1 A Framework for Conceptualizing, Representing, and Analyzing Distributed Interaction Dan Suthers Work undertaken with Nathan Dwyer, Richard Medina and Ravi Vatrapu Funded in part by the U.S. National Science Foundation Our Context Conducting methodologically diverse studies of distributed interaction Experimental studies of representational guidance Coding acts for statistical hypothesis testing Micro-analysis to understand members methods Online learning and online communities ESDA and SNA analyses of participant roles and boundary spanning We sought a common framework (data representations and concepts) for this work
2 Generalized Motivations Understanding interaction and learning in technology-mediated settings Challenges Interaction is distributed across actors, space, time Synchronous, quasi-synchronous, asynchronous Multiple data artifacts; interaction not visible Complex contingency of human action on context Desiderata Media-independent conception of interaction Render interaction visible in single analytic artifact Support multiple methods of analysis Impartiality and evidence-based interpretation Preview of Framework Empirical Foundation Events (by actors and actants) Contingencies: relationships between events Representational Foundation Contingency Graph as abstract transcript (contingencies between events) Conceptual Foundation Coordinations Uptake
Interaction Interaction ranges from co-present maintenance of joint conception (Teasley & Roschelle) to networked individualism (Castells) Assumption: Interaction can be understood in terms of manifest relationships between acts Empirical Foundations Events, Acts Events include ANT's agency of objects (actants) Acts restricted to object-oriented agency (AT) Contingencies any observed relationship between events that may evidence how one event may have enabled or influenced other events. "many metaphysical shades between full causality and sheer inexistence" (Latour, 2005) may be non-local 3
4 Some Contingencies Contingencies of e i on e j Media Dependency Temporal Proximity Spatial Organization Inscriptional Similarity Semantic Relatedness e i operates on object created or modified by e j e i took place soon after e j e i takes place in configurational context created by e j e i creates inscriptions similar to those created by e j The meaning of inscriptions created by e i and e j overlap Conceptual Foundations: Coordination Actors coordinate between personal and public realms via visible acts Inspired by DCog: bringing internal and external structures into alignment, but A non-representationalist interpretation is possible More general than contribution, message, utterance, etc not assuming conversational setting not limited to expressive acts: includes perception
Conceptual Foundations: Uptake When a participants' coordination takes aspects of prior events as having certain relevance for ongoing activity Interpretative: transforms the taken-up object by foregrounding aspects of the object as relevant for ongoing activity Composable: if u 1 transforms o 1 into o 2 and u 2 transforms o 2 into o 3 then u 2 takes up and interprets o 1 u 1 o 2 : meaning making is embedded in a successively expanding history More on Uptake Attitudinal and intentional as well as informational Applies across media and modalities Can be intrasubjective as well as intersubjective More abstract than argument moves, thematic connections, etc. More general than (includes but does not assume) transactivity Epistemological utility, not ontological claim 5
Representational Foundation: Contingency Graph Directed hypergraph G=(V, E) V = events selected for analysis E = set of tuples (e u, {e 1,... e n }), e i V, where e u is contingent on e 1 through e n Respects chronology: if indexed by time stamps, (e u, {e 1,... e n }), u > i, i = 1, n Example: ({e 1, e 2, e 3, e 4 }, {(e 3, {e 1 }), (e 4, {e 1,e 2 })} CG as Abstract Transcript Abstracts (further) from media-specific transcripts to common format Formal object, not to be confused with implementations: can specify without creating an actual data structure visualizations: can be visualized in different ways Contingencies do not automatically imply uptake: evidence must be interpreted Uptake graph: composite contingencies that are taken to evidence uptake 6
Constructing CGs Identifying Events and Coordinations recording and segmentation is an analytic act includes expressions and perceptions Identifying Contingencies (e u, {e 1,... e n }), annotated with type and pointer to data Iteration and Densification Clustering (raising level of description) Directions of analysis (e.g., functional analysis, grounded theory, CA, goal-driven search) Examples from a Laboratory Study Asynchronously interacting dyads in laboratory Public heath problem with hidden profile materials Original study: representational guidance of evidence maps vs. threaded discussion 7
8 Example 1: Basics How participants converged on "duration of exposure" Manually constructed contingency graph (180 events, 220 contingencies) Small fragment (uptake graph) shown here to illustrate several points Example 1: Notation
