Transcript
Page 1: Analysis of interaction in collaborative activities; the Synergo approach

1/60

Analysis of interaction in collaborative activities:

the Synergo trail

Nikolaos AvourisUniversity of Patras, GR

Keynote Talk

INCoS 2010 – Thessaloniki November 24th

Page 2: Analysis of interaction in collaborative activities; the Synergo approach

2/60

outline- on analysis of collaboration- the synergo testbed- synergo studies- models from synergo data

Page 3: Analysis of interaction in collaborative activities; the Synergo approach

3/60

On analysis of collaborative

activities

Page 4: Analysis of interaction in collaborative activities; the Synergo approach

4/60

Typical analysis objectivesfocusmethod

Participant’s perceptionsInquiry methods

Interaction processQuantitative, qualitative, sequential methods

Learning outcomesPre-post testing

Collaborative technologyUsability evaluation

Page 5: Analysis of interaction in collaborative activities; the Synergo approach

5/60

Focus on the interaction process

– Dillenbourg: “the basic instrument for understanding collaborative learning is understanding the interaction that takes place during a learning process”

– Koschmann: “CSCL research is not focused on instructional efficacy, but it is studying instruction as enacted practice”

Page 6: Analysis of interaction in collaborative activities; the Synergo approach

6/60

Quantitative analysis• Frequency counts of events such as:

- messages posted per student per period of time- hits on particular discussion forum pages - actions taken on objects of a shared workspace- number of files read in a shared file system etc.

• Defining metrics (indicators) that combine different kinds of frequency counts

• Suitable for all kinds of collaborative learning• They can lead to models of interaction (e.g.

Social Networks etc.)

Page 7: Analysis of interaction in collaborative activities; the Synergo approach

7/60

Qualitative content analysis• “Content analysis refers to any process

that is a systematic replicable technique for compressing many words of text into fewer content categories based on explicit rules of coding” (Kripendorf, 1980)

• Suitable for every means of dialogue oriented collaborative learning (synchronous & asynchronous, collocated & distant)

Page 8: Analysis of interaction in collaborative activities; the Synergo approach

8/60

Content analysis models• Henri’s scheme• Garrison’s model• Gunawardena’s Interaction

Analysis Model• Language/action OCAF

Page 9: Analysis of interaction in collaborative activities; the Synergo approach

9/60

Content analysis resources

• The content analysis guidebook http://academic.csuohio.edu/kneuendorf/content/

Page 10: Analysis of interaction in collaborative activities; the Synergo approach

10/60

Small group synchronous interaction: Integration of dialogue and action• Treats language acts and actions taken to objects

in an integrated way• Uniform annotation (eg. the OCAF framework)• Shifts the focus to the objects of a shared

workspace• Objects have an ‘owner’ just like language acts• Can visualize uptaking actions (Suthers 05)

Page 11: Analysis of interaction in collaborative activities; the Synergo approach

11/60

Dialogue: Chat tool affordances• Visual and/or auditory cues are not available• No production blocking->overlapping exchanges• Persistence of messages – substantiation of conversation• Loose inter-turn connectedness - but possibility of

simultaneous engagement in multiple threads• Verbal deixis spans throughout the whole history of

dialogue (no restricted time window is adequate for analysis)

• Posters may reply rapidly, using short messages and split long messages to increase referent/message coherency (Garcia and Jacobs 1999)

• Participants begin new topics fairly much at will in a manner that would not happen in a formal face-to-face group discussion (O’Neil & Martin, 2003)

Page 12: Analysis of interaction in collaborative activities; the Synergo approach

12/60

Action: Shared Activity spaceaffordances• Feedthrough (Dix et. al., 1993)• Various degrees of coupling (Salvador

et. al., 1996)• Workspace can be used as an external

representation of the task that allows efficient nonverbal communication

• Workspace artefacts act as conversational props (Hutchins, 1990)

Page 13: Analysis of interaction in collaborative activities; the Synergo approach

13/60

Types of communication acts / gestures in shared workspace

• Deictic references• Demonstrations • Manifesting actions• Visual evidence (Gutwin, Greenberg, 2002)

Page 14: Analysis of interaction in collaborative activities; the Synergo approach

14/60

Grounding through actions on a workspace representation (Suthers, 2006)Sequences of actions :(1) one participant’s action in a

medium…(2) is taken up by another participant

in a manner that indicates understanding of its meaning, and

(3) the first participant signals acceptance

Page 15: Analysis of interaction in collaborative activities; the Synergo approach

