A large part of today's seminar was given over to a presentation on "Information Visualization" by Xia Lin on Information Visualization. There was also some initial feedback on the KMT, which I'll summarize again later when more comments come back.

An important point occurred to me during this class, as we discussed relevance feedback, and why users generally don't use it. It's not because it's not effective, but more because they can't be bothered.

I pointed out that it's important for designers to avoid the trap of thinking that this is because users are lazy or stupid. Instead, users are (for the most part) busy and distracted by a dozen different systems with which they have to interact (email, word processing, browsing, etc.). They are actually fairly smart about how to prioritize their time and attention on each system, and they tend to avoid spending more time on one (such as to learn easier/better methods) than absolutely necessary.

The critical bit here is that users don't have time to invest in a given system, so the interface had better draw on their previous experiences to cut the handling time and make more efficient use of the time they do spend learning to use it.

Xia Lin on Information Visualization (IV)
"Search is not enough- it's only the beginning". IV helps us go further with learning and "amplifying cognition". Visual representations that allow for in-depth interaction (reorganizing objects, changing views, etc.) allows us to draw on other parts of how our brains work, such as pattern recognition.

Power of Perception: people are good at scanning, recognizing, organizing, and remembering. Analyzing a static pattern of dots draws on one cognititive strength, while seeing those dots moving in a familiar pattern (such as a human walking motion) draws on another.

Power of Graphics: we use graphics (such as graphs, ER diagrams) to draw on visual cues for resolving logical problems.

Power of Association: with sight and hearing, memory and cognitive processing power is amplified; data that you didn't consciously know before is reconstructed when you need it from associative context. For example, how many windows did you have in the house you lived in 5 years ago?

Further IV resources:

One important thing to note is that IV is not just for retrieval, but rather it is for those who wish to interact with the system and do deeper analysis. This might explain why these ideas are not being developed in a more commercial way.

Discussion of the KMT
Some initial feedback on the KMT interface, from one student.

  • Despite what initially appears to be a very confusing interface, she that it was "carefully designed to follow the user's actual workflow"
  • However, she thought that there should be contact information ((phone number, email) in the UI to allow for direct communication with reviewers and admins. In truth, there are email links, but clearly they weren't obvious to her. The Same applies to the review status information
  • She pointed out that the FAQ editor provides feedback on XML errors, but doesn't provide any when the XML is correct. She felt that some positive feedback would be helpful to indicate that the system has actually responded to the user action (clicking the validation button).

    • >

More KMT feedback later.

Relevance Feedback
Designers of search (IR) interfaces often have to find the balance between user control and automatic processing. For example, in the KMT, what is the best way to set up the default search, as opposed to the level of control provided in the "advanced search"?

Why don't users use relevance feedback (the "more like this" feature included in some search engines, for example)? It is proven to increase relevance in their search results, so what is the barrier?

Likely answer: it requires too much investment in the system.

"HCI/UI designers must be user advocates"