Upload
angrytx
View
219
Download
0
Embed Size (px)
Citation preview
8/13/2019 Unobtrusive Social Network Data Through Email
http://slidepdf.com/reader/full/unobtrusive-social-network-data-through-email 1/7
8/13/2019 Unobtrusive Social Network Data Through Email
http://slidepdf.com/reader/full/unobtrusive-social-network-data-through-email 2/7
Report Documentation Page Form Approved
OMB No . 0704-0188
Public reporting burden for the collection of information is estimated to average hour per response, including the time for reviewing instructions, searching existing data sources, gathering and
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,
including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 21 5 Jefferson Davis Highway, Suite 204, Arlington
VA 22202-4302. Respondents should be aware that notwithstanding an y other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it
does not display a currently valid OMB control number.
1. REPORT DATE
DEC 2008
2. REPORT TYPE
N/A
3. DATES COVERED
4. TITLE AND SUBTITLE
Unobtrusive Social Network Data From Email
5 a CONTRACT NUMBER 5 b GRANT NUM ER5 c PROGRAM ELEMENT NUMBER
6. AUTHOR(S) 5d. PROJECT NUMBER
5e. TASK NUMBER
5f. WORK UNIT NUMBER
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES)
Network Science Center United States Military Academy West Point,
New York 10996
8. PERFORMING ORGANIZATION
REPORT NUMBER
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10 . SPONSOR/MONITOR'S ACRONYM(S)
11 . SPONSOR/MONITOR'S REPORT
NUMBER(S)
12 . DISTRIBUTION/AVAILABILITY STATEMENT
Approved for public release, distribution unlimited
13 . SUPPLEMENTARY NOTES
See also ADM002187. Proceedings of the Army Science Conference (26th) Held in Orlando, Florida on 1-4
December 2008
14 . ABSTRACT
15 . SUBJECT TERMS
16 . SECURITY CLASSIFICATION OF:
a. REPORT
unclassified
b. ABSTRACT
unclassified
17. LIMITATION OF
ABSTRACT
c. THIS PAGE
unclassified UU
18 . NUMBER
OF PAGES
19a. NAME OF
RESPONSIBLE PERSON
Standard Form 298 Rev. 8-98) Prescribed by ANSI Std Z39-18
8/13/2019 Unobtrusive Social Network Data Through Email
http://slidepdf.com/reader/full/unobtrusive-social-network-data-through-email 3/7
outlook archived iles nd tores ha t ata nto n SQL-
database.
Communication via mail an e ivided nto wo
types f relationships: he human-computer nteraction;
an d he omputer-computer nteraction. eople re
usually os t amiliar ith he uman-computer
interaction, where they sit at a computer, write an email,
an d push send ; or they login to their email account an d
read messages ontained n the inbox . he omputer-
computer nteraction, s ctually n utomated xchange between wo omputers, ften ith everal ther
computers erving s ntermediaries n he elivery
process. A message en t rom ne omputer s eceived
by he arget omputer(s), n ts lectronic orm, via client oftware program that ultimately opies he mail
message from its host server. The email message is stored
on esignated entral erver ntil he eceiver picks
up he message rom he erver. hi s rocess s he
electronic ersion of picking p ackage t he os t
office. The lectronic mail an be elivered to he post office repository fo r yo u to physically pick up , or directly
to your ersonal mail ox or ou o pick p. Once target omputer icks p he essage, he uman- computer nteraction llows he human to ead, rint, r
store he lectronic essage ia heir mail lient
software.
There are everal different ways n which mail ca n be elivered hrough he omputer-computer nteraction
in he orld-wide lectronic mail rchitecture. he
message can be delivered to the equivalent of post office lobby-box, alled n MAP erver. he mail an e
delivered o ersonal mailbox, alled ost ffice Protocol POP) erver. he mail an lso e outed through a Microsoft Exchange (MSEx) server. here are
many echnical ifferences between hese mail ervers,
bu t heir purpose s he ame. owever, he rincipal
difference between n MAP nd POP mail erver s
the torage eature of the erver. An MAP erver will
allow ou or your mail lient oftware) o ersist, or
store, your email physically on that server. A PO P server only serves as a temporary holding station fo r a message
that is removed once it ha s been retrieved by your email client oftware. n SEx erver s icrosoft proprietary ystem ha t s widely se d hroughout he
DoD. While it ha s some additional security features, it is more ifficult o xtract mail etwork ata rom his
system ecause f he ropriety ata ormat ha t
Microsoft institutes. n IMAP server is designed to store
the essage ve n fter he mail as een nitially
retrieved. t hould be noted that the POP protocol alls
for an email to be removed from the incoming mail bo x
once t as een etrieved, owever om e oftware
extensions o llow or ead-only ccess o he POP inbox, esulting n he message emaining n he nbox
when etrieved nd s herefore managed by he lient
software level. The popular Yahoo mail service
implements this feature fo r paying customers.