15/60

Merging Action and dialogue Annotated model=collection of objects (OCAF Avouris et al. 2003)

MEF = { Entities= E (ABC) = 1/EP, FA , EI

E (VELO) = 2/ EP, FA , EI E (TRUCK) = 3/FP, FI E (STOREHOUSE) = 4/FP EC, FA, FI E (STORE) = 5/FP EC, FA, FI Ε(DELIVERY)= 11/ FP, EX, FI

Relations= R (VELO-owns-SH) = 9/FPI R (VELO-owns-ST) = 10/FPI R(TRUCK-transports- DELIVERY)=17/ EP, FI, EC R(SH-are-suppplied-by-TR) = 18/ FIM R (ABC-owns-TR) = 25/ FPI R(ST-owns-SH) = 24/ EP FP FI EC, EM R (ABC-owns-TR) = 25/ FPI

Attributes= A (DEL.id) = 13/FIM A (DEL.volume) = 14/FIM A (DEL.Weight) = 15/FI A (DEL.Destination) = 16/FI A (TR.Max_Weight ) = 19/FI A (TR.id ) = 21/EP , FI A (TR.Journey_id ) = 23/FI A (TR.volume ) = 20FIM A (SH.id ) = 24/FI

Items not in the final solution -R (SH-DEL) = 12/EP , FR , -A(VELO.Storehouse)=6/ EP , FC -A(VELO.Store)= 7/ EP , FC -A(ABC.Truck)= 8/ FP , EX -A (TR.max_journeys_per_week) = 22/EP , FR }

A(volume)

A(destination)

A(Journey _id)

A(id) A(volume)

E(VELO)

2/EP, FA , EI

E(ABC) 1/EP, FA , EI

E(STORE-HOUSE)

4/FP EC, FI

E(STORE)

5/FP , EC, FAI

E(TRUCK) EP, FI

E(DE-LIVERY)

11/FP, EX, FI

20/FI,M

23/FI

21/EP, FI

14/FIM

16/FI

R

9/FPI

R 24/EP FPI, EM

A(id)

24/ FI

R

10/FPI

R 18/FIM

R 25/FPI

R 17/EP,FI,EC

R 12/EP, FR

A(Max_ weight)

19/FI

A(Id)

13/FIM

A(Weight

15/FI

A (max-journeys/week 22/EP, FR

A (storehouse)

6/EP, FC

A (store)

7/EP, FC A (truck) 8/FP, EX

Page 16: Analysis of interaction in collaborative activities; the Synergo approach

16/60

Synergo

Avouris et al. 2004hci.ece.upatras.gr/synergo

Chat

Act

Page 17: Analysis of interaction in collaborative activities; the Synergo approach

17/60

Synergo

Chat tool

Shared Activity Space

Drawing objects libraries

Partner selectiontool

Page 18: Analysis of interaction in collaborative activities; the Synergo approach

18/60

Synergo Drawing librariesConcept maps

Flow charts

Entity-Relationship Diagrams

Free Drawing

Page 19: Analysis of interaction in collaborative activities; the Synergo approach

19/60

Activity logging

used for :• Build a history of interaction at server• support latecomers during synchronous collaboration• analysis and playback of the activity •Support replication/ reduce bandwidth requirements

Page 20: Analysis of interaction in collaborative activities; the Synergo approach

20/60

Analysis tools

20

Page 21: Analysis of interaction in collaborative activities; the Synergo approach

21/60

Log Data Preprocessor

21

Analysistools

Page 22: Analysis of interaction in collaborative activities; the Synergo approach

22/60

Typed events automatically annotate the diagram

Object A

I C

Actor A Actor B Actor C

Types of events I (Insert), M (Modify), D (Delete) C (Contest)

M R

( )itoaii TOAtE ][],[,,=

Page 23: Analysis of interaction in collaborative activities; the Synergo approach

23/60

Playback of annotated view

Page 24: Analysis of interaction in collaborative activities; the Synergo approach

24/60

What about the chat? Can we annotate chat automatically?

