Conversations Presenters Maggie Law and Vivien Petras 42602

Conversations Presenters: Maggie Law and Vivien Petras 4/26/02 1

Chats and Newsgroups • A chat room is a Web site, part of a Web site, or part of an online service, that provides a venue for communities of users with a common interest to communicate in real time. • Forums and discussion groups allow users to post messages but don't have the capacity for interactive messaging. From: www. whatis. com 2

Goals of Chat Visualizations • To improve communication capabilities of text -only-based chats (use gestures, mimic etc. to enhance and nuance conversation) • To convey contextual information (number of participants, interactions between participants, show participants that are not “talking”) 3

Readings • Comic Chat -- David Kurlander, Tim Skelly, David H. Salesin Proceedings of SIGGRAPH'96, 225 -236 http: //grail. cs. washington. edu/pub/papers/comics. pdf • Stories and Social Networks -- Warren Sack Proceedings of the Workshop on Narrative Intelligence (Cape Cod, MA: American Association of Artificial Intelligence, November 1999). http: //www. sims. berkeley. edu/~sack/Pubs/sack-aaai 99. doc • Chat Circles -- Fernanda Viegas and Judith Donath Proceedings of CHI 99. http: //web. media. mit. edu/~fviegas/chat_circles. pdf 4

Focus of Readings Two tools for small-scale real-time chat: • Comic Chat – conversation represented as a continuous comic strip • Chat Circles – conversation represented as a “cocktail party” of abstract shapes One tool for large-scale newsgroup analysis: • Conversation Map – conversation as medium for computational sociolinguistic analysis 5

Comic Chat 6

Problems with Chat according to Microsoft R&D 1. 2. 3. Users multi-task while engaging in online chat Users must do things other than chatting while chatting (heavy mouse control, navigation) Scenery is generally static, boring 7

Comic Chat: A Solution? • • Graphics are “essential, not merely ornamental. ” Comic strips are a familiar medium to most. Strips themselves are history logs (intuitive). Automatic settings: • • • which avatar to include in each panel placement & orientation of avatars word balloon type and placement camera zoom factor when to begin a new panel default avatar gestures & expressions 8

Development Methodology: • Analyze chat transcripts (how to represent in comic form? what can a computer extract? ) • Automate characters, balloons, panels • Semantic representations, driven by popular topic predictions • Other considerations…. 9

History Logs Comic strips are themselves history logs. The primary visual representation is the historical transcript. All context is captured in one place. Is this a benefit or a shortcoming? 10

Comic Chat in action Example #1: A “typical session” session (http: //www. digitalspace. com/avatars/book/fullbook/chcc 4. htm) Example #2: Some social anthropology… 11

But… does it work? What makes a comic strip interesting? What makes a conversation flow? 12

What makes a comic strip interesting…. ? 6. Surface 5. Craft 4. Structure 3. Idiom 2. Form 1. Idea/Purpose Illustration from Understanding Comics, by Scott Mc. Loud, p. 171 13

Illustration from Understanding Comics, by Scott Mc. Loud, p. 171 14

Conversation = Negotiation “Conversations demand individual and collective decisions over who will speak, when, and for how long. But in informal conversation the negotiations involved in such turn taking, though rapid and efficient, are all but invisible. To claim a turn, people will merely shift their gaze, subtly alter body position, or wait for a break to interject. Others will turn from or toward them, inviting them in or leaving them out. Speakers will hand off the conversation with a look, rising intonation, or a direct question to another person. Or they may attempt to keep hands on by avoiding eye contact, speaking rapidly, rejecting counteroffers, or pausing only at works like and so. (One of the reasons that conference calls can be so awkward is that they remove many of these cues used for effective negotiation. )” -- John Seely Brown & Paul Duguid, The Social Life of Information, pp. 48 -49 15

User Testing? (None. ) “…we look forward to performing user studies to compare Comic Chat to other graphical chat systems. ” • Would also be interesting to compare with nongraphical (text-based) chat systems. • Lots of users still out there… loyal (cult? ) following • Our experience: Takes practice; amusing to experiment with; at times confusing and complicated to control (like other chat tools!). 16

Comic Chat Conclusions + + An interesting idea with a novel approach Innovative ideas (e. g. emotion wheel, semantic extractions) Familiar format & conventions, cleverly drawn Fun & interesting for those who enjoy role-playing - Experience of following a chat session ≠ experience of reading a comic strip - Some touted benefits not as great as they might seem (e. g. history logs, mouse use) - Heavy reliance on automation can be prone to error - Even authors concede it’s an unachievable goal to reproduce creativity of human comic artist 17

18

Chat Circles 19

Comparison I … Text-based chat 20

Comparison II … MS V-Chat The Palace 21

Chat Circles • “Chat Circles is a graphical interface for synchronous communication that uses abstract shapes to convey identity and activity. ” 22

Chat Circles - Identity • Each participant appears as a colored circle, which is accompanied by the user name • Location of circles will also identify participants (important for many users having similar colors associated) 23

Chat Circles - Activity • Participants’ circles become larger when posting occurs (circle adapts to text length) • Circle appears bright when posting occurs • Circles of inactive users fade in the background 24

