CSCL 2011 | Keynote	Augmented Social Cognition: How SocialComputing is Changing eLearning              	Ed H. Chi		GoogleR...
2008-05-13   CSCL 2011 Keynote
Prelude:	  A	  personal	  learning	  story	  To:	  chi@acm.org	  From:	  Brad	  Barrish	  <brad@…removed.for.privacy….com>...
Talk	  in	  3	  Acts	       The	  Importance	  of	  Social	  Signals	  in	  eLearning	                n    Act	  I:	  The...
Act	  I:	  	  Invisible	  Social	  Signals	  from	  the	  Crowd	    Joint	  work	  w/	  Todd	  Mytkowicz,	  Rowan	  Nairn,...
Using	  Information	  Theory	  to	  Model	  Social	  Tagging	  [Ed	  H.	  Chi,	  Todd	  Mytkowicz,	  ACM	  Hypertext	  200...
Tagging	  Behavior	  H(Tag)	  shows	  tag	  saturation	       H(Doc	  |	  Tag),	  browsability	      2008-05-13           ...
Implication	  I(Doc;	  Tag)	  	  Mutual	  Information	            Raise	  in	  avg.	  tag	  /	  bookmark	        2008-05-1...
TagSearch:	  MapReduce	  Implementation	                                 Tags                      URLs                   ...
TagSearch:	  Use	  Semantic	  Analysis	  to	   Reduce	  Noise	  	  	  	  	  http://mrtaggy.com	  	  Semantic Similarity Gr...
2008-05-13   CSCL 2011 Keynote
Experiment	  Design	  	  [Kammerer	  et	  al.	  CHI2009]	  n     2	  interface	  x	  3	  task	  domain	  design	         ...
Evauation	  Results	  [Kammerer	  et	  al.,	  CHI2009]	  n    Exploratory	  interface	  users:	         –    performed	  ...
Act	  II:	  Visible	  Social	  Signals	  from	  	  Shared	  Highlighting	            Kudos	  to	  Lichan	  Hong,	  Les	  N...
Finding	  a	  Restaurant	  n    Appropriate	  for	        the	  occasion	   2008-05-13                    CSCL 2011 Keynote
Heuristics	                   Poor heuristic                               Good heuristic  2008-05-13           CSCL 2011 ...
“Hints”	                Solo              Cooperative (“good hints”) 2008-05-13    CSCL 2011 Keynote
SparTag.us:	  Social	  Highlighting	     2008-05-13           CSCL 2011 Keynote
SparTag.us:	  Social	  Highlighting	    n    In	  situ	  tagging	  while	  reading	    n    Highlighting	    n    Share...
Highlighting	  as	  Importance	  	  Indicator	  recall                          first-visit                               ...
Evaluation	  Task	  &	  Metric	  [Nelson	  et	  al.,	  HCII2009]	  n     Sensemaking	  task	          –  Find	  and	  rea...
Procedure	                              SparTag.us	                       SF    with	   Friend 	                          ...
Results:	  Learning	  Gain	  N=18	  	  SparTag.us	  +	  Friend	  superior	  to	  both	  individual	  conditions	  No	  diff...
URL	  Kind	       Code	                                                Blog	              B	  Observation	                ...
Von	  Restorff	  Isolation	  Effect	  [1933]	  n    As	  applied	  to	  highlights,	  the	  von	  Restorff	  isolation	  effe...
Act	  III:	  	  Abstracted	  Knowledge:	  The	  Science	  of	  Understanding	  Wikipedia	    Kudos	  to	  Bongwon	  Suh,	 ...
Exponential	  Growth	  of	  Wikipedia:	  an	  accepted	  ‘fact’	                            Number of Articles (Log Scale)...
Growth	  of	  Edits	   2008-05-13                 CSCL 2011 Keynote
Something	  happened	  in	  early	  2007	   2008-05-13            CSCL 2011 Keynote
Growth	  of	  Active	  Editors	  *In thousands       2008-05-13            CSCL 2011 Keynote
Slowing	  Growth	  in	  Global	  Activity	  *In thousands       2008-05-13            CSCL 2011 Keynote
Earlier	  Exponential	  Growth	  Model	      n     Preferential	  Attachment:	  Edits	  beget	  edits	              –  mo...
Logistic	  Growth	  Model	  n     Ecological	  population	  growth	  model	          –  Also	  depend	  on	  environmenta...
Match	  to	  Data:	  #	  of	  New	  Articles	  n    Follows	  a	  logistic	  growth	  curve	                             ...
Struggle	  for	  Existence	  -­‐	  Darwin	  n    Biological	  system	         –  Competition	  increases	  as	           ...
“Showering”	  Hypothesis	  What	  drives	  contributions	  to	  Wikipedia?	  Cooperation	  is	  not	  the	  main	  driver?...
Conflict/Coordination	  Effects	  in	  Wikipedia	                                     (Kittur, Suh, Pendleton, Chi, CHI2007)...
Ratio	  of	  Reverted	  Contributions	  	                Monthly Ratio of Reverted Edits 2008-05-13                    CSC...
Visual	  Analytics	  over	  Wikipedia	  data	  Mediator	  Pattern	  -­‐	  Terri	  Schiavo 	            	  [Suh,	  et	  al....
WikiDashboard.com	  2008-05-13       CSCL 2011 Keynote
Coda:	  A	  Challenge:	  A	  modified	  logistic	  model	  n    Carrying	  Capacity	  as	  a	  function	  of	  time.	   20...
What	  Did	  We	  Learn?	  n    The	  Common	  Thread:	         –  Utilization	  of	  Social	  Signals	  for	  Learning	 ...
Research	  Vision	  Augmented	  Social	  Cognition	  n    Cognition:	  the	  ability	  to	  remember,	  think,	  and	  re...
From	  Rote	  Learning	  to	  Interaction	   2008-05-13               CSCL 2011 Keynote
Thank	  you!	                n    chi@acm.org	                n    http://edchi.net	   2008-05-13                       ...
What	  I	  will	  not	  talk	  about	  …	  n    Motivation	         –  Cultural	  and	  economic	  incentives	         – ...
2008-05-13   CSCL 2011 Keynote
Lowering	  Participation	  /	  Interaction	  Costs	  n    Interaction	  costs	                                           ...
Using	  Machine	  Learning	  to	  Detect	  Conflicts	  n                          Counting	  ‘Controversial’	  labels	  n...
Collaborative	  Knowledge	  Building	  n    “They	  cannot	  even	  begin	  to	  coordinate	  on	  content	        withou...
Google	  Plus	  as	  a	  Research	  Platform	   2008-05-13              CSCL 2011 Keynote
Upcoming SlideShare
Loading in...5
×