One approach is to ask the user to do it - open sentences (e.g. Epsilon (Soller et al. 97)

Page 25: Analysis of interaction in collaborative activities; the Synergo approach

25/60

Abstract objects

Dialogue messages

Model objects

Deleted objects

(b)

Annotation of chat events

Page 26: Analysis of interaction in collaborative activities; the Synergo approach

26/60

Define types of actions (annotation scheme)

Page 27: Analysis of interaction in collaborative activities; the Synergo approach

27/60

Overview: Visualization of logged actions

Page 28: Analysis of interaction in collaborative activities; the Synergo approach

28/60

Teachers view and tool support

• E. Voyiatzaki, M. Margaritis, N. Avouris, Collaborative Interaction Analysis: The teachers' perspective, Proc.ICALT 2006 - The 6th IEEE International Conference on Advanced LearningTechnologies. July 5-7, 2006 – Kerkrade , Netherlands, pp. 345-349.

Page 29: Analysis of interaction in collaborative activities; the Synergo approach

29/60

Teacher support (supervisor tools)

Page 30: Analysis of interaction in collaborative activities; the Synergo approach

30/60

Study of the use of tools by teachersComputer Engineering University degree program (A’ Semester)

Level of Education

1 Teacher + 5 Teaching Assistants Teachers involved

80 students (46 students 2004-2005, 34 students 2005-2006)

Learners involved

Problem solving activity: Development and Exploration of an Algorithm

Students in Dyads , no roles assignedTypical Laboratory lesson (2 didactic hours)

Collaborative Activity

SYNERGO Collaborative EnvironmentSYNERGO Analysis Tools

Collaborative Tools

Page 31: Analysis of interaction in collaborative activities; the Synergo approach

31/60

teacher

The Teachers Used the proposed views and gave feedback…

Quantified Overview:Class and

Group

The ProcessView

(Playbackof the

activity)

Qualitativeview

Rowdata

researcher

Teachers: “The process view is the most important tool for in depth insight .”

Page 32: Analysis of interaction in collaborative activities; the Synergo approach

32/60

studies

Vrachneika Gymnasio-3rd year

UnivPatras Algorithms

Page 33: Analysis of interaction in collaborative activities; the Synergo approach

33/60

Typical tasks- Collaborative Cognitive Walkthrough of an interactive system

- Designing Data bases (ER-D)

- Building and exploring Flow Charts

Page 34: Analysis of interaction in collaborative activities; the Synergo approach

34/60

Joint Univ Patras -UnivDuisburg croos-national collaborative activities (2004-2005)

• A. Harrer, G. Kahrimanis, S. Zeini, L. Bollen, N. Avouris, Is there a way to e-Bologna? Cross-National Collaborative Activities inUniversity Courses, Proceedings EC-TEL, Crete, October 1-4, 2006, LNCS vol. 4227/2006, pp. 140-154, Springer Berlin

Page 35: Analysis of interaction in collaborative activities; the Synergo approach

35/60

Similar models with different tools (Synergo, Freestyler)

Page 36: Analysis of interaction in collaborative activities; the Synergo approach

36/60

Findings of the Patras-Duisburg study

•Mixture of synchronous and asynchronous approaches.

•Only partly use of the provided tools •Engaging activities - examples of sessions of many hours (4-5 h) in joint activity and discussion

• Innovative use of media and coordination mechanisms

•Good strategies for division of labor•Excellent social dynamics and group spirit.

Page 37: Analysis of interaction in collaborative activities; the Synergo approach

37/60

A distance learning course of Hellenic Open University (HOU) (2003-2004)

M. Xenos, N. Avouris, D. Stavrinoudis, and M. Margaritis, Introduction of synchronous peer collaboration activities ina distance learning course, IEEE Transactions inEducation, vol. 52 ( 3), Aug. 2009, pp. 305 - 311,

Page 38: Analysis of interaction in collaborative activities; the Synergo approach

38/60

Synergo server

ODL Server (forum, exchange of material,

help desk) Asynchronous interaction

Synchronous interaction (share

drawing / chat communication)

Synergo client

Synergo client

ODL repository

Activity logging

Submit final solution Record

activity

Student #1 Student #2

Post assignments, form groups

Tutor

Facilitator

Arrangements on sessions plan- direct contact

Respond to technical and organizational problems –

follow activity

GroupGroup

Mixed media and collaboration approachesAsynchronous group activity

Synchronous activity

Page 39: Analysis of interaction in collaborative activities; the Synergo approach

39/60

Synergo- Discussion forum

Page 40: Analysis of interaction in collaborative activities; the Synergo approach