Chat Circles – Conversational Groupings • There is only ONE room in Chat Circles • Groupings are achieved by moving closer to other participants • At any time, a participant can view all other participants • A participant can also detect interesting conversations in different areas of the room by looking at how many circles are gathered and how often circles become larger • Overview panel in Chat Circles II nice example of focus + context 25

Chat Circles History 26

History Log Patterns + Easy to see “lurkers” + Sequence and size of messages quickly visible - Not very scalable 27

History Log Patterns +/point User-centric: only 1 of view represented the Impossible to see all text at once – requires individual mouse rollovers Easy to see “out of range” conversations – but why would you want to? 28

Chat Circles Demo http: //chatcircles. media. mit. edu/ 29

Comments on Viz + Attractive UI + Consolidated overview of what is going on in the chat room (in version II) + Intuitive and easy handling + No “gratuitous” use of graphics + Does not attempt to convey emotion (unlike other tools, which inevitably fall short) 30

Comments on Viz - Difficult to avoid associating meaning with color - Confusing when many people use the same color - Not very scalable (but does it matter? ) - Text fades quickly, requires full attention - History needs mouse-over to display text 31

Conversation Map 32

Stories & Social Networks • Social Network = a set of interrelated people • Stories weave people into communities or social networks by constituting a kind of common sense • Goal of analysis: understanding a story / conversation – What is told and how is it told? Who tells it? How many people are in the conversation? 33

Very Large-Scale Conversations • What does a Very Large-Scale Conversation look like? – Large – Supports network-based communities – Public • Question: How do we navigate those public spaces? What can we learn about the social interactions? 34

Story understanding • Sociology / AI collaborative filtering – Stories as locations in a social network of producers – Form & content is black-boxed – Who tells it? • Computational Linguistics / IR – Look at form & content of stories – Social context is black-boxed – What is told? 35

Conversation Map • Combination of the social and semantic networks approach • Facilitates navigation of VLSC (very large-scale conversations) • Methodology = Computational Sociolinguistics = Computational Linguistics + Quantitative Sociology 36

Conversation Map • Analyzes newsgroup messages several thousands at a time) (up to • Provides a graphical summary of – A set of social networks (who is corresponding with whom) – A menu of discussion themes important to the conversation – A semantic network that articulates some of the synonyms and metaphors of the discussion – A message thread visualization 37

Conversation Map - Viz Social Network • Discussants • Who is correspon • • ding with whom Shows reciprocity Acts as a filter for spammers Identifies main players in discussion Mouse click isolates discussant network and threads the discussant posted 38

39

Conversation Map - Viz Themes • Summarized topics of discussion • Ranked by frequency • By clicking on a theme, the discussants and threads talking about it will be highlighted 40

41

Conversation Map - Viz Semantic Network • connected terms have been "talked about" in similar ways • Central terms are connected to many terms • Clicking on a node shows threads, discussants and themes that are associated with it • Double click shows all associations (verbs, nouns, adjectives) in the message archive 42

43

44

Conversation Map - Viz Message Threads • Plotted as spider webs • In chronological order (from upper left) • Thread with many messages will show up as almost completely green • Double click zooms in a thread 45

Another click will show the message text 46

Analysis – what is possible? • First level – main panels: – – Themes of a discussion Main discussants Synonyms and metaphors Popular and unpopular threads • Second level: – Comparison of discussions over time (e. g. shift of themes) – Coherence of discussion (are people talking to each other? ) – Semantic cohesion (which themes are associated with a concept? ) – Pattern of social network (center, multiple hubs, cluttered) • Etc. (for examples see also: http: //www. sims. berkeley. edu/~sack/SIGGRAPH 01/ 47

Conversation Map System output corresponds to Metafunctions of language (Michael Halliday) • Interpersonal = social network (language connects people together) • Textual = themes (language connects itself by referencing) • Ideational = semantic network (language contains / carries ideas that are associated with other ideas) 48

Comments on Viz + Summarizes a lot of information in one display + Allows for multiple analyses / interpretations + Uses brushing and linking to connect associated nodes / terms / threads + Uses zooming to display more specific information 49

Comments on Viz - It is clear that the emphasis of this tool in on the many features of analysis and not on the visualization itself (no usability studies are mentioned) - So much information – what does it mean? - Color could be better utilized - Many controls must be learned (mouse clicks, alt-shift combinations) - Spider webs have to be manually untangled 50

Conversation Map – Future work • Improve interface • Speed up creation of conversation map for any given newsgroup so that analysis can happen simultaneously (now it takes several hours to create a map, which doesn’t help a user navigating the newsgroup at a certain point of time) 51

Overall Conclusions • Online Chat seems to be a leisure time activity (no hard agenda, as with other online communication media) • Is a graphical representation useful for a chat? It’s fun to look at but can also be a distraction. • Has limitations. Current technologies (even graphical ones) always fall short of delivering the subtle physical and emotional cues of face-to-face interaction. • Study of tools such as the 3 covered this week reveals interesting lessons about ourselves and how we communicate. 52