SOCIALMEDIA.ORG/SUMMIT2014 ORLANDO OCTOBER 27–29, 2014 Measuring social media as a complex, adaptive system GERALD KANE BOSTON COLLEGE AND MIT-SLOAN MANAGEMENT REVIEW
Jul 15, 2015
SOCIALMEDIA.ORG/SUMMIT2014ORLANDOOCTOBER 27–29, 2014
Measuring social media asa complex, adaptive system
GERALD KANEBOSTON COLLEGE AND MIT-SLOANMANAGEMENT REVIEW
Measuring Social Media as a Complex Adaptive System
Gerald C. Kane, @pro0ane Associate Professor Guest Editor, Social Business Boston College MIT-‐Sloan Management Review [email protected] [email protected]
The opportunity… • Social media allows interacGons on size and scope not previously possible.
• “Digital trace” allows unprecedented opportuniGes to measure and analyze these behaviors.
• It’s what got me interested in SM – Facebook & Wikipedia.
…The problem The resulGng interacGons are oSen complex…. 1. Non-‐linear 2. Co-‐evoluGon 3. Self-‐organizaGon 4. Emergent dynamics …which can create problems for measuring them effecGvely.
The “canary in a coal mine.” • Founded in 2001
• 4.3M English arGcles • 6th most heavily trafficked website.
• 15 years of excellent data for studying how people collaborate online. • Can learn much about implicaGons for measurement in social media.
Insight #1: Non-‐linear (Ransbotham and Kane 2011, MISQ)
• More may be be`er, but only to a certain point. • How does membership turnover affect arGcle development?
• Most online community research usually assumes membership retenGon is posiGve (e.g. Ma and Agarwal 2007, Butler 2001).
• Yet, research on organizaGonal turnover suggests that some moderate amount of turnover is beneficial (e.g. March 1991).
• Study: 2065 Featured ArGcles between 2001-‐2009 (3M revisions, 186 GB data)
• Findings: Moderate turnover beneficial in online communiGes for both likelihood creaGng and retaining knowledge.
0
2
4
6
8
10
12
-‐2 -‐1.5 -‐1 -‐0.5 0 0.5 1 1.5 2 2.5
Per
cent
age
Cha
nge
in P
ropo
rtio
nal R
isk
Average Experience (standardized)
Promotion Demotion
Impact of Social Business on companies. (Kane et al. 2014, MIT-‐SMR)
0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
70.00%
80.00%
90.00%
100.00%
1 2 3 4 5 6 7 8 9 10
1) PosiGvely impact business outcomes 2) SB IniGaGves meet objecGves
Takeaway #1 Avoid Oversimplifying: Understanding and managing social media is rarely as simple as you think it is or want it to be.
Insight #2: Co-‐evolution (Ransbotham, Kane, Lurie 2012, Marketing Science)
• Changes in one part of the plamorm can affect outcomes in others. • The proverbial bu`erfly flapping its wings.
• Does turnover have effect beyond focal arGcle? Is “quality” contagious? • Collaborators may join new communiGes when leave old ones, transferring knowledge from one community to another.
• Like a bee pollinaGng flowers, contributors can spread knowledge from one community to another.
• Study: 40K contributors to 16K medical arGcles on Wikipedia 2001-‐2009 (2M revisions, 50GB data). • Created 2-‐mode affiliaGon network of arGcles and shared collaborators.
• Finding: Centrality in both local (degree) and global (closeness) centrality predicts quality and popularity of content. • Online collaboraGon may involve mulGple interdependent communiGes.
Squares = authors Circles = articles
Red = Featured Articles Orange = A-quality Articles Yellow = Good Articles
Light Blue = B-quality Articles Dark Blue = Start-quality articles
Results
Takeaway #2 Small, unexpected changes in one part of the social media environment can o@en have a big impact on another.
Insight #3: Emergence (Kane, Johnson, Majchrzak, Management Science)
• Can order evolve without any management intervenGon or formal leadership structure?
• Study: In-‐depth case study of 8K edits from 3K contributors to AuGsm arGcle from 2001 – 2010. • One of handful of arGcles that promoted to, demoted from, and re-‐promoted to featured arGcle status.
• Among most heavily visited arGcles, recognized by outside sources for quality.
• Finding: CommuniGes are both structured AND emergent, depending on the stage of development. • DeliberaGon types occurred in ways similar to soSware development lifecycle, despite li`le formal coordinaGon mechanism.
• Knowledge arGfact served as coordinaGng mechanism
0
2
4
6
8
10
12
14
16
18
20
Qtr2
Qtr3
Qtr4
Qtr1
Qtr2
Qtr3
Qtr4
Qtr1
Qtr2
Qtr3
Qtr4
Qtr1
Qtr2
Qtr3
Qtr4
Qtr1
Qtr2
Qtr3
Qtr4
Qtr1
Qtr2
Qtr3
Qtr4
Qtr1
Qtr2
Qtr3
Qtr4
2004 2005 2006 2007 2008 2009 2010
Num
ber o
f Delibera.
ons
Sum of ChaoGc GeneraGng
Sum of Joint Shaping
Sum of Defensive Filtering
Duration of Contribution for Top 20 Contributors (contributors rank ordered by activity level)
!
!
2/13/2004' 5/13/2006' 8/10/2008' 11/8/2010'
E'S'M'W'R'T'B'
RN'C'N'AC'D'
CO'TD'AT'CE'L'
SO'Z'Q''
Top$20
$Par*cipan
ts$
Dura=on'of'Par=cipa=on'
Highest Activity
Moderate Activity
#!of!Edits!
!Promotion(#1(((Demotion(((Promotion(#2(!
Insight #4: Dynamics (Kane and Ransbotham, ICIS 2013)
• Feedback loops can develop, where two characterisGcs can mutually reinforce one another.
• Dynamics of online peer producGon have never been tested. • Presumed that content leads to viewers, more viewers lead to be`er content.
• Does this dynamic exist, does it change over Gme? • Study: Same sample of 16K medical arGcles, 40K contributors used earlier. • 3SLS regression, using “protected” as the idenGficaGon variable (i.e. affects contribuGons but not viewership).
• Findings: We find evidence for hypothesized collaboraGon dynamics, but a`enuates over Gme. • Age of an arGcle is posiGvely related to viewership, but negaGvely related to contribuGon acGvity.
• Anonymity improves both contribuGon and viewership.
Implications for Organizations • CombinaGon of qualitaGve and quanGtaGve data is powerful.
• Embrace paradox between leading and following – including customer communiGes.
• Provide Gme for employees to learn new ways of working – internal use for external experience.
• Its not mainly about the technology – culture is key.
• Look for leadership examples outside business (e.g. military, non-‐profits).
To Conclude… • InteracGons on social media exhibit characterisGcs of complex adapGve systems • Non-‐linear • Co-‐evoluGon • Self-‐organizaGon • Emergent dynamics
• If we do not account for these complex features, we risk making mistakes in our analysis and interpretaGon of our data. • “With great data, comes great responsibility.”