Edward Tufte on the Iraq War

Ironically, that graphic is *great* at showing how stupidly rosy-eyed the neocon vision of Iraq (or at least the one they were selling to the American public) was,

And it's actually a pretty accurate picture of what's happened in Iraq, if you run the time scale in reverse.

Ah, yes. PowerPoint: The language of the generals. Briefings shed information the higher they go. They might have coined the phrase KISS for generals, except I think it's really supposed to be more like Keep It Simple for the Stupid. Oh, they're not stupid, just busy.

Does "Beautiful Evidence" still contain the ghastly critique of phylogenetic trees that Tuft had on his blog and was considering including? Not that Tuft is a creationist, but he clearly didn't understand what trees represent and seemed to think they were more or less arbitrary diagrams and not the output of algorithms working on hard data.

Powerpoint is a huge bugbear for me, because the medical curriculum is fulsome with them, just insanely corpulent with them. And nobody does them well. I once saw a resident so blearily-eyed glozing over his ppt, he pronounced several words he had misspelled in the powerpoint as-is, without correction, apparently having refined those neural pathways leading from writing recognition to speech areas.

Any teachers or profs in the house: do you use Ppt? How extensively? I only ask because I've *never* seen it used right.

Badger: I didn't read the whole book, so I didn't notice -- I'll check again on another day.

Burghardt: Yes, I use powerpoint (actually, its more cleanly designed sister program, Keynote). Do I use it right? I don't know. What's right?

I use it as a vehicle to put fairly complex images on the screen, which I then try to spend a little time explaining. I find that I typically use 15-20 slides for an hour talk, but I think I'd be doing a better job if I cut that back to ten or so. The problem with powerpoint is the temptation to throw every little point on one slide after another, and rattle them off rather than actually, you know, talking about the data.

Yes, I do succumb to sometimes stringing together too many little bits and pieces, and calling that a lecture. I could be a better teacher.

J. Badger:

In this manuscript page,
http://www.edwardtufte.com/bboard/q-and-a-fetch-msg?msg_id=00018e&topic…
it certainly does seem that Tufte is betraying a significant ignorance of the field of phylogenetics. Phylogeneticists are of course acutely aware that consensus trees (or max parsimony, or whathaveyou) are not necessarily what happened in truth, and that the cladogram is a reduced, illustrative display of the results of a phylogenetic analysis.

I hope he passed it by someone in the field before publishing it on paper.

Regarding Powerpoint, Tufte has *excellent* advice for giving talks in his first (and best) book, the Visual Display of Quantitative Data. One point he makes is that any display that has significant detail should be available to the audience in paper form, so that they can refer to it at leisure at any time during or after the talk.

He also makes the EXCELLENT recommendation never to go OVER your allotted time, quoting I-forget-who as saying, "Nobody ever complained about a seminar ending too early" or words to that effect.

About 2/3s or the way down this thread on ET's blog is another truly horrid Iraq War powerpoint:

http://www.edwardtufte.com/bboard/q-and-a-fetch-msg?msg_id=0000fv&topic…

One of my observations about briefings is never to include too many words on a chart. If you do, the audience will have a choice of reading the chart or listening to you; they can't do both at the same time. I am not sure giving a printed version of a briefing to the audience to read during the briefing is a good idea. It's the same problem. Do you want them reading or listening? I agree that a handout for later is a good idea.

"I am not sure giving a printed version of a briefing to the audience to read during the briefing is a good idea."

If you mean the silly practice of printing out your PP slides and giving them to the audience ahead of time, I agree with you.

That is not what Tufte recommends. He recommends handing out data-rich displays that are keys of the talk. In PP, the audience would get a minute or two to decipher these displays and then *poof* onto the next slide and gone.

I suppose if *all* you want to convey in a talk are very specific points, that is OK. But in the scientific community we are supposedly trying to share data/results/ideas and get different viewpoints on them.

"Do you want them reading or listening?"

Whichever works best to convey complex ideas in a short amount of time. Talking isn't necessarily the best way to do that.

Stephen, my point is that it's counterproductive to give an audience two things to do at the same time. I perfer to have few slides and talk to them. If you have a complex slide, then spend the time necessary to explain the major points. If you can't do that in your talk, then you shouldn't use that slide. If you are trying to convey a lot of complex information that you can't explain in a talk, maybe you should consider doing it in a publication. Then the next time you talk about your work, people will already be familar enough that they can give you different viewpoints.

PZ--
in Geoff Pullum's "5 Golden Rules for giving academic presentations", point #4 mentions it being a good rule of thumb for each transparency to be up for 3 minutes, or at least 2. I've seen Pullum give talks and he's really good. So your timing re slides sounds like it's about right on, FWIW.

Point well taken about not covering too much in a seminar. Reality is that you will always be giving a simplified version of the work. Furthermore, members of your audience will have different levels of expertise in different areas, and will be paying attention to different parts of the talk.

If done correctly, you are simply arming your audience with data by which they can better understand and/or critique your work. It's a "nothing up my sleeves" approach that people like.

I personally like it when speakers have handouts, and have never received a complaint for giving them out; I've even gotten positive comments about the handouts when the talk is over.

I think it's an unrealistic view of the presentation process, that people will be "receiving" the wisdom at exactly the rate and time at which you present it.

Doesn't that dreadful powerpoint capture mean there was an Iraqi plan afterall? I thought we went in with no plan? That's what I've been hearing. Should I brace for a paradigm shift?

Jonathan Badger: I agree that phylogenetic trees can be useful, but just because something came as a result of an algorithm operating on data doesn't make it useful, or doesn't entail that there isn't a better representation ...

I agree that phylogenetic trees can be useful, but just because something came as a result of an algorithm operating on data doesn't make it useful, or doesn't entail that there isn't a better representation

Well, I take objection to Tufte claiming that phrases like "consensus tree" "sounds like marketing" (his phrase). It doesn't. Everybody reading a paper about phylogeny knows *exactly* what that means because there are standard methods for making consensus trees.

Secondly, the difference between a phylogenetic tree and the typical sort of thing Tufte deals with, such as charts of oil price, is that there is a standard representation (okay, a couple standards, if you consider v-shaped trees to be different from square-shaped). I'm not convinced that someone from outside the field like Tufte can really make improvements that won't confuse readers of phylogenetic papers.

I'll try to look through the published book to see what Tufte finally wrote (if anything) on the subject. There were improvements between the two drafts on the blog, so maybe the final version is less objectionable yet.