1: Interactional Patterns Information Sharing: (13p, {13}) and (20p, {20}) Integration: (10, {13, 20p}) Round Trip: ((20p, {20}), (10, {13, 20p}), (10p, {10})) 1: Advantages of CG Temporal partial ordering Tracking availability and access 9
10 Example 2: Discovery of Interactional Pattern ( W ) Information Sharing / Round Trip in Evidence Map Subsequent Negotiation in Threaded Discussion Example 3: Quantitative Hypothesis Testing One experimental group (evidence map) had higher agreement and higher post-test scores on integration questions. Why? Analysis of artifacts suggested no difference in information sharing Undertook quantitative comparison of patterns to test whether round trips explained the result Patterns were traced out algorithmically and counted
11 4: Multi-Level Analysis of Representational Practices Analysis originally undertaking to explain convergence & divergence, but discovered emergence of representational practices CG generated from MySQL database and visualized in Omnigraffle Interactive utility enabled selective viewing of CG (Richard Medina s work) 4: Multi-Level Analysis of Representational Practices
12 4: Multi-Level Analysis of Representational Practices Lemke: "look at at least one organizational level below the level we are most interested in (to understand the affordances of its constituents) and also one level above (to understand the enabling environmental stabilities)" 4: Multi-Level Analysis of Representational Practices Points illustrated: Automated generation of contingency graphs is possible and can be useful Contingency graph helps select relevant portions of the source data for further analysis (leading to insights into nonverbal behavior) Aggregation of coordinations and contingencies exposes patterns at coarser granularities
13 5: Exposing Roles in Asynchronous Discussion 1 P2 9/23 In your first post, your needs assessment seems to be talkin g about sociocognitive 3:39 rather than socio-cultural 2 P1 9/23 What is the "socio-cognitive" approach? I'd like to read mo re about this 11:15 approach since I am not f amiliar with it. I was really interested in the sociocultural approach b ecause it seems to imply that intellectual development is directly related to socialization. 3 P1 9/23 11:31 4 P2 9/24 2:33 5 P2 9/24 2:34 6 P1 9/24 3:35 7 P3 9/25 10:14 I didn't see any description of the "socio -cognitive" approach in the assigne d readings. I was not familiar with this approach...what is unique about socio-cultural (or CHAT - cultural historical activity theory) is the emphasis on cultural and social context. But you are right, it doe s give an account of ind ividual cognitive change as a function of social interaction......sorry, I meant socio-constructivist (though I have used socio-cognitive to include the former).... Thank you - that clears it up for me! :) I noticed that several of our grant proposals mixed up socio -cognitive for th e socio-constructivist. I was thrown a little at first. Anyone know where the confusion stems from? 5: Exposing Roles in Asynchronous Discussion
14 Summary Framework developed to resolve practical problems in our laboratory (distributed interaction) and provide several lines of work (ideographic and nomothetic) with theoretical coherence Empirical Foundation: Events, Acts, Contingencies Abstract Transcript: Contingency Graph Conceptual Foundation: Coordinations and Uptake Meeting the Desiderata Render interaction visible in single analytic artifact: Contingency graph Media-independent conception of interaction: Coordination and Uptake are independent of media Contingency graphs capture any event-based interaction Support multiple methods of analysis: Statistical analysis of sequential structures (Example 3) Sequential analysis (Examples 1, 2, 4) Computational support (examples, 3, 4 and future work) Impartiality and evidence-based interpretation: No prior coding scheme Contingencies capture observed data Uptake asserts that there is interaction Research programs interpret that interaction in specific ways
Limitations and Caveats Event-based ontology does not explicitly include developmental, cultural or historical situation, identity, etc. except via prior events CGs are partial: avoid inferences that assume they are complete CGs are not to be relied on alone: return to source data More generally, avoid fixing analysis at one level New Work: Associograms 15
Final Comments CGs enable us to deal with complexity of data Software support needed to realize this potential Framework offers Boundary Objects for CSCL Suthers, D. D., Dwyer, N., Medina, R., & Vatrapu, R. (in press). A framework for conceptualizing, representing, and analyzing distributed interaction. International Journal of Computer Supported Collaborative Learning, 4(4). 16
17 Mahalo Dan Suthers suthers@hawaii.edu Laboratory for Interactive Learning Technologies lilt.ics.hawaii.edu Information and Computer Sciences University of Hawaiì at Manoa www.ics.hawaii.edu Bridging