CSCL 2011 Keynote on Social Computing and eLearning

2,704

Published on

Ed H. Chi
Google Research (Work done at Xerox PARC)

CSCL2011 Keynote Abstract:

Our research in Augmented Social Cognition is aimed at enhancing the ability of a group of people to remember, think, and reason. Our approach to creating this augmentation or enhancement is primarily model-driven. Our system developments are informed by models such as information scent, sensemaking, information theory, probabilistic models, and more recently, evolutionary dynamic models. These models have been used to understand a wide variety of user behaviors, from individuals interacting with social bookmark search in Delicious and MrTaggy.com to groups of people working on articles in Wikipedia. These models range in complexity from a simple set of assumptions to complex equations describing human and group behaviors.

Indeed, increasingly, new social online resources such as social bookmarking sites and Wikis are becoming central in eLearning. By studying them, we further our understanding of how knowledge is constructed in a social context. In this talk, I will illustrate how a model-driven approach could help illuminate the path forward for social computing and social learning.

-----

Published in: Technology, Business

CSCL 2011 Keynote on Social Computing and eLearning

  1. 1. CSCL 2011 | Keynote Augmented Social Cognition: How SocialComputing is Changing eLearning Ed H. Chi GoogleResearch Work done whileat Palo AltoResearch Center(PARC) 2008-05-13 CSCL 2011 Keynote
  2. 2. 2008-05-13 CSCL 2011 Keynote
  3. 3. Prelude:  A  personal  learning  story  To:  chi@acm.org  From:  Brad  Barrish  <brad@…removed.for.privacy….com>  Subject:  Pancreatic  cancer  Date:  Thu,  1  Feb  2007  21:37:55  PST    Hey  Ed.  Im  a  fellow  del.icio.us  user  and  noticed  you  bookmark  a  lot      of  pancreatic  cancer  stuff.  Im  at  home  with  my  dad  who  was  diagnosed      a  little  over  a  year  ago  and  is  now  at  the  tale  end  of  things.  Ive      learned  a  lot  through  his  treatments  and  about  whats  out  there.  I      dunno  if  its  something  you  or  a  family  member  has,  but  just  wanted      to  drop  you  an  email.  Be  well.    Brad   2008-05-13 CSCL 2011 Keynote
  4. 4. Talk  in  3  Acts   The  Importance  of  Social  Signals  in  eLearning   n  Act  I:  The  Invisible   –  Social  Search   n  Act  II:  The  Visible   –  Shared  Annotations   n  Act  III:  The  Abstracted   –  Shared  Knowledge  Space   2008-05-13 CSCL 2011 Keynote
  5. 5. Act  I:    Invisible  Social  Signals  from  the  Crowd   Joint  work  w/  Todd  Mytkowicz,  Rowan  Nairn,  Lawrence  Lee     [Chi  and  Mytkowicz,  Hypertext2008]   [Kammerer  et  al.,  CHI2009]     2008-05-13 CSCL 2011 Keynote
  6. 6. Using  Information  Theory  to  Model  Social  Tagging  [Ed  H.  Chi,  Todd  Mytkowicz,  ACM  Hypertext  2008]   Concepts   Topics   Users   Documents   Noise   Tags   Decoding   Encoding   T1…Tn   2008-05-13 CSCL 2011 Keynote
  7. 7. Tagging  Behavior  H(Tag)  shows  tag  saturation   H(Doc  |  Tag),  browsability   2008-05-13 CSCL 2011 Keynote
  8. 8. Implication  I(Doc;  Tag)    Mutual  Information   Raise  in  avg.  tag  /  bookmark   2008-05-13 CSCL 2011 Keynote
  9. 9. TagSearch:  MapReduce  Implementation   Tags URLs P(URL|Tag) P(Tag|URL) n  Spreading  Activation  in  a  bi-­‐graph   n  Computation  over  a  very  large  data  set   –  150  Million+  bookmarks   2008-05-13 CSCL 2011 Keynote