Once a target computer receives an email, the human- computer interaction involves the computer displaying the
message sing lient oftware. mail messages t he computer-computer nteraction evel re os t ften
formatted in a world-wide standard format called MBOX. MBOX llows or ifferent mail lient oftware
programs o ccess he mail rom he erver without
confusion. Th e MBOX ormat pecifies wo ections of the mail, he header section an d the body section. he
header ection ncludes he rom:, o: , C: , CC:, Subject:, nd ate:, nformation. he ody ection
contains he message ex t nd ny ttachments o he
email.
The MSEx erver oe s ot tore messages n he
MBOX format. icrosoft's proprietary tandards reate
technical nd icensing urdles n ccessing mail ata
directly rom the erver n ny manner ther than using Microsoft oftware. Unfortunately, he MSEx ormat s widely se d hroughout oD , aking mail ata
extraction more difficult. here are three approaches that we have discovered for extracting email content from an
MSEx ormat. ne pproach s ustom lient-side
visual asic atch McCulloh, t. l. , 007). nother
client-side pproach nvolves sing NetMap, which s plug-in fo r Microsoft Excel 2007 that extracts email data from proprietary .pst ile nto n Excel ormat. he
data an hen e manipulated r aved o ther ile
formats. he third approach involves parsing header data
from erver og ile. hese pproaches ill e
discussed in more detail in this paper. nalysis of dyad counts ill e se d o ompare he erformance f client-side ata ollection ith entralized ata collection.
3. METHOD This study involves monitoring the email traffic of 24
mid-career Army officers in a one-year graduate program
administered jointly by Columbia University an d the U.S. Military cademy USMA). ac h f he fficers
participating ere sked o ign onsent orm n
accordance ith he nstitutional eview oard IRB),
approved y he SMA uman ubjects esearch Review oard llowing heir ata o e ollected or
research purposes.
As part of this study, the participants permitted us to place a custom developed program (McCulloh et al , 2007) that works in conjunction with their MSEx Outlook email
accounts. This program llowed us o ollect mail ata
from the sent items folder found on participants' personal
computers. he nformation ncluded ll f he eader
information associated with an email. W e did no t view or
8/13/2019 Unobtrusive Social Network Data Through Email
http://slidepdf.com/reader/full/unobtrusive-social-network-data-through-email 4/7
include the body of th e email in the study. We were also able o ollect imilar mail header nformation irectly
from he og iles maintained y he Directorate Of Information Management DOIM). he ata ollected
from the custom program is referred to as the Client-Side
Method, while data collected from the DOIM log files is
referred o s he entralized ethod. e id ot
investigate NetMap s n pproach s t as dentical underlying mail ata-sourcing apabilites nd
functionality only with a different, al beit a more elegant,
user interface. he email data collected from all methods was analyzed using a dynamic network analysis approach
(Carley, 2003).
3 1 Client Side Method A lient ide Visual Basic or Applications VBA)
program was installed on the personal computers (PC) of all participants, n the session window of their Microsoft
Outlook. Details of this data collection scheme to include
the visual basic code are outlined in detail by McCulloh, et al. (2007). t is designed to overcome the difficulty in pulling information from a subject's sent mail folder in a
proprietary Outlook Exchange system. This patch is easy
to mplement n Visual Basic nd works harmoniously with Microsoft Outlook. The principal nvestigator could
then compile the data from all participants into on e master
file an d ensure anonymity of the names.
On e of the chief advantages in managing a client-side
patch s he ow-level ontrol n athering ata. researcher oe s ot have o btain ermissions rom network administrator to collect email data. hey merely
need the consent of the monitored individuals, who must login o heir Outlook or he lient-side atch o e
installed. urthermore, he rogram esigners an pick an d choose which data to import from the local client. If,
for example, we wanted to include message content, then
that could have been an option. W e could have also just as
easily gathered incoming email traffic, as opposed to only monitoring utgoing mail. hi s ould rovide urther
insight into reas uc h as whether a user classifies mail
as junk mail, whether they delete n ncoming message, or even if they flag a particular message as important.
