Photos Spare Cycles MythBusters

CHI Tuesday notes

alt.chi Re-thinking Humans, Compuers, Interaction, and Design

The Three Paradigms of HCI

Steve Harrison, Deborah Tatar, Virginia Polytechnic Institute and State University, USA
Phoebe Sengers, Cornell University, USA

Informal histories of HCI commonly document two major intellectual waves that have formed the field: the first orienting from engineering/human factors with its focus on optimizing man-machine fit and the second stemming from cognitive science, with an increased emphasis on theory and on what is happening not only in the computer but, simultaneously, in the human mind. In this paper, we document underlying forces that constitute a third wave in HCI and suggest systemic consequences for the CHI community.

  1. Human Factors

    • rooted in engineering
    • interaction as human-machine coupling
    • no guiding set of models or theories
    • critical incidents
    • oldest of HCI traditions, Frederick Taylor early 20th century
    • Fitts Law
  2. Cognitive Revolution

    • rooted in cog psych and information theory (signal/noise)
    • interaction as communication of information
    • cognitive models
    • experimental methods
    • validity is key
  3. Phenomenological matrix

    • rooted in phenomenology
    • interaction as (component of) a situation
    • embodied cognition, critical theory, activity theory
    • ethnography, cultural probes
    • Dourish, Suchman

Kuhn's Structure of Scientific Revolutions * foundational epistemologies * common representations and understanding * tractable questions * broad procedures (methods) * rules of interpretation of procedural results * each paradigm independent * classic experiments as pedagogy -- students perform experiments to learn paradigm * not entirely accurate for social sciences

Agre's Computation and Human Experience * the system of scientific metaphors * more appropriate for social sciences

Warning voice example: 1950s B47/52. Complex cockpits and concerned that pilots would miss warning messages, so went with audio warnings in the pilot's earphone. * P1: male pilots react faster to female voice (all male flight crews, ground control). In fact, react even faster to wives and girlfriends, so they had personalized messages recorded by them. * P2: female voice increases signal to noise ratio * P3: how does the pilot feel about the voice? Have we gendered aircraft as female?


  • Human factors: how can specific problems that arise be fixed? Cool hacks. You tried it and it worked out.
  • Cog rev: what is the user's model? How can this be operationalized? You compare two or more isolated and specific alternatives.
  • Phenom: What existing activities should be support? Support appropriations? What are politics, values? You argue about the relationship of data to what you seek to understand. What you don't build as important as what you do.

Questioning the Technological Panacea: Three Reflective Questions for Designers

Eric Baumer, Bill Tomlinson, University of California, Irvine, USA

This paper argues that asking whether or not a technological solution is appropriate should be an explicit and exposed part of the design process. It raises three questions that should be addressed during the design process: Are there other, possibly non-technological, solutions that could address the problem equally well, if not better? Are designers creating solutions to problems that users themselves do not need to have? Are these technological solutions treating a problem rather than its cause?

Many types of problems -- social, emotional, educational, health, etc... -- but one solution, i.e. computational technology?

Three Questions: * Non-technological solutions? * Problematization? Do the users see a problems and a need for a solution? * Treating symptoms or cause? Complex causes and effects.

Non-tech solutions can be less expensive, less time consuming, less environmentally detrimental.

Miro (Sengers and Gaver, 2005): projection on a wall based on works of Joan Miro. Probes judge the emotional climate. * Alternatives: bulletin board, group art wall, etc...

Mother's Work/Artful Design (Taylor and Swan, 2005): studied use of home calendar systems, catalogues many low-tech solutions (e.g. flower-petal calendar with a petal for each day). Proposed solutions without asking if there was a problem. Should the division of the labor have been questioned instead? (i.e. not mother's work) Example challenge: a washing machine with a fingerprint scanner that didn't allow the same person to do the laundry twice in a row.

Road Warriors: notion that downtime/disconnection is bad. Bell (2006) questions constant connectivity. Downtime important for relaxation and reflection in some cultures.

Ticket to Question/Social Awkwardness (McCarthy et al, 2004): give people a ticket to talk (ala Harvey Sachs) at conferences to try and address social awkwardness of question asking. RFID badges. Possibly created a different awkward social interaction rather than fixing hypothetical problem. Question causes? (e.g. hierarchy, prestige, i.e. situation rather than people)

Computer-Human Interaction: self-selection for technological solutions.

Not Luddism: important to push boundaries, but do so reflectively (it's OK not to use technology).