  10. 10. TagSearch:  Use  Semantic  Analysis  to   Reduce  Noise          http://mrtaggy.com    Semantic Similarity Graph Web Tools Reference Guide Howto Tutorial Tips Help Tip Tutorials Tricks 2008-05-13 CSCL 2011 Keynote
  11. 11. 2008-05-13 CSCL 2011 Keynote
  12. 12. Experiment  Design    [Kammerer  et  al.  CHI2009]  n  2  interface  x  3  task  domain  design   –  2  Interface  (between-­‐subjects)   n  Exploratory  vs.  Baseline   –  3  task  domains  (within-­‐subjects)   n  Future  Architecture,  Global  Warming,  Web  Mashups  n  30  Subjects  (22  male,  8  female)   –  Intermediate  or  advanced  computer  and  web  search  skills   –  Half  assigned  Exploratory,  half  Baseline.  n  For  each  domain,  single  block  with  3  task  types:   –  Easy  and  Difficult  Page  Collection  Task  [6min  each]   –  Summarization  Task  [12min]   –  Keyword  Generation  Task  [2min]   2008-05-13 CSCL 2011 Keynote
  13. 13. Evauation  Results  [Kammerer  et  al.,  CHI2009]  n  Exploratory  interface  users:   –  performed  more  queries,     –  took  more  time,     –  wrote  better  summaries  (in  2/3  domains),     –  generated  more  relevant  keywords  (in  2/3  domains),  and   –  had  a  higher  cognitive  load.  n  Suggestive  of  deeper  engagement  and  better  learning.  n  Some  evidence  of  scaffolding  for  novices  in  the  keyword   generation  and  summarization  tasks.   2008-05-13 CSCL 2011 Keynote
  14. 14. Act  II:  Visible  Social  Signals  from    Shared  Highlighting   Kudos  to  Lichan  Hong,  Les  Nelson       [Hong  et  al,  AVI2008]     [Nelson  et  al.,  HCII  2009]   2008-05-13 CSCL 2011 Keynote
  15. 15. Finding  a  Restaurant  n  Appropriate  for   the  occasion   2008-05-13 CSCL 2011 Keynote
  16. 16. Heuristics   Poor heuristic Good heuristic 2008-05-13 CSCL 2011 Keynote
  17. 17. “Hints”   Solo Cooperative (“good hints”) 2008-05-13 CSCL 2011 Keynote
  18. 18. SparTag.us:  Social  Highlighting   2008-05-13 CSCL 2011 Keynote
  19. 19. SparTag.us:  Social  Highlighting   n  In  situ  tagging  while  reading   n  Highlighting   n  Shared  notebooking     n  Sharing!   2008-05-13 CSCL 2011 Keynote
  20. 20. Highlighting  as  Importance    Indicator  recall first-visit 2008-05-13 CSCL 2011 Keynote
  21. 21. Evaluation  Task  &  Metric  [Nelson  et  al.,  HCII2009]  n  Sensemaking  task   –  Find  and  read  material  about   Enterprise  2.0  mashups  in  order  to   write  two  essays  n  Seeds:   expert  content  for  scaffolding   –  Tags  from  del.icio.us   –  URLs  from  Google/PageRank   –  Constructed  and  then  shared  through  social  mechanisms  (i.e.,  a   SparTag.us   friend )  n  Performance  Measures   –  Learning  gain:  Pre/Post  Knowledge  Test   Posttest score - Pretest score Gain = Max score - Pretest score 2008-05-13 CSCL 2011 Keynote
  22. 22. Procedure   SparTag.us   SF with   Friend   SparTag.us  Demographics   SO Only   Posttest   &  Pretest   Without   WS SparTag.us   2008-05-13 CSCL 2011 Keynote
  23. 23. Results:  Learning  Gain  N=18    SparTag.us  +  Friend  superior  to  both  individual  conditions  No  difference  between  the  two  control  conditions   2008-05-13 CSCL 2011 Keynote
  24. 24. URL  Kind   Code   Blog   B  Observation   Conference   C   Employment   E   My.Spartag.us   M   News   N   URL KindOpenSource   Code O   Blog Search  B S   Conference C Vendor   V   Employment E Wikipedia   MySpartagus M W   News Consultant   N X   OpenSource O Search S Vendor V Wikipedia W Consultant X 2008-05-13 CSCL 2011 Keynote
