Interpretation is a vital step in scholarship. Studies can be misinterpreted, and how findings are understood and internalized by a community can be more important than the findings themselves.
Last month, the New York Times reported on the military’s use of PowerPoint. Commanders called the continual use of bullet points and other PowerPoint conventions “dangerous.” General H.R. MacMaster said:
It’s dangerous because it can create the illusion of understanding and the illusion of control. Some problems in the world are not bullet-izable.
One slide in particular became the poster child of the problem — a multicolored spaghetti diagram with arrows linking all sorts of concepts, some big, some small. General Stanley A. McChrystal joked:
When we understand that slide, we’ll have won the war.
He may have said more than he realized. When I first saw this story, which was followed by mocking coverage on “The Daily Show” and “The Colbert Report,” among many other places, it bothered me at a subconscious level. This subconscious burr was revealed during a recent business trip with two very smart people. We were shooting the breeze about various topics when they mentioned this article and that infamous slide’s relationship to systems dynamics. They helped me realize what was wrong with the PowerPoint/military narrative, and why McChrystal’s joke might have been both prescient and damning.
We had the picture, but we were being misinformed about what it was and what it meant.
Systems dynamics is a way of portraying the web of consequences, interactions, and relationships that invariably exist in a complex system. Want to avoid unintended consequences? Use system dynamics. Want to understand which features of a system drive behavior? Use system dynamics. Want to know how to efficiently and effectively change a web of dependencies? Use systems dynamics.
Mind maps and causal loop diagrams are primary tools of systems dynamics.
They are also called “spaghetti charts” because they look like that when rendered visually.
It’s actually reassuring that the military is attempting to use systems dynamics to understand complex social/military/cultural/economic systems. People who use these approaches create long-term solutions and avoid pitfalls.
As C.V. Harquail writes on the Authentic Organizations blog:
This diagram is nothing to laugh at, and nothing to make fun of. This diagram is something to celebrate, because it shows us that our military leaders are trying to take a systems approach to the complex problems in Afghanistan. . . . Mind maps and causal loop diagrams are EXACTLY the kind of graphics that our military leaders should be creating, should be sharing, and should be using. This is not the kind of visual presentation that’s part of ‘death by PowerPoint.’ This is the kind of analysis that might save lives.
So, McChrystal was right — understand this diagram, win the war.
But systems dynamics does look messy to people who don’t know what it is or how to use it. In its raw form, it looks like a wiring diagram for your house — if you’re not familiar with electrical wiring symbols, flows, and schematics, it will look like a sort of art deco spaghetti diagram. You just want the lights to work — but it takes the wiring diagram to make that happen.
In its finished form, systems dynamics looks like SIMS or other sophisticated software; it looks like smoothly functioning distribution systems; it looks like things you rely on. Basically, systems dynamics makes complex things work.
Even the military is speaking out, with the bloggers at an official Army blog — the Combined Arms Center’s DRLO Force Management blog — writing:
. . . developing models of complexity is exactly how you go about making the unknown a little more knowable. . . . A systems dynamics model is the first step towards building collaborative understanding of complexity.
The damning part of McChrystal’s statement equating comprehension of that slide with winning the war comes when you ponder that top commanders apparently don’t uniformly appreciate the importance of complex systems when it comes to plans for victory. If military action is still believed by some to be simply to invade, subdue, and defeat, then we still have a problem in the modern military age.
That said, it is worth shaming the use of bullet points to boil down complex ideas. Bullet points can help to organize, but more often than not, they oversimplify. In fact, bullet points are the military’s main complaint about PowerPoint, if you look beyond the spaghetti diagram. But it’s been hard to see beyond the infamous slide because the narrative emerging from the New York Times’ article — that PowerPoint diagrams are silly, overly complex, and useless — is an oversimplification based on poor information and graphical convenience.
Fortunately, the narrative wasn’t completely based on reporting since the New York Times allows comments from readers. Some readers pegged it as a systems dynamics chart from the get-go.
Allowing comments may introduce complexity to the business of news dissemination. But if oversimplification of complex information is a problem the mass media still has, we’re better off with a little more complexity.
(Thanks to SM and JC for the conversation that inspired this post.)
Discussion
17 Thoughts on "Get the Picture: PowerPoint, Systems Dynamics, the Military, and the New York Times"
Nicely put Kent. There is an underlying contradiction in this story, as one slide is criticized as being too complex and the others for being too simple. If this were true then we would ban all graphic display.
I earlier told the story of having my project to map the dynamics of Naval technology cancelled because the Admiral thought the diagrams were too complicated (despite being accurate). I should mention that others, especially the R&D program managers, feared the accuracy, not the complexity, as it made them more accountable for progress.
The problem with bullet charts is that they are still overhead slides, a form of print (or clay tablet) technology. Rooms are long but not nearly as high, so screens are relatively short and the print has to be very big to be seen from the back. The amount of information is correspondingly very small.
Our normal tools for visual communication, such as pages of text, maps, pictures, computer screens, etc., cannot be used in a group setting, because they are all unreadable. This is why slides are so dumb; the ratio of symbol size to slide size has to be relatively quite large. In fact the “spaghetti chart” used in the NYT story would be illegible in a meeting. (Because I do graphic display of complex situations I have fought this problem for 37 years.)