Managing he ata ollection rom he ndividual
participant required minimal effort. Once fully developed an d nstalled, he isual asic atch s ittle o o
overhead on the part of the user to manage. Furthermore,
these articipants elt more omfortable nowing ha t
they av e om e egree f ontrol n ow he ata s
collected. While his ould mpede he ata ollection
process, the subjects felt more comfortable knowing what
was actually monitoring their email. nitially, most of the
participants' mail were en t to ther students r people
affiliated with heir raduate rogram. ithin wo o three weeks, the participants began to email family
members an d friends. W e suppose that this represents an
increased level of trust. n the beginning, participants felt
that their email needed to appear strictly business related.
Gradually, s he y ncrementally en t ersonal mail
messages while they were at work without an y negative
consequences, they began to feel comfortable an d appear
to av e eturned nto ormal adence f mail
communication. os t of the participants ne w ow o remove he atch when heir participation n he project
ended. everal articipants aid hey el t ore
comfortable nowing ha t he oftware ending he principal investigator information was on their computer,
an d that Big Brother was not pulling their information
from somewhere else.
3 2 Centralized Method As n lternative method, we eveloped oftware
application which nalyzes mail ata athered irectly
from a centralized mail xchange erver. This oftware
gathered ata ve r ive month eriod nd xtracted those email messages which were sent an d received from the participants in this study. he server log files contain the mail eader nformation. hi s nformation as
parsed into the same format as the client-side method.
With this method of data collection, he participants
were ot ware of the precise im e ha t he ollection
process tarted. hey did provide consent in accordance
with the RB, owever, we were ot required o nform them f he xact ate he n ollection ould egin.
There as o ignificant bservable hange n he
participants' pattern of communication. he entralized
method was completely unobtrusive.
3 3 Dyad Analysis It was not clear at the beginning of this investigation
whether mail ommunication within omogenous
group of people would appear random, if it would remain relatively consistent from week to week, or if there were
identifiable factors ha t would ffect changes n network
structure. To investigate the structure of the network, we
computed the dyad count. he dyad count, defined as the communication between tw o nodes (Wasserman, & Faust,
1994) istinguishes hree ifferent ypes f communication:
symmetric,
utual,
nd
ull.
n
n asymmetric dyad, on e node talks to another, bu t does no t
receive a response. This type of communication could be
an example of a group that ha s members who are sending
ou t nformation. mutual ya d ignifies wo odes communicating ith ach ther. his yp e f
communication might occur n group ha t ollaborates
equally, r ne n which ubordinates erify or larify
directives. inally, ull ya d ccurs when wo odes which re art f he etwork o ot ave ny communication ctivity. n ya d ount, we onduct
8/13/2019 Unobtrusive Social Network Data Through Email
http://slidepdf.com/reader/full/unobtrusive-social-network-data-through-email 5/7
census nd abulate he umber f ull, mutual, nd
asymmetric yads. ith 4 embers n ur tudy
comprising network, here xists 76 ombinations of
possible undirected pairs. ach of the 276 dyads could be
either null, mutual, r asymmetric. he dyad counts re
compared or data collected with the lient-side method,
centralized method, nd with alendar f ignificant
events.
4. RESULTS
There were ignificant ifferences n he lient-side
an d centralized methods of data collection. he data from
both methods as oded s meta-network Carley, 2002). onsidering ha t he articipants re andom
sample f mid-career Army fficers ha t ll ulfill he same ole f tudent n he rganization, e ight
hypothesize ha t he mail elationships ormed n he network are random. Given that there are 24 nodes in the
network, here xist 24 3 = 52 ossible yads. e can test the hypothesis:
H0: Graph ~ Binomial(552, 0.5)
HA: Graph ̂ Binomial(552, 0.5),
using the test statistic z =( / - E(/)) / Sqrt( V(/) ), where /
is the number of directed links in the graph. hi s reduces
to z = / - 276 ) / 11.75, where / is the su m of the mutual
an d asymmetric dyad counts. Under the null hypothesis,
this number follows standard normal distribution. he
p-value was ignificant at the 0.05 evel fo r most weeks,
providing vidence o eject he ypothesis ha t mail
communication atterns re andom inomial ith probability arameter f .5 . eek ith corresponding p-value that was no t significant at the 0.05
level ca n be identified in Table by the 95% confidence
interval on the Binomial parameter/? that includes 0.5.