  25. 25. Von  Restorff  Isolation  Effect  [1933]  n  As  applied  to  highlights,  the  von  Restorff  isolation  effect   suggests  that  readers:  n  (a)  tend  to  focus  on  and    n  (b)  learn  what  is  marked,    n  whether  the  information  is  important  or  not.   –  Nist  and  Hogrebe  87   2008-05-13 CSCL 2011 Keynote
  26. 26. Act  III:    Abstracted  Knowledge:  The  Science  of  Understanding  Wikipedia   Kudos  to  Bongwon  Suh,  Niki  Kittur     [Kittur  et  al.,  CHI2007]   [Suh  et  al.,  WikiSym  2009]     2008-05-13 CSCL 2011 Keynote
  27. 27. Exponential  Growth  of  Wikipedia:  an  accepted  ‘fact’   Number of Articles (Log Scale) http://en.wikipedia.org/wiki/Wikipedia:Modelling_Wikipedia’s_growth 2008-05-13 CSCL 2011 Keynote
  28. 28. Growth  of  Edits   2008-05-13 CSCL 2011 Keynote
  29. 29. Something  happened  in  early  2007   2008-05-13 CSCL 2011 Keynote
  30. 30. Growth  of  Active  Editors  *In thousands 2008-05-13 CSCL 2011 Keynote
  31. 31. Slowing  Growth  in  Global  Activity  *In thousands 2008-05-13 CSCL 2011 Keynote
  32. 32. Earlier  Exponential  Growth  Model   n  Preferential  Attachment:  Edits  beget  edits   –  more  number  of  previous  edits,  more  number  of  new  edits   Growth rate depends on: N = current population r = growth rate of the population N(t) = N 0 " e rt dN = r" N dt Growth rate Current of population ! population! 2008-05-13 CSCL 2011 Keynote
  33. 33. Logistic  Growth  Model  n  Ecological  population  growth  model   –  Also  depend  on  environmental  conditions   –  K,  carrying  capacity  (due  to  resource  limitation)   dN N = rN(1" ) dt K 2008-05-13 CSCL 2011 Keynote
  34. 34. Match  to  Data:  #  of  New  Articles  n  Follows  a  logistic  growth  curve   New Article 2008-05-13 CSCL 2011 Keynote
  35. 35. Struggle  for  Existence  -­‐  Darwin  n  Biological  system   –  Competition  increases  as   population  hit  the  limits  of  the   ecology   –  Advantage  go  to  members  of  the   population  that  have  competitive   dominance  over  others  n  Analogy   –  Limited  opportunities  to  make   novel  contributions   –  Increased  patterns  of  conflict  and   dominance     2008-05-13 CSCL 2011 Keynote
  36. 36. “Showering”  Hypothesis  What  drives  contributions  to  Wikipedia?  Cooperation  is  not  the  main  driver?  n  Hypothesis:  Conflicts  drives  most  of  the  contributions.   –  How  do  we  measure  conflicts?  n  Conflicts  cause  coordination  costs  to  go  up.   –  How  to  measure  coordination  costs?  n  “negotiation  is  critical  to  helping  multiple  perspectives   to  converge  on  shared  knowledge.”     –  Stahl,  Group  Cognition,  Ch8,  2004   2008-05-13 CSCL 2011 Keynote
  37. 37. Conflict/Coordination  Effects  in  Wikipedia   (Kittur, Suh, Pendleton, Chi, CHI2007) 2008-05-13 CSCL 2011 Keynote
  38. 38. Ratio  of  Reverted  Contributions     Monthly Ratio of Reverted Edits 2008-05-13 CSCL 2011 Keynote
  39. 39. Visual  Analytics  over  Wikipedia  data  Mediator  Pattern  -­‐  Terri  Schiavo    [Suh,  et  al.,  VAST2007]   Anonymous (vandals/ spammers) Sympathetic to husband Mediators Sympathetic to parents 2008-05-13 CSCL 2011 Keynote
  40. 40. WikiDashboard.com  2008-05-13 CSCL 2011 Keynote
  41. 41. Coda:  A  Challenge:  A  modified  logistic  model  n  Carrying  Capacity  as  a  function  of  time.   2008-05-13 CSCL 2011 Keynote