Disruption and Recovery of Computing Tasks: Field Study, Analysis, and Directions

Shamsi T. Iqbal, University of Illinois, Urbana-Champaign, USA
Eric Horvitz, Microsoft, USA

Presents results from a field study investigating user behavioral patterns during disruption and recovery from notifications in computing environments. Based on the findings, design implications for recovery tools are discussed.

Interruption affecting users and tasks (Cutrell et al '00, '01; Czerwinski et al '00; Iqbal and Bailey '05, '06; Horvitz et al '03; Fogarty et al '04; Mark et al '05)

Four phases of the interruption cycle

  1. Pre-interruption: user performing task
  2. Preparation: user finishes ongoing tasks and transitions to respond to interruption
  3. Diversion: opening of other applications, etc...
  4. Resumption (instrumented as return to original application, some remnants of diversion can remain)

Study not interested in cases where user breaks from cycle

Logging: Email and IM notifications, Window status, Keyboard activity

Interruption rates and time spent: * Pre-interruption: ~4 emails and ~3IM notifications/hour. Task switch every 72s * Preparation: 5 minutes. Task switch every 84s. Higher rate of 'enter' key (marking resumption area with whitespace). Higher rate of paste. No diff for 'save'. * Diversion: 10 minutes (includes diversions of diversions). Task switch every 46s -- use time to perform unrelated diversions. Less focused. * Resumption: 11-15 minutes. Task switch every 25s. Trying to locate original application by cycling through open applications (several instances of same app). Applications <25% visible took significantly longer to return to that >75% visible. Visual cues can help here. Takes time to read cues to reinstate task state. More time spent on app before interruption, the less time it takes to resume.

Conclusion from above: users stabilize ongoing task states, clutter of interruption contributes to resumption time. More stabilization = faster resumption.

Visible windows are reminders

Users unaware of extent of time spent in diversion

Possible solutions: * support activities to stabilize suspended task state: automated saves, cues to current state and resources * reminders of suspended tasks during diversion * easy access to suspended applications during resumption * support to reinstate context during resumption (e.g. replaying past actions from undo stack)

Side note: no significant difference between managers and non-managers, though developers receive less communications

CAAD: An Automatic Task Support System

Tye Rattenbury, John Canny, University of California, Berkeley, USA

We present a system that automatically discovers and displays task representations. Through a feasibility study, we demonstrate that automation is a viable direction for future task support and management applications.

User: knowledge and information worker, both organized and unorganized

"To what extent can task-relevant information be automatically organized"

Multi-monitor setup

Differs from semi-Automatic approaches -- TaskTracer, UMEA, Kimura (Voida et al) -- in that it is designed to work without any explicit input by user

"Context structures": similar to "tasks" or "sub-tasks" but no notion of goal or intention. Sets of people and information that regularly co-occur

Architecture: Logging - Pattern Mining - Activity Display

Activity Display: * (1) feedback for self-reflection * (2) improve access to relevant information * Circles representing context structures, sized by level of activity * Direct-manipulation editing: edits are implicit feedback to CAAD

Future work: * add search terms based on active context structures * modeling inter-context structure relationships * file-sharing recommendation system

Understanding and Developing Models for Detecting and Differentiating Breakpoints During Interactive Tasks

Shamsi T. Iqbal, Brian P. Bailey, University of Illinois, Urbana-Champaign, USA

Demonstrates the feasibility of building models that are able to detect and differentiate breakpoints during free-form tasks. These models can enable interruption management systems to realize defer-to-breakpoint policies in practice.

Breakpoints: * Coarse: switch from programming to e-mail * Medium: switch from MSWord to Firefox to search for a reference; switch between editing two images * Fine: switch between editing and compiling

Interrupting people at breakpoints in task execution significantly reduces cost; even better if done at coarse breakpoint

Existing approaches: * Activity switching: can only detect on level of breakpoint * Create machine parsable description of tasks: requires many descriptions

People can identify 2-levels of breakpoints in observing someone else perform a task (Zack & Traversky, '01; Newtson et al '73, 76, 77) -> determine features that characterize breakpoints in interactive computing tasks

Used observers to identify coarse, medium, and fine breakpoints by watching videostream of others performing their own task

Candidate features: * Switches in unrelated apps * Large shift in focus within app * Change in action within app * Derived features from keyboard and mouse action

Post a comment

related entries.

what is this?

This page contains a single entry from kwc blog posted on May 1, 2007 9:44 AM.

The previous post was CHI Monday notes.

The next post is The Sports Illustrated way... in 2004.

Current entries can be found on the main page.