microsoft research talk: ben shneiderman

Quotes a participant in one of his workshops who says "I've spent 20 years thinking about collaboration, but only 2 hours thinking about creativity."

Much of the literature on creativity doesn't mention computers at all. This is a new space that he's claiming. Not making machines more creative (AI approach), but developing/improving computing tools to make people more creative.

Heifferentiates between revolutionary, paradigm-shifting creativity, impromptu everyday creativity, and his area of interest, "evolutionary, normal science, music, and art, creative knowledge work."

How do we enable professional workers to move a little higher up the ladder in knowledge work using creativity?

Talks about the difficulty of doing empirical research into creativity, and reviews some theorists in creativity research:

• Structuralists: Polya's How to Solve It (1957), Couger (1996) review of 22 "creative problem solving methodologies" and Atman's design steps (Atman et al Design Thinking Research Symposium 2003)
• Inspirationalists: Free associations, Breaking set, Visualization (Aha!)
• Situationalists (context, community, collaboration): Personal history, Consultation, Motivations

Tells about a student who sent him email saying "My PhD proposal is attached, can you tell me what you think?" No context, no definition of expectations, and no reason for him to look at it.

Highly recommends Csikszentmihalyi's book _Creativity_ (1993) -- not quite a software requirements document, but close enough to allow him to take the next step towards that.

1. Domain: e.g. mathematics or biology "consists of a set of symbols, rules and procedures
2. Field: "the individuals who asct as gatekeepers to the domain...decide whether a new idea, performance, or product should be included" (originality is a necessary but not a sufficient component)
3. Individual: Creativity is when a person...has a new idea or sees a new pattern, and when this novelty is selected by the appropriate field for inclusion in the relevant domain"

(At this point there are a lot of comments and questions from the room; this is like a professor's dream seminar enviroment--a room full of very bright and interesting people who are here because they want to be, and are enthusiastically engaged in what you're talking about.)

Shneiderman proposes these eight activities:

1. Searching and browsing digital libraries
2. Consulting with peers and mentors:
3. Visualizing data and processes
4. Thinking by free association
5. Exploring solutions -- "what if" tools
6. Composing artifacts and performances
7. Reviewing and replaying session histories
8. Disseminating results

Talks about each individually:

For search, we need not just effective basic search, but also improved multimedia search, overviews and previews, result set categorization and visualization, multiple session searches. (Gets a laugh when he says that if you can't find something to be creative about in search that Google hasn't done, you shouldn't be in this business.) We need faceted search, and the ability to preview cardinality of results. (MSN Search folks, take note: you should watch this presentation on the internal resnet site.)

Consulting with Peers and Mentors: Prefaces by saying the next killer app is responsibility, trust, empathy. The problem with consultation is a lack of trust. Need negotiated expectations.

Visualizing data and process--refers to _Readings in Information Visualization: Using Vision to Think_. Many data types to deal with in visualization, scientific visualization is different from information visualization. Getting to networks (from trees) is the challenge.

Talks about exploration and discovery. Shows a photo of birds, asks people to answer the question "what's interesting here?" High dimensional spaces are messy. You can't be hunting unless you know what you're hunting for. If you don't have hypotheses and/or specific queries, you can't explore effectively. In any field, peopel are trained to look at things in thorough, orderly way. He says we need to develop the same skills for looking at high-order data. (Can't you argue, though, that those orderly ways can cause blinders?)

Shows a chemical table of elements in spreadsheet form. Anything interesting? Hard to tell. Shows it in a scatter graph, asks the same question. People note outliers and correlations. So, can we build an outlier detector? A correlation detector? (Ah...yes, it's the outlier detector that's interesting to me in the context of social networks. We do a lot with correlation, less with outliers...)

Demos Multi-V: Hierarchical Clustering Explorer. Displays demographic data using it. Very complex tool, but at least demonstrates clustering methods. Also provides some very nice correlation tools; this is a data set with 14 different variables relating to US counties. So...counties with a large number of young people have high unemployment. Decreasing #s of highschool grads correlate to increasing levels of poverty. These are not surprising, but you can find interesting second and third level correlations. For example, slight increases in income cause dramatic reductions in poverty. Also possible to find quadratic relationships, outliers, and other kinds of clusters and relationships. Biologists were particularly interested in how the tool could identify gaps.

This is a fascinating tool. Would love to play with this in the context of social networks, tagging behavior, etc.

He's running out of time, so flies through the rest of the slides:

Free association--suggests Axon idea processor.

Exploring solutions: mentions the work of Michael Terry and Beth Mynatt at Georgia Tech. Create multiple versions at a time, not one at a time.

Composing artifacts and performances--provide templates and exemplars. Need better tools, for better presentations.

Reviewing and rpelaying session histories...record, review, annotate, disesminate. Treat histories as first class objects. Replay them, step through them, etc.

Adobe photoshop history tools...each event is a discrete component; we should be able to do that in all tools.

Disseminating results--need better ways to do this. (Doesn't mention blogs, but that seems like an obvious tool here.)

Challenges for creativity work: Domain knowledge is vital, it may take years, individuals have highly varied approaches, and the theories (about support tools) are shallow, evaluation is difficult, need better ways to do triangulation and multi-dimensional representation.

Influencing our colleagues: Want NSF to incorporate creativity into existing programs, also encouraging a new NSF program on software tools and socio-technical environments to enhance creativity. (Hmmm...the "socio-technical environments" piece of that is intriguing.)

Mentions Richard Florida's compelling argument about creativity as a potent economic driver.

And that's all, folks. Wow. A lot of content packed into a short time frame, and a lot of great ideas to think about. These research talks are definitely one of the big perks of being here...