  42. 42. What  Did  We  Learn?  n  The  Common  Thread:   –  Utilization  of  Social  Signals  for  Learning  and  Information  Access   –  Whether  it  is  invisible,  visible,  and  abstracted.  n  The  Establishment  of  Common  Ground   –  Implicit  Coordination   –  Explicit  Coordination   –  Negotiation  n  “All  collective  actions  are  built  on  common  ground  and   its  accumulation.”   –  Clark  and  Brennan,  1991   2008-05-13 CSCL 2011 Keynote
  43. 43. Research  Vision  Augmented  Social  Cognition  n  Cognition:  the  ability  to  remember,  think,  and  reason;  the  faculty  of   knowing.  n  Social  Cognition:  the  ability  of  a  group  to  remember,  think,  and   reason;  the  construction  of  knowledge  structures  by  a  group.   –  (not  quite  the  same  as  in  the  branch  of  psychology  that  studies  the   cognitive  processes  involved  in  social  interaction,  though  included)  n  Augmented  Social  Cognition:  Supported  by  systems,  the   enhancement    of  the  ability  of  a  group  to  remember,  think,  and   reason;  the  system-­‐supported  construction  of  knowledge   structures  by  a  group.    Citation:  Chi,  IEEE  Computer,  Sept  2008   2008-05-13 CSCL 2011 Keynote
  44. 44. From  Rote  Learning  to  Interaction   2008-05-13 CSCL 2011 Keynote
  45. 45. Thank  you!   n  chi@acm.org   n  http://edchi.net   2008-05-13 CSCL 2011 Keynote
  46. 46. What  I  will  not  talk  about  …  n  Motivation   –  Cultural  and  economic  incentives   –  Personal  and  societal  values   –  Psychology  (e.g.  cognitive,  personality,  social)  n  Policy  and  Investment   –  Resources   –  Teacher  training   –  Technological  investment  n  With  the  Assumption  of  Motivation  and  Resources,   how  to  make  information  universally  accessible  and   useful  in  a  Web2.0  world?   2008-05-13 CSCL 2011 Keynote
  47. 47. 2008-05-13 CSCL 2011 Keynote
  48. 48. Lowering  Participation  /  Interaction  Costs  n  Interaction  costs   # People willing to produce for “free” determine  number  of   people  who  participate  n  Surplus  of  attention  &   motivation  at  small   transaction  costs  n  Therefore…  n  Important  to  keep   interaction  costs  low   Cost of participation 2008-05-13 CSCL 2011 Keynote
  49. 49. Using  Machine  Learning  to  Detect  Conflicts  n  Counting  ‘Controversial’  labels  n  5x  cross-­‐validation,  R2  =  0.897   10000 9000 Actual controversial revisions 8000 7000 6000 5000 4000 3000 2000 1000 0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 Predicted controversial revisions 2008-05-13 CSCL 2011 Keynote
  50. 50. Collaborative  Knowledge  Building  n  “They  cannot  even  begin  to  coordinate  on  content   without  assuming  a  vast  amount  of  shared  information   or  common  ground….  And  to  coordinate  on  process,   they  need  to  update  their  common  ground  moment  by   moment.  All  collective  actions  are  built  on  common   ground  and  its  accumulation.”   –  Clark  and  Brennan,  1991  n  At  Web-­‐scale  social  learning,  what  we  know  about  the   nature  of  conflict  and  negotiation  is  woefully   inadequate.   2008-05-13 CSCL 2011 Keynote
  51. 51. Google  Plus  as  a  Research  Platform   2008-05-13 CSCL 2011 Keynote
  1. A particular slide catching your eye?

    Clipping is a handy way to collect important slides you want to go back to later.

×