LECTURE 09: INTERACTION PT. 2: COST AND COLLABORATION April 1, 2015 COMP Topics in Visual Analytics Note: slide deck adapted from R. Chang

Announcements Reminder: complete “Needs Assessment” phase of your project (before next class) - Having trouble picking a technique? Go back to slides by D. Staheli, or talk to Jordan - Tip: if you don’t have access to actual users, “Personas” technique can be incredibly useful - Post a brief discussion on Piazza by April 6 th, 5:59pm - What technique you chose - Why that one was appropriate - What you learned about your users (esp. any changes between your original post and what you think now) - Ways this has influenced your design, if applicable

Recap: Yi, Kang, Stasko and Jacko (2007) 1. Select: mark something as interesting 2. Explore: show me something else 3. Reconfigure: show me a different arrangement 4. Encode: show me a different representation 5. Abstract/Elaborate: show me more or less detail 6. Filter: show me something conditionally 7. Connect: show me related items Yi, J. S., ah Kang, Y., Stasko, J. T., & Jacko, J. A. (2007). Toward a deeper understanding of the role of interaction in information visualization. Visualization and Computer Graphics, IEEE Transactions on, 13(6), Select: mark something as interesting 2. Explore: show me something else 3. Reconfigure: show me a different arrangement 4. Encode: show me a different representation 5. Abstract/Elaborate: show me more or less detail 6. Filter: show me something conditionally 7. Connect: show me related items

Discussion Is this really a taxonomy of interactions, or is it a taxonomy of visualizations? Are the two separable? Who we we handle interaction on different visualization types?

Case Study: Brushing and Linking

Collaborative Brushing and Linking Isenberg, Petra, and Danyel Fisher. "Collaborative Brushing and Linking for Co ‐ located Visual Analytics of Document Collections." Computer Graphics Forum. Vol. 28. No. 3. Blackwell Publishing Ltd, 2009.

Recap: Questions to Ask Yourself 1. What is the goal of the analysis? 2. What kinds of operations do we need to enable? 3. How can the visualization support those operations? 1. What is the goal of the analysis? 2. What kinds of operations do we need to enable?

Questions? Thoughts?

Interaction: Benefits and Costs So far, we’ve talked about interaction (at all levels) in terms of what it enables - Maintaining context - Supporting hypothesis generation - Etc. Question: are there any downsides? Costs? Put another way: how do we decide when it’s worth it?

Interaction Costs Lam (2008) surveyed 484 papers, tried to break down “cost” into logical parts: 1. Decision costs to form goals 2. System-power costs to form system operations 3. Multiple input mode costs to form physical sequences 4. Physical-motion costs to execute sequences 5. Visual-cluttering costs to perceive state 6. View-change costs to interpret perception 7. State-change costs to evaluate interpretation

How hard is it to decide where to start? Human intuition: give me more choices! Caveat: decisions require effort As interfaces become more complex and display more data points, users may need to decide to decide on - a subset of data - interface options Decision costs (  goals)

System-power costs (  system operations) Once a person has decide on a question they want answered, how hard is it to translate it into logical operations? Deciding on the correct operation sequences may be difficult (especially for complex systems) When the set of available operations isn’t immediately clear, users may have expectations based on previous systems (!!)

Multiple input mode costs (  physical sequences) Given a sequence of logical actions, how hard is it to figure out how to perform them? Translating system operations to device operations may be difficult due to: - inconsistent mode operations on multiple views (e.g. zooming) - mode change with inadequate visual feedback (e.g. MS ribbons) - overloaded input controls (e.g. gesture-based interaction)

Physical-motion costs (  execute sequences) Once you’ve set a sequence of actions to perform, how hard is it to physically execute them? Fitts’ Law can estimate actions performed with a mouse MT = a+b*log 2 (A/W +1) where MT is average movement time, A is distance between the two targets, W is target width, and a and b are experiment constants.

Visual-cluttering costs (  perceive state) Given a visual representation, how hard is it to perceive the system state? Interaction such as mouse hovering can cause visual cluttering that makes state perception difficult.

View-change costs (  interpret perception) Given a sequence of two views, how hard is it to reorient after changing between them? Interactions usually result in view changes that requires re-interpretation based on expectations: Interpretation requires object association of: - temporal objects, as in zooming; - spatial objects, as in view coordination, and - local and global objects, as in navigation

State-change costs (  evaluate interpretation) Given a sequence of two logical states, how hard is it to reorient yourself (or get back to where you started)? Data analysis often requires reflection on multiple data views or analysis states Lack of refinding support may inhibit exploration. Fisheye vs. coordinated frames

Total Cost (according to Lam) t_cost(…)=cost(startup) + cost(filtering) + cost(decomposing_to_actions) + cost(translating_to_logical_actions) + cost(translating_to_physical_actions) + cost(executing_physical_actions) + cost(percieve_system_state) + cost(reorient_after_view_change) + cost(reorient_after_state_change) Ouch.

One way to organize these costs

Flashback to 1986: Norman’s Gulfs The gulf of execution is the degree to which the interaction possibilities on an object correspond to a person’s intentions and their perceived possible actions Goal and perceived actions Actual steps I need to take to accomplish goal

Flashback to 1986: Norman’s Gulfs The gulf of evaluation is the difficulty of assessing the state of the system and how well the artifact supports the discovery and interpretation of that state Perceived system state Actual system state

Interaction Costs

Scalability Scale of data Scale of representation Scale of analytical capacity

Discussion: Collaborative Scalability Do people scale? Different modes of collaboration?

Dimensions of Collaboration Tasks: - Single vs. multiple Workspaces: - Single vs. multiple - Shared vs. private - Co-located? Coupling - Tight or loosely coupled Synchronicity - Synchronous vs. asynchronous

Example: Coupling - Tang, Tory, et al. (2006) SPSA - Same problem, same area VE - View engaged: one working, another viewing in an engaged manner SPDA - Same problem, different area V - One working, another viewing D - Disengaged, one working, one disengaged DP - Different problems

Example: Coupling - Tang, Tory, et al. (2006)

Design Considerations Heer and Agrawala (2007) 1. Division and allocation of work Parallelizing the work at different phases Competitive vs. cooperative 2. Common ground and awareness Sharing visual environments Sharing awareness through history, or collaborative B&L 3. Reference and deixis “Look at this!” Putting something in the pre-designated area and refer to it later 4. Incentives and engagements Usefulness (Name Voyager) Fun! (Von Ahn’s ESP Game:

Design Considerations Heer and Agrawala (2007) 5. Identity, trust, and reputation Person perceived to be more reputable will be accepted more quickly Explicit reputation formation (seller ratings on eBay) 6. Group dynamics Group management (allows for scaling) Group size and diversity 7. Consensus and decision making Voting and ranking Presentation, reporting, and externalization

Thoughts/Questions?

Reminders Final Project: Phase 2 (Needs Assessment) due before class on Monday