Sunlight or or info "lite"?

A good vizualisation illuminates and stimulates, but a bad one is just about worse than fluff, little substance and lots of eyecandy, if you're lucky. I recently saw a visualisation on a web page which purports to display statistical information about funds 'earmarked' by the US Office of Management and Budget. This viz uses the 'Many Eyes' toolkit created by the excellent crew at the Visual Communication Lab at IBM research. Many Eyes provides a basic toolkit for creating visualisations of all kinds - you just have to plug in some data and with a bit of fiddling, off you go with an interactive visualisation. It's pretty neat. The folks at Sunlight Labs have used Many Eyes to create a vizualization that maps these OMB earmarks to circles of various areas and colors. The resulting charts, which look like very pretty Ishihara Cards, may be aesthetically pleasing, but have problems.

The main one is that the human visual processing system is pretty poor at discriminating numerical values from areas, even ones that are symmetrical, like these circles. People tend to consistently under-report that value that is associated with an area of a given size. Take the two circles in this picture for example. The smaller circle is 20 units wide, and the larger one is 50 units wide. (The units are arbitraty - we could call them pixels, or even 'meow-meows'.) For making a comparative judgement about the costs associated with various expenditures we might ask, "How many smaller circles fit into the bigger one?" Our visual intuition consistently misleads us. People tend to estimate about 4-5 circles.

But a bit of simple geometry can illuminate the issue further. Taking the radii of the circles, we can work out the exact areas that are encoded. Since area=pi (π) times the radius squared, we can calculate the actual area of each circle. The smaller one is: 10 x 10 x 3.14159, or about 314 pixels and the larger one is 50 x 50 x 3.14159 or about 1963. dividing 1963 by 314, we can see that in fact, about 6.25 smaller circles can fit into the larger one.

Thus for a precise reading, relying on area to make an accurate judgement is not a good idea, and so designers of charts and graphics should avoid this pitfall. It could be argued that the folks at Sunlight Labs were generally trying to give an impression of a constellation of values, but the resulting scattershot image makes it difficult to make accurate comparisons and moreover forces the viewer to "hunt" for individual items of interest, such as States.

A better alternative for the purpose of accurate comparisons, rather than just impressions, which would avoid all of these problems would be the humble histogram...

You heard it here first...

It has been said - and there is some debate (http://en.wikipedia.org/wiki/Lies,_damned_lies,_and_statistics) as to the origin of this quote - that there are three kinds of untrtuths: "Lies, damned lies, and statistics."

Well I think it's time to extend this a bit further. There are actually four kinds of untruth: "Lies, damned lies, statistics, and information graphics". Statistics can be used to generate some pretty funky or even improbable numbers. Because most people are unfamiliar with the underlying mathematics, they must simply take it on faith that the people who know are both accurate an honest. For the researcher. statistical results are usually reported with numbers. But when they are presented to others they often get translated into graphics, illustrations, pie charts and other sorts of junk, which can easily lead to misinterpretations. Because visual graphics compellingly take advantage of our visual processing abilies, we tend to forget that they can be misleading. Not only ore there problems with perception (think of all of the optical illusions you may have seen before), but there are simple problems of misleading representations.

Pie charts are the worst, especially for making comparisons. usig false perspective, or misleading manipulations of scale can also cause misinterpretations. And needless graphics can be visually distracting.

Fortunately, one doesn't have to be a mathematician to avoid the trouble with misleading information graphics. Some simple principles can help anyone to weed out the lies from the truths. Edward Tufte has a very nice summary of the possibilities in his excellent series of books. The data-to-ink ratio is probably one of the best ways of avoiding problems. Use only enough ink (or pixels) as necessary to communicate the data. Everything else is extraneous and possible confusing.

Just a little thought about this can help one to avoid troubles and falling for the damned lies...

some words are my friends

You've just gotta love certain words. I have a list I've been keeping lately of some of my favourites:

 

surcease

succor

hyssop

surfeit

 

They all seem to have something in common, but i just can't put my finger on it.  A little je ne know quoi.

 

Enter the Cheese

So, I need (okay, want) to know the melting point of Velveeta. What's so hard about that? Somehow, they can make the stuff, but they don't have any clue about when it melts? You've gotta be kidding!

My original question is at the very bottom of the email I sent to the helpful folks at Kraft foods (no relation)... After you've entered the chees try this.

-----------------------------------------------------------------

From:"Kraft - Online Team2" <OnlineTeam2@Casupport.com>
To: thatbrock
Subject:RE: Your Comment/Question
Date:Thu, 9 Feb 2006 14:43:47 -0500

Thank you for visiting http://www.kraftfoods.com/.

I'm sorry as I'd like to assist you, the information you're requesting
isn't currently available. I apologize for any disappointment this
may cause you.

If you haven't done so already, please add our site to your favorites
and visit us again soon!