The e-solution is to pan and zoom on the slides, but nobody does it, not that I have seen anyway. PPT does not have that capability. The PPT slide is still basically a clay tablet, a 6000 year old technology.
You’re totally right. This slide is a good example of where Acrobat/Reader could serve as a better presentation tool than Powerpoint.
Acrobat is indeed a start, but it is pretty clunky for a slide show. People don’t need to watch you pan and zoom. You want to be able to create a preset sequence of views, including changing slides, with the ability to override to respond to questions. Doing this with a hand held remote is also needed.
Surprising things happen when you zoom, like seeing features that are hard to see when looking at the whole slide. This is especially true with artworks. But people can find it disconcerting at first. Google maps is changing that.
There are separate issues that are being slightly intermingled here.
“If this an useful graphic for analysis?” is a different question than “Is this a useful graphic for an oral presentation?”
There are many complex figures or tables that work fine in a journal article, where someone has lots of time to examine details, that would not work well in a spoken presentation.
To someone who understands systems dynamics, that chart is useful as a presentation element, because you can directly translate it into understanding. It all depends on the audience. When held up as an example of PPT communication as generally understood, it’s easily mocked. When viewed as a systems dynamics diagram, it’s impressive.
Of course it depends on what one is trying to present. But if it is a system of some sort, and the screen is big enough, then a single map is better than 20 slides with 100 bullets phrases. Consider the difference between a road map and verbal instructions.
That is, there is nothing intrinsically wrong with using a complex graphic in an oral presentation. (However it may take a lot more work to develop an accurate graphic than to bang out the slides.)
If I can pan and zoom, or have a big enough screen, I sometimes use a single slide for a one hour talk. Here is an example, a system of info in the form of an issue tree:
http://www.bydesign.com/powervision/NOx_trees/NOx-02.pdf
The real problem with bullets is that people have to intuit the structural relations among the ideas. Thoughts come in systems, while bullets come in linear lists. Since these underlying idea structures are not simple people make mistakes and misunderstand. Hiding the complexity does not help to convey it.
Get ready for RDF . . . “spaghetti diagrams” of triples (subject-predicate-object) and quads (subject-predicate-object-context) are going to be increasingly common, and increasingly necessary, because they are a way of formally describing semantic relationships that are as complex as the system dynamics prompted by this post — relationships that are _networks_, not _hierarchies_. This is as profound a shift as giving up our “files-in-folders-in-directories” metaphor. We need to guard against our reflex to say “that’s too complicated, put it in bullet points.” The reality is complicated. We want to simplify it so we can get a handle on it, but too often we simplify so far that what we are left with is the comforting illusion of understanding, not real understanding. — Bill Kasdorf
Indeed, automatic RDF network generation is the newest buzz in the Semantic Web world, so we may be in for a deluge. Unfortunately there are dozens of different structures embedded in any given body of information, so actual intelligence (and a lot of human work) will be needed to discern them. This has been the Semantic Web’s problem all along, the lack of a basic theory of semantic structures. (I just happen to have one lying around:
http://www.bydesign.com/powervision/Mathematics_Philosophy_Science/information.html)
Some important structures are networks, but others are trees and some are linear. Knowing which is which is vital to accurate mapping. The RDF triples are just raw data, not a picture of anything.
That may be true of isolated triples, but a lot of data modeling today is done in complex networks of triples and quads to describe relationships. Fundamentally these are not for _visualizing_ the relationships, they’re for _describing_ the relationships. So, true, this is not a “picture” if you mean a visual representation (although we typically want one so we can “visualize” the relationships).
These can also be used for inference. A trivial example: “X works at Z” and “Y is a co-worker of X” enables you to infer “Y works at Z,” thus creating a new triple you didn’t have before. Very useful. But often resulting in “spaghetti diagrams.”
BTW I was not implying that these RDF networks would necessarily be automatically generated. I was thinking of them more in the data modeling context.
And my point was really just that any of these modeling/descriptive metaphors or mechanisms are really all about describing some _reality_, not “dumbing things down” so we can grasp them at a glance (or think we do), which I think is the real PowerPoint issue.
Unfortunately these RDF data networks have not proven to be useful, so far as I know that is. I would love to see one that is useful, for purposes of discussion.
“Thoughts come in systems, while bullets come in linear lists.”
This is a useful reminder insofar as it relates to many of the conversations and debates that we have here re: information transmission.
Comparatively simple linguistic and numeric information can be conveyed effectively in linear arrangements when the object—as it often is in teaching, research, and presentation—is to organize, clarify, and guide for users who typically have a common level of understanding.
However, our cognitive and physiological processes are associative and non-linear. To convey more complex and dynamic ideas, mindmapping, systems dynamics, and information architecture represent linerar/heirarchical organization and radiants, trees, layers, loops, plus characteristics of the relationships that intermediate between information objects—largely in 2D, as per the example in Kent’s post.