A onfidence nterval n he robability f communication an e onstructed or ac h eek
according to the expression given by ,
P±z„ PII552
where ̂
is
he
maximum
ikelihood
stimate
f
he unknown arameter n he ssumed inomial
distribution nd qual o 52. able hows he
mutual, asymmetric, an d null dyad counts recorded using
the lient-side nd entralized methods. he ight most
column of Table hows the 95 % confidence interval on the random probability of communication. confidence
interval that spans .5 will correspond to a significant p-
value in the random binomial hypothesis test above. or
each week in Table , tw o values re hown fo r each of
the ya d ounts: Mutual, symmetric, nd Null. The
numbers n the top of each cell in Table orrespond to
the lient-side ata ollection method. he umbers n
the ottom f ach ell n able orrespond o he
centralized data collection method. he data presented in
Table orresponds to the time period beginning with the first week of the pring emester nd nding with he
week efore pring reak. he tudents oo k heir
comprehensive xa m ollowing pring reak nd he n
began to transition to their military duties t West Point.
Therefore, his ata epresents easonable im e eriod
for comparison of the client-side an d centralized methods of data collection.
Table 1 . Recorded directed links using client-side an d
central methods. Week Mutual Asymmetric N u l l C o n f i d e n c e
1 3 J a n 2 8 0
5 4
44
8 9
232
1 3 3
0 . 0 6 , 0 . 1 0
0 . 2 2 , 0 . 3 0
2 0 J a n 2 8 6
2 8
8 8
8 3
1 8 2
0
0 . 1 4 , 0 . 2 0
0 . 5 0 , 0 . 5 9
2 7 J a n 2 8 0
1 1 8
7 8
9 2
1 9 8
6 6
0 . 1 1 , 0 . 1 7
0 . 3 4 , 0 . 4 2
3 Feb 2 8 8
2 2
1 6 2
8 1
1 0 6
0
0 . 2 7 , 0 . 3 5
0 . 4 7 , 0 . 5 5
Feb2 8 0
1 1 2
1 4 8
1 0 0
1 2 8
6 4
0 . 2 3 , 0 . 3 1
0 . 3 4 , 0 . 4 2
1 7 Feb 2 8 6
23
1 1 4
7 9
1 5 6
0
0 . 1 8 , 0 . 2 5
0 . 5 2 , 0 . 6 0
24 Feb 2 8 26
2 4
1 0 8
9 2
1 4 2
0
0 . 2 1 , 0 . 2 8
0 . 4 9 , 0 . 5 8
2 Ma r 2 8 8 4
32
1 9 2
5 1
0
0
0 . 4 6 , 0 . 5 4
0 . 5 8 , 0 . 6 6
9 Ma r 2 8 26
2 4
1 4 3
7 3
1 0 7
0
0 . 2 7 , 0 . 3 4
0 . 4 6 , 0 . 5 4
Client-side dyad counts are above central dyad counts.
It ca n be
een in the Confidence
olumn of Table
that here s tatistically ignificant ifference n he probability of communication between the client-side an d
central ata ollection ethods or ll eeks, y
observing ha t he 5% onfidence ntervals o ot
overlap. n ll ases, he lient-side ethodunderestimates he probability of communication n he
network. he general pattern of the probability parameter
is correlated at a value of 0.69, which is low considering
they re stimates n he am e roup f ndividuals
during he am e week. he lient-side ata ollection
method is therefore biased.
The ya d ount nalysis an rovide dditional
insight nto he rganizational ynamics f he participants by comparing their probability of interaction
to ignificant vents n heir cademic alendar. e
restrict ur nvestigation o ata ollected sing he
centralized method ince it s omplete. he entralized
method aptures ll ata en t r eceived hrough he
central erver. he maximum likelihood stimate of the parameter, , n he inomial istribution f yads s
plotted over time an d displayed in Figure 1 .
8/13/2019 Unobtrusive Social Network Data Through Email
http://slidepdf.com/reader/full/unobtrusive-social-network-data-through-email 6/7
Maximum Likelihood Estimate of inomial Parameter
Figure 1 . ML E of parameter » using centralized method.
Th e lowest ML E of p s hown in the irst week of
the emester, when he participants were just eturning
from Christmas leave. hi s was followed by an increase
in ommunication s he roup begins to plan or group
academic ssignments, arpooling, nd ther
administrative ssues. he ow points n the ML E of p
occur uring he weeks of 27 anuary nd 0 February
when major roup cademic rojects r resentations were ue . his s onsistent with he indings f
McCulloh, et. l. 2007) who observed a similar decrease in mail ommunication uring imes of group ctivity.