40/60

Findings of the HOU study

• Infrastructure overhead higher than expected (unforeseen technical problems)

• Peer tutoring patterns emerged in higher degree than younger students

• Multiple media engaged• Strong social aspects of community

building

Page 41: Analysis of interaction in collaborative activities; the Synergo approach

41/60

Study on Mecahnics of Collaboration:Coordination protocol

Group B No floor control: all partners can act in the shared work space

Group A Explicit floor control: Only the key owner can act in the shared work space

0

20

40

60

80

100

120

140

160

180

200

Critical Insert Delete Move Chats

Type of events

Num

ber o

f eve

nts

GROUP A (with key)

GROUP Β (without key)

T+ T-

Page 42: Analysis of interaction in collaborative activities; the Synergo approach

42/60

Findings of the study§ Explicit floor control of the shared activity area did not inhibit problem solving

§ Similar patterns of activity in both groups.

§ group T- was more active than T+

§ T+ students have been obliged to negotiate on possession of the activity enabling key and thus argue at the meta-cognitive level of the activity and externalise their strategies, a fact that helpedthem deepen their collaboration

Page 43: Analysis of interaction in collaborative activities; the Synergo approach

43/60

models

Page 44: Analysis of interaction in collaborative activities; the Synergo approach

44/60

#1 Support for Group Awareness through a Machine Learning ApproachTrain a classifier to be used for estimation of the quality of collaboration using historical data of problem solving activities of students engaged in building concept maps and flow-chart diagrams in Hellenic Open University and University of PatrasM. Margaritis, N. Avouris, G. Kahrimanis, On Supporting Users’Reflection during Small Groups Synchronous Collaboration, 12th International Workshop on Groupware, CRIWG 2006 Valladolid, Spain, September 17-21, 2006, LNCS 4154

Page 45: Analysis of interaction in collaborative activities; the Synergo approach

45/60

Logfile segmentation L={S1, S2, … Sk}

NE

quality of collaboration per segment (bad, average, good)

Page 46: Analysis of interaction in collaborative activities; the Synergo approach

46/60

Correlation based feature selection(CFS) for different segment sizes

NE=60 NE=80 NE=100 NE=200 (2) num_chat (2) num_chat (2) num_chat (2) num_chat

(3)symmetry_chat (3)symmetry_chat

(4) altern_chat (4) altern_chat (4) altern_chat (4) altern_chat

(5) avg_words (5) avg_words (5) avg_words (5) avg_words

(6) num_quest (6) num_quest (6) num_quest

(7) num_draw (7) num_draw (7) num_draw (7) num_draw Correlation based Feature Selection (CFS)

technique:

makes use of a heuristic algorithm alongwith a gain function to validate theeffectiveness of feature subsets.

NE= number of events per segment

Page 47: Analysis of interaction in collaborative activities; the Synergo approach

47/60

Performance of classificationalgorithms

• Naïve BayesianNetwork

• Logistic Regression• Bagging• Decision Trees• Nearest Neighbor 75

80

85

90

60 80 100 200Fragmentation factor NE

Suc

cess

rate

(%)

NaiveBayesLogisticBaggingSimpleLogisticRandomForestNNge

Page 48: Analysis of interaction in collaborative activities; the Synergo approach

48/60

Visualization of group awareness indicator

State of Collaboration

Page 49: Analysis of interaction in collaborative activities; the Synergo approach

49/60

Evaluation study• 11 groups of 3 students each were given a

collaborative task. • 6 of these groups were provided with the group

awareness mechanibsm. • 5 groups did not have that facility• The mean values of collaboration symmetry

were significanlty different between the two sets (p=0,0423).

Page 50: Analysis of interaction in collaborative activities; the Synergo approach

50/60

Side-effect

• in four (4) out of the six (6) groups therewas an explicit discussion about the groupawareness mechanism.

• A side-effect:

Page 51: Analysis of interaction in collaborative activities; the Synergo approach

51/60

#2 Measuring quality of collaboration in Synergoactivities using a rating scheme and an automatic rating model

Based on: Meier, A., Spada, H., & Rummel, N. (2007). A rating scheme for assessing the quality of computer-supported collaboration processes. International Journal of Computer-Supported Collaborative Learning, 2, 63–86.

