Week Six of HCI saw a presentation by Cathy Marshall on digital interfaces and archiving, and some general discussion about relational browsing and how to best give users access to enormous numbers of documents (such as on the BLS website, with 17k documents). One attempted solution that we looked at is the RAVE "relational browser" developed at SILS, which uses rollover behavior and scripting to do on-the-spot queries behind the scenes to modify the presentation of the content.

Reviewing last week's discussion
Why should we categorize things? Some suggest that the act of categorization forces the user to spend time thinking about the thing, creating a cognititve connection to the item and where it is.

This seminar is about HCI Retrieval – the interfaces that people use to get information (not usability, ergonomics, and other such HCI issues)

ZUI (zoomable user interface) creates a third type of movement in the interface (jump and scroll, and now zoom)

Visible Human UI

Watched the "Visibile Human" video, which concerns an interface to handle images of millimeter slices of a human body. Example of a really good mapping of control mechanisms and representation (the UI) to the type of data being represented.

Speaking in terms of agile views, in this UI the overall body is the "Overview", while the specific slice is the "Preview".

Categorization to support browsing

BLS website page analysis, using the Relation Browser:

using a database to index individual web pages (17000 of them), the entries are sorted by topic (energy, crime, agriculture, etc.), geographical coverage (national, regional, etc.), and agency (census, labor stats, disease control, etc.).

UI uses rollovers to do instant queries and return immediate information (number of documents in a given facet); clicking "freezes" that choice, and then the user can further query by rollovers, or do a final click for a query on those choices. Further refinement of the results can be accomplished by title search, with each letter typed acting as sub-search.

"The real trick of clustering is naming the clusters" – cataloging after that is a lot easier. (this seems to relate to our thesaurus/topic tree issues)

The best method is a machine/human collaboration about deciding on the names (term frequency reports and clustering algorithms for heavy lifting, subjective judgement by people to decide)

"Folksonomy" approach, (like del.icio.us) where everybody helps to catalog the items.

Interface Server Concept
http://idl.ils.unc.edu/rave

————————————-

Presentation: "Looking at our digital belongings from very close up and from very far away"
by Cathy Marshall

she works for Microsoft Windows; ethnographer with a comp sci background

"digital belongings" = our personal stuff, stored away ("lifecaching")

  1. detailed analysis of reading and material form vs. ePeriodical reading data
  2. personal archiving over the long term (20 years)

How we Read
British Library's "turning the pages" project – is this sort of interface sufficient to support reading on the screen? if not, what is?

study of reading suggests that paper gives "lightweight navigation" (folding pages to narrow focus, looking ahead in the text to preview, interrupting text flow to look at other elements, scanning back to re-read for context, etc.)

Personal archiving

we lose our information due to changing technologies, viruses, etc. (ex. floppy disks).

Barriers and challenges (conclusions)

  • Good:

    • predicting future values
    • format opacity
  • bad:
    • people didn't understand file system structures; most go through applications, rather than directly to the files
    • long term intelligibility (being able to tell what is stored where; organization)
    • desktop search; folks understand Google, but not necessarily "search"
  • Ugly:
    • trust and attitudes; optimism, fatalism, paranoia
    • reliance on ad-hoc IT people
    • viruses, spyware, malware are everywhere
    • even in the best cases, something inexplicable would go wrong