emergencies from Social Information
(KDD-LESI 2014)

to be held on August 24, 2014, New York City
co-located with ACM SIGKDD 2014
--------------------------------------------------------------
Important Dates:
** Submission Deadline: June 10, 2014
** Notification of Acceptance: July 1, 2014
** Workshop date: August 24, 2014

Workshop Organizers:
Yu-Ru Lin (yurulin@pitt.edu)
James Bagrow (jbagrow@uvm.edu)

This workshop will bring together researchers interested in problems
of information processing to meet the growing challenges in emergency
situations. Our aim is to foster a research conversation among
computer scientists, social scientists, and other interested
participants to discuss issues and challenges relating to emergency
understanding, including theoretical, methodological, ethical, and
political questions regarding the study of large-scale social
information data.

We invite submissions on various research topics within the contexts
of emergency study using social communication data, including but not
limited to the following:
* Extracting emergency events from big data
* Measurement of relevance and user activities through emergency
information retrieval in social media
* Identifying misinformation during emergencies and crisis events
* Evaluation framework for the emergency mining algorithms
* New technologies (e.g., mobile applications) for mining and
deploying emergency information
* Fusion of social communication features, metadata, user generated
content, and social context within the emergency situations.
* Scalable or real-time architecture for large-scale emergency
information processing, mining and visualization
* Emergency social and information structure pattern discovery and
predictive modeling.
* Security and privacy management for emergency information processing.
* Human computer interfaces for emergency data mining and crowdsourcing

We provide different submission formats: full papers, short papers,
and posters. We encourage submissions which present early stages of
cutting-edge research and development. The format is the standard
double-column ACM Proceedings Style. Submissions that do not comply
with the above guidelines may be rejected without review. Additional
information about formatting and style files are available online at:
http://www.acm.org/sigs/publications/proceedings-templates
* Full research papers (5 to 8 pages)
* Short research papers (up to 5 pages)
* Posters and software demo (up to 2 pages)

All submissions must be entered into the reviewing system.


Contact: kddlesi14@gmail.com

By David Lazer | 4:22 PM | Comments (0)