They hypothesized that during times of increased face-to-
face ommunication, people ommunicate verbally nd have es s im e nd eed or mail ommunication.
Furthermore, uring hese imes f ncreased ubgroup
activity, people have es s im e o write nd espond o
emails rom ndividuals utside f heir mmediate
subgroup. ollowing he roup ssignments ue uring the week of 10 February, the next major academic vent
was the comprehensive exam following Spring break.
A similar dyad analysis or the client-side method is
shown n igure . he haracteristic ip n mail communication orresponding o roup ctivity s ot
clear. areful eview f the participants' cademic calendar oes ot eveal ny ctivities r vents hat
would explain the behavior of the plot in Figure 2. hi s
further suggests the importance of centralized email data collection.
Maximum Likelihood Estimate of inomial Parameter
Figure 2. ML E of parameter » using client-side method.
Th e client-side method of ata collection is not completely without merit. It ca n still be seen in Figure 2
that the first week ha s the lowest ML E ofp. here is also a dip in the plot for the group assignment for the week of
27 January. he identification of the week of 10 February
is missed owever. hi s uggests hat ven lient-side
data ca n provide some insight into group behavior. his
may e n ppropriate method o se when omplete
centralized data is unavailable. entralized data may be
unavailable or easons of security, rivacy, amage, r
other technical difficulties. n these situations, the client-
side method may till provide valuable nformation n social network behavior.
5 CONCLUSION We found that the primary advantage to utilizing a server-
side method to gather data is the improved data integrity. Every user with n mail ccount must oth end nd
receive ata rom hat ccount's ssociated mail erver.
Therefore, o nsure that ll data s gathered t must be
collected t ts ource. ll ata ontained within he
centralized server is available for collection, such as from, to, cc, bcc, subject, time of receipt at the mail server, etc. Copying data directly rom he erver llows he ocial
network nalyst o ccurately tudy ll mail
communications within tudy group or those utilizing their given email address.
Implementing erver ased pplication lso
precludes he ubjects nvolved n he tudy rom corrupting an d inserting bias into the data. With a client-
side application, users had the ability to turn off, remove or disrupt the xecution of the program used to monitor
email. With a server-side collection technique, the clients
are completely unaware or knowledgeable about when or what s ollected. W e ound hat while t akes more
overhead o nitiate he etrieval of email raffic rom mail erver, here s urprisingly ittle overhead n he
part f erver dministrator o ctually ssist he
research ffort n athering ata. ince og iles re
typically tored in ommon location on the erver, the
administrator need only make these files available. When
operated cross network, he/she an asily op y these log iles o ommon ocation rom where he erver-
based data collection program ca n import the data.
By resenting wo ethods or athering nd
analyzing email data, we have shown both advantages an d disadvantages or he ocial etwork nalyst. hese
strengths an d limitations must be considered by an y social network analyst when studying email traffic. Even though
gathering ata t ts ource oes rovide etter ata
integrity, uch ata ollection means re ot lways
feasible. n hese ases, mail ata ollected n decentralized manner ca n still provide insightful analysis of the underlying social network.
8/13/2019 Unobtrusive Social Network Data Through Email
http://slidepdf.com/reader/full/unobtrusive-social-network-data-through-email 7/7
We advise a practitioner to be highly sensitive to the
privacy mplications f his rocess, specially n he
public nd rivate ectors. eople within he military
typically o ot maintain he xpectation f mail nd
internet rivacy. his ay ot e rue n ther
populations. are ust lso e xercised ith
interpreting he esults of these ypes of social networks.
It is important that trained social network analysts provide
proper interpretation of the organizational behavior, while
respecting the privacy of individual dentities. evealing
the position an individual maintains n the social network
of an organization may lead to an overall decrease in trust
and dversely ffect he eadership limate within he
organization. hen sed roperly, owever, ocial
network nalysis an rovide ealth f aluable
information o he rganization. everal ommands
within he rmy ave lready mplemented ocial
network data collection from email. hese methods have
been utomated n he oftware package ORA, which s
maintained by CASOS at Carnegie Mellon University and
can be freely downloaded by the military and academia.
Future esearch n his rea will ikely xplore he
impact of cellular phone communication and blackberries
on ocial etworks within he military. his ine f
research will urther upport he fficacy f Netcentric
Operations within he Army. ocused esearch nto he
usage f cell phones, lackberries, -mail, nd ace-to-
face ommunication uring major roup ctivities will
provide greater insight into social network data collection.
Understanding the desired channels of communication for
military eaders, may ignificantly ontribute o haping
the communication echnologies hat he DoD nvests n.