Kim McMiller
Associate Director, Consumer Relations

~~TLXEA_12689413~~N
------------------------------------------------------------------

At what temperature does Velveeta melt? In farenheit.

getting an idea of scale

Texas is a pretty big place. This was confusing when I got to the UK, because the scales of distance are no longer intuitive. I have an in-born (inbred??) idea of how long it takes to drive from El Paso to Houston or from Dallas to Austin. Moving to England screwed all this up - not to mention the additional complication of having to drive on the other side of the road. So I had to
make this map to get a handle on things.

After this, I had it all figured out. It made sense. Before I saw the lay of the land before my very eyes, I had no idea that the UK is such a tiny little place. And next time I contemplate a quick little jaunt up to Edinburgh, I'll imagine I am going from Lubbock to San Antonio - though the journey will surely be more picturesque.

Thanksgiving Anarchy in the UK

Well I think we were planning on Thanksgiving dinner (Nov. 24) in the evening. This is somewhat unorthodox. I shall explain.

Now, usually - at least where I come from, where we are gun-toting and God-fearing - Thanksgiving dinner is at about 12 or 1, "dinner" being a term we Texans often use for what people in other countries like California (or England) call "lunch". This leaves plenty of time for eating, feeling bloated, and watching real football in the afternoon. There is always a game on TV during Thanksgiving, and one may find after years of 'training' that to watch men brutally crush one another aids the digestion (it is more civilized than Rugby, which occasionally draws blood - the horror!). Rooting for the Cowboys after dinner (when they are playing) is required in Texas, followed, or occasionally interrupted by, passing out on the Lay-Z-Boy. At this point, the remote control tends to fall to the carpet - take care less'n a toddler come over and chew on it. Somehow childern abound at Thanksgiving, even amongst the unmarried and they get bored with the adults rather quickly, turning to their own forms of mischief. But I digress.

Around 4 or 5pm is time for seconds, usually more dessert. Speaking of, "pudding" is invariably a custardly substance that comes out of a box marked: JELLO in big red, sans-serif letters. Dessert - be it pudding, pecan pie, pumpkin pie, or otherwise - is usually destined for an encounter with CoolWhip (see picture above), a ready made "whipped topping" of which it is mandatory to keep a tub in stock, should guests request it. This does not mean necessarily that it is edible or that it need be. But traditions must be guarded. Breaking with tradition, however we'll not have any CoolWhip since they are still inspecting the shipment at HM Customs and Excise.

We plan however, to have other delights on hand this year, though we'll not be serving the traditional Thanksgiving Turkey for two reasons. First, that they aren't usually very tasty anyway, unless a senior matriarch, ( e.g., grandma) has laid hands to the poor thing - and my partner doesn't qualify yet, thankfully. Second, that they are hard to cook, which is kinda related to the first but I had to come up with at least two reasons. Therefore, we are likely to have a perfectly respectable Thanksgiving ham again - another good ham this year I hope, accompanied by his old friends, the baked sweet potatoes. We shall not want for starch! It would be un-American.

If there aren't leftovers, we will have failed utterly, as it is mandatory to send off the guests with Tupperware containers brimming great helpings of food that one cannot again bear to look upon. We plan to bless our guests with the same. Happy Turkey-day to all.

robot or marionette?

These days, most radios have a digital tuner; one that changes the station by pressing a button. Pressing the button alters the frequency that the radio is receiving by a very small, discrete amount, allowing one to tune in a station more clearly. Formerly, most radio tuners were analog. Tuning in a station was accomplished by turning a knob, which in turn, moved a piece of wire or string inside the radio. This string was attached to an electronic component called a capacitor. The movement of the string moved metal plates inside the capacitor, and this physical change of the position of the plates caused the radio to turn up or down the freqency spectrum. The mechanical linkage allowed a person to adjust the signal very precisely (especially important for the weaker signals in the shortwave spectrum).

Changing the frequency by turning a knob produces a very subtlely shifting auditory experience, one that has an infinitely fine gradient. A person can use this to find exactly the right place, "the sweet spot" where the reception is best. On modern radios, this is usually not necessary unless the signal is very weak. If so, it can be very difficult or impossible to tune such a radio, with a digital tuner, to the very best spot that is available. At times like these, an analog control can be a nice thing to have around. Recognizing this need, manufacturers of digital music synthesizers and samplers have responded by adding to newer models special knobs that simulate the analog controls that former synths used to have - with the benefit of being able to store and recall the position, and thus the sound. Pretty neat. Systems that use digital interfaces that replace analog, mechanical linkages are called "fly-by-wire" systems in the airline industry, and "guide-" or "drive-by-wire" in other industries. Paradoxically, the phrase "-by-wire" refers to the electrical wire connecting the circuity, not the mechanical wire that connects parts of analog control systems.

In terms of user interface design, the difference is important to consider. Such decisions change how people perceive and interact with devices. The digital guide-by-wire devices increment in steps - hopefully ones that are small enough to be useful, but not so small that they are tedious. They also can usually store their state for later recall. Analog devices can be better for some tasks and provide extremely fine control, but are often more expensive to manufacture, and cannot store their state. The mertis of each approach must be weighed carefully in the design of any physical interface.