14 April 2009
Info Viz and the Guru: A story in four parts with a surprising twist in part three
Part 1: Trev makes an observation
Here's an observation I've made after many years of study: "Science is quite hard".
In particular, it's fiendishly difficult to show real statistical evidence in a popular medium like a web site. This sort of evidence tends to involve a lot of numbers and graphs and, generally speaking, we human beings hate these.
So imagine my frabjous joy when I heard that Ben Shneiderman was coming to our shores to give a talk on Information Visualisation. "Big Ben", as we like to call him, is one of the Founding Fathers of interface design. My professional colleague - whose name really is "Jelly" - and I popped our skates on and scooted up to Cambridge to see him in action.
Part 2: Trev and Jelly get the good oil
Back in 1980, Ben Shneiderman wrote one of the first books which dealt seriously with problems of software interface design. His "Eight Golden Rules" were a major influence on the current official Usability Body of Knowledge guidelines. He's one of the originals, a venerated elder, a spirit from the Ancient World, a pioneering voice in blah blah blah, you get the idea.
The content of Big Ben's talk was deep and wide, and a blog version summary won't be able to do it justice. But here are a few stand-out examples that were discussed.
Roughly-speaking, there are two types of information visualisation:
- Things that allow ordinary people to see moderately complex info and understand it at a glance
- Things that allow experts to see very complex data differently, so they can extract its overall meaning better
One example of the first type is the Tree Map. (So called because it doesn't look anything like a tree. See my earlier blog on silly names for things)
This Tree Map shows the daily movement of stocks in the US stock market:
Another groovy thing is the Starfield (so called because it sounds more exciting than "Clever Scatter Plot Tool"). Basically, this is just a scatter plot, but it has some clever tools on the side which allow the user to zoom and filter in useful ways. Here, for example, is the FilmFinder screen:
The graph shows the year films were made plotted against their length. So you can easily see which films are lengthy and recent, as opposed to short and old (like Woody Allen). Clicking on a data point shows the movie details, including a picture of Woody Allen.
The clever bit is the tools which allow the user to zoom and filter information in helpful ways. One of these is the "zoombar":
This is like a scroll bar, but rather than just moving it from side to side, you can move either end, and so select the range of data. So, in this example, only films which are between 67 and 181 minutes are shown, and you can easily change this range. What's more the data displayed changes as you move the bar, so you can see how it changes as you go along. This is faster and easier than entering and re-entering ‘from' and ‘to' numbers. There's more to it than just this, but space is limited so I'll press on. View a video of Ben Shneiderman explaining starfields
A third quite groovy method is called "LifeLines" (actually quite a good name). Here's a simplified version which shows how patterns in patients' medical records can be analysed.
The timeline of each patient is marked with the significant events:
You can then choose to align one event type, e.g. a heart attack. This moves the timelines so that the "heart attack events" (assuming just one for each patient) are all vertically aligned:
The graph at the bottom shows when the patients started using Medication X, relative to when they had their heart attack. We can see that more patients started using it just before their heart attack than at any other time, which suggests Medication X is part of the problem. (Please note: in this simplified example, only four timelines are shown but there are many more unseen ones.) See a fuller description, including a video
Big Ben gave a whole lot more examples, including methods of searching for particular patterns within a large number of line graphs and ranking sets of scatter plots by how closely they fit certain shapes.
It was quite inspiring if you're a bit nerdy like me. However, back at the office the next day a disturbing surprise awaited me.
Part 3: Trev gets the disturbing surprise that was awaiting him
So there I was, all fired up about Info Visualisation. But when I started to look up background info on Big Ben I found something disturbing. Many of the examples he talked about, he has in fact been talking about for over a decade. And yet they aren't widely used. They don't seem to have caught on. The starfields video in the link given above is 14 years old.
I found this quite surprising. And disturbing. It was like some sort of "disturbing surprise" or something. I showed it to Jelly and she looked a) surprised and b) disturbed.
However, fourteen years ago Big Ben was talking about using specialised statistical software to do all this which requires highly motivated, expert users. Now we have the capacity to deliver this sort of thing over the web, which leads me to speculate ...
Part 4: Speculation on putting rich, queryable data up on the web
The sorts of methods described here are for analysts who need to extract meaning from data. Most web sites aren't aimed at people who need to do this.
However corporate web sites are aimed at analysts. Numerical information on corporate sites usually relates mainly to finances and operations, and is covered by graphs, tables and some infographics. Some sites provide data in a downloadable spreadsheet so that analysts can load it into their own statistics software and create their own graphs.
Could there be a part for corporate sites to play here? Could they provide a leading role in introducing a range of different ways of analysing complex data, providing tools online which allow analysts to zoom, filter and re-analyse the info?
Some analysts will still take downloadable data and put it into a specialised statistical package. But in providing it in a convenient form, with no set-up or downloading required, even simplified versions of these methods might help guide analysts towards engaging more deeply, and taking a deeper interest in the numbers.
Hmmm. I don't have all the answers, need to ponder on it a bit. Drop me a line if you have any thoughts.
Any statements made in these blog posts are the views of the blogger and do not necessarily represent the views of The Group.