This ine of research may also provide data or real-time monitoring of organizational change. t will certainly be
valuable in enhancing command and control systems used
by the military.
ACKNOWLEDGEMENTS This research is part of the IkeNet project in the U.S. Military Academy
Network Science Center an d the Dynamics Networks project in CASOS
(Center for Computational Analysis of Social an d Organizational Systems, http://www.casos.cs.cmu.edu) at Carnegie Mellon University. This work wa s supported in part by:
• Th e Army Research Institute for th e Behavioral an d Social Sciences, Army Project No . 611102B74F/ Grant No . CM U - W91WAW07C0063
• Th e Army Research Labs Grant No . DAAD 19-01-2-0009
• Th e views an d conclusions contained in this document are those of the authors an d should not be interpreted as representing the official policies, either expressed or implied, of the National Science Foundation or th e U.S. government.
Carley, K.M. 2002) Smart Agents and Organizations of the
Future. The Handbook ofNew Media. Edited by Leah
Lievrouw an d Sonia Livingstone, Ch . 2, pp. 206-220,
Thousand Oaks, CA , Sage.
Carley, .M., olumbus, ., eReno, M., eminga, . nd
Moon, I. 2008). OR A User's Guide 2008. Carnegie Mellon
University, School of Computer Science, Institute or
Software Research,
echnical Report, CMU-ISR-08-125.
Carvalho, V.R.,, Wu, W. & Cohen W. W., (2007). Discovering
Leadership Roles in Email Workgroups. In Proceedings,
CEAS 2 7 Fourth Conference on Email andAnti-Spam,
Au g 2-3, 2007 Mountain View, CA , www.ceas.ee.
Diesner, ., Frantz, T. , Carley, K. M. 2005) Communication
Networks rom he Enron Email Corpus t's Always About
the People. Enron is no Different. Computational and
Mathematical Organization Theory, 1, 20 1 - 228.
Ducheneaut, N; Bellotti, V. (2001). Email as habitat: An
exploration of embedded personal information management.
ACM Interactions. Sep-Oct, 30-38.
Frantz, T. , Carley, K.M. 2008a) Transforming raw-email data
into ocial-network nformation. n hristopher . ang,
Hsinchun Chen, Michael Chau, Kuiyu Chang, heau-Dong
Lang, Patrick . Chen, Raymond Hsieh, Daniel Zeng, Fei- Yu e Wang, Kathleen Carley, Wenji Mao, nd ustin Zhan
(Eds.). Intelligence nd ecurity nformatics Workshops,
PAISI, PACCF and SOCO 2008 Springer, Lecture Notes in
Computer cience, o. 5075. acific Asia Workshop n
Intelligence an d Security Informatics (PAISI 2008).
Frantz, T. & Carley, K.M. (2008b). CEMAPII: An Architecture
and Specifications to Facilitate the Importing ofReal-World
Data nto he ASOS oftware uite. Carnegie Mellon
University Technical Report (ISR-08-130)
Gloor, P. & Zhao, Y. (2004). TeCFlow - A Temporal
Communication Flow Visualizer fo r Social
Networks Analysis. In Proceedings, ACMCSCW
Workshop on Social Networks, N ov 6, Chicago, IL ,
McCulloh, I.A., Carley, K.M. (2008). Social Network Change
Detection. Carnegie Mellon University, Technical Report, CMU-CS-08-116.
McCulloh, I.A., Garcia, G, MacGibbon, J. , Tardieu, K, Dye,
H. , Moores, K, Graham, J. (2007). IkeNet: Social Network Analysis ofe-mail Traffic in the Eisenhower Leadership
Development Program. rm y Research Institute Technical
Report 1218.
Tyler, J. R. , Wilkinson, D. M., Huberman, B. A. (2003). mail
as Spectroscopy: Automated Discovery of Community
Structure within Organizations. ommunities and
Technologies, 81-96.
Viegas, F.B., Golder, S. , & Donath, J. (2006). Visualizing email
content: portraying relationships from conversational
histories. In Proceedings of the SIGCHI conference on
Human Factors in computing systems. Montreal, Quebec,
Canada, 979-988.
REFERENCES
Carley, K.M. 2003). Dynamic Network Analysis. n Dynamic
Social Network Modeling and Analysis: Workshop Summary and Papers, Ronald Breiger, Kathleen Carley, nd Philippa
Pattison, Eds.) ommittee n uman actors, ational
Research Council, National Research Council. Pp. 33-145.