Page 52: Analysis of interaction in collaborative activities; the Synergo approach

52/60

Original setting New setting

Desktop-videoconferencingsystem with shared texteditor

Synergo: shared whiteboardand chat

Medical decision making Computer programming(algorithm building)

Intermediates;complementary prior

knowledge (psychology andmedicine)

Beginners;similar prior knowledge

CSCL tool

Domain

Collaborators

Meier et al. (2007) rating scheme

Page 53: Analysis of interaction in collaborative activities; the Synergo approach

53/60

Meier et al (2007) rating scheme dimensions

Page 54: Analysis of interaction in collaborative activities; the Synergo approach

54/60

Kahrimanis et al. (2009) adapted collaboration rating scheme

7. Individual Task Orientation (for dyad mean or absolute difference)

Motivation

6 .Cooperative Orientation Interpersonal Relationship

5 .Structuring the Problem Solving ProcessCoordination4. Argumentation

3. Knowledge Exchange Joint information processing

2. Sustaining Mutual Understanding 1. Collaboration Flow Communication

Dimensions Aspect of collaboration

Page 55: Analysis of interaction in collaborative activities; the Synergo approach

55/60

Development of a Collaboration Quality Estimation Model

Data set used• 350 students of 1st year working in

dyads to solve an algorithmic problem using Synergo (academic year 2007-2008) duration of activity 45’ to 75’

• 260 collaborative sessions• Grading according to the quality of

solution and quality of collaboration

Page 56: Analysis of interaction in collaborative activities; the Synergo approach

56/60

36 derived metrics used(Kahrimanis et al. 2010)

Page 57: Analysis of interaction in collaborative activities; the Synergo approach

57/60

Quality of Collaboration Estimator(Kahrimanis et al. 2010)

Observed vs. Estimated CQ average

-2

-1

0

1

2

3

-2 -1 0 1 2 3

Estimated(collaboration quality avg)

Obs

erve

d (c

olla

bora

tion

qual

ity a

vg)

VIPs (1 Comp / 95% conf. interval)

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

Variable

VIP

collaboration_quality_avg = 0.460 + 0.004*C4 - 0.005*C6 + 0.011*C8_17.5 - 0.012*C7

+ 0.602*EV3 + 0.447*STC - 0.001*MW1 + 0.008*MW6

Stone & GeisserCoefficient

(cross validation)

Partial Least Squares Regression Model

Page 58: Analysis of interaction in collaborative activities; the Synergo approach

58/60

Use of Quality of Collaboration Estimator as discriminator between cases of good and bad collaboration

• The model scored between 76.6% to 79.2%, with the exception of one dimension of lower quality.

Page 59: Analysis of interaction in collaborative activities; the Synergo approach

59/60

Use of Quality of Collaboration Estimator as automatic rater

• The model had acceptable performance as rater as the inter-rater reliability with human raters had the following values: ICC=.54, Cronbach’s α=.70, Spearman’s ρ=.62 acceptable for α και ρ (George, & Mallery, 2003; Garson, 2009), not for ICC(.7) (Wirtz & Caspar, 2002) . This applies both for the average collaboration quality value and the individual dimensions.

Page 60: Analysis of interaction in collaborative activities; the Synergo approach

60/60

Current developments• Study of tablet-based collaboration patterns

(synergo v. 5)

• Study of Attention mechanisms (Chounta et al. 2010)

Page 61: Analysis of interaction in collaborative activities; the Synergo approach

61/60

More on Synergo:hci.ece.upatras.gr/synergo

Page 62: Analysis of interaction in collaborative activities; the Synergo approach

62/60

Some more key references• Avouris N., Margaritis M., & Komis V. (2004). Modelling interaction

during small-group synchronous problem-solving activities: TheSynergo approach, 2nd Int. Workshop on Designing ComputationalModels of Collaborative Learning Interaction, ITS2004, Maceio, Brasil, September 2004.

• Κahrimanis, G., Meier, A., Chounta, I.A., Voyiatzaki, E., Spada, H., Rummel, N., & Avouris, N. (2009). Assessing collaboration quality insynchronous CSCL problem-solving activities: Adaptation andempirical evaluation of a rating scheme. Lecture Notes in ComputerScience, 5794/2009, 267-272, Berlin: Springer-Verlag.

• Kahrimanis G., Chounta I.A., Avouris N., (2010) Determiningrelations between core dimensions of collaboration quality - A multidimensional scaling approach, In the 2nd InternationalConference on Intelligent Networking and Collaborative Systems(INCoS 2010)


Top Related