As technologies evolve our presentations shift from being distillations (strictly) to incorporating simulation. We are increasingly able to re-enact processes that include dynamic, multi-dimentional simulations and video with gestural navigation. With a pass of the hand, we can switch back and forth between complex representations and simpler visual and linear summaries.
I wonder to what degree the attachment to 2D is necessary versus habitual. For example, I can envision a 3D presentation/simulation but may yet require 2D schematics to translate the concept to stakeholders, planners, and builders in order to bring it to fruition. There are industries outside publishing that are more comfortable with 3D modeling CAD/CAM.
I’d be curious to know more about the degree to which—in the adjacent animation and gaming media industries, for example—3D has supplanted 2D in the planning and presentation phases.
3D is a hot button for me, Alix. In reality the network chart Kent started with above is a 2D rendering of a 3D network. You can tell it is 3D because of all the crossing lines. When 2 lines cross one has to be in front, like pipes. In fact some of the nodes that appear close together will be far apart in a proper 3D rendering, like the Moon and Venus in the night sky. Much of the confusion in charts like this comes from mashing the 3D into 2D.
If you look closely a lot of 2D charts are actually 3D networks. We just don’t have the technology to draw the 3D version. My dream is to build 3D fly-thru graphics of important information structures. It is the only way to really see and understand them.
Scholarly citation or co-author networks might be a good place to start. OSTI has an energy thesaurus with 30,000 terms and 200,000 term-term relations. It captures the concept structure of an important subset of science and technology, but we have no idea what it looks like. There is no lack of candidates for 3D visualization, we just can’t do it yet.
You are correct that 3D CAD and video games are likely places to start when building a 3D graphing technology, also simulation, but none is ideal. Nobody deals with lines in space, nor keeping the labels facing the viewer as they rotate or navigate the structure. I built some simple prototypes in 1992, after formulating my general theory of info structures, so you would think somebody could do it.
FYI more information on OSTI’s 30,000 term semantic structure is at:
http://www.cendi.gov/activities/09-11-08_kos_wkshp_agenda.html
See my slide set: “OSTI Word Web: A Search Tool Using an Integrated Taxonomy and Thesaurus of Science”
“OSTI has an energy thesaurus with 30,000 terms and 200,000 term-term relations. It captures the concept structure of an important subset of science and technology, but we have no idea what it looks like.”
Just wondering how important it is, in this case, to know what the structure looks like? The relationships provide a structure for S/T subset, but to what degree are they artificially imposed references, attempting to describe how science actually occurs? It would be interesting to overlay the 30,000/200,000 against the archival journal literature to see how the structure of disciplinary terms and subsets map to the time series of scientific discoveries—which discoveries were pivotal to other research outcroppings, leading to other discoveries. (This is not to suggest that the timeline of discovery is not also “artificial”.)
Once one knows what one has, which you do, the project will be designed from close analysis of a defined audience, audience needs, and organizational goals. What’s the vision? I agree that one does not necessarily wish, at this moment in semantic technology, to get into heavy post-semantic search manual editing. ‘Tis better to work with what can be done well today—with extensibility.
The relational structure is highly artificial in that it was laid down by expert consensus, mostly in the 1970s, so it is not observational data by any means. It is a folk artifact, like all controlled vocabularies. Still it is clear that there are concept families in there, so we would like to see their structure. One interesting thing is how little the energy related concepts have changed in the last 25 years, or so it seems. That could be tested against the literature.
The key point is that we have only the vaguest understanding of what the concept structure of science is like. In part this is due to the deeper fact that concepts per se are among the most poorly understood things in the world. That is why I study them.
For fun, if you have an iPhone, you may want to try out the Planets app by Q Continuum (tip: test drive during non-daylight hours).
From a review on Appolicious: http://www.appolicious.com/tech/articles/138-planets-app-points-you-skyward:
There are three sections: “sky,” “visibility” and “globe.”
The “sky” section provides a circular diagram with a compass overlay and the currently visible planets, sun and moon in their respective positions. Click on each planet (or sun or moon) to see the name, rising and setting times, and altitude, or angular distance above the horizon.
The “visibility” section of the Planets app has a mini-timeline for each planet as well as the sun and moon with rise- and set-time indicators as well as a red line marking the current point in time. Tap on a particular celestial body and basic stats are revealed: type, radius, mass, moons, etc.
The “globe” section is the wow feature of this app’s updated version. After selecting the celestial body from a list at the top of the screen, you can manipulate a three-dimensional view by swirling your finger around the screen. The Earth even has the sunshine and darkness parts coordinated with the time of day you view it. One minor flaw is a label mismatch between Neptune and Uranus.
For a free app, Planets is slick and complete. If the Planets app doesn’t quench your astronomical thirst, try the Pocket Universe app by Craic Design for $2.99. There are more features, including one called Planetarium, which has a motion sensor as you hold your iPhone or iPod Touch up to the sky. From there you can see labeled celestial bodies in their current positions.
Hi Kent-
Thanks so much for quoting my post… it continues to irk me how few people ‘get’ how useful diagrams like this one can be.
Some of the commenters above mention the difficulty of using maps like this one in presentations (especially presentations projected onto a screen). One tool I like is Presi, which allows you to put up a map and move around, in and out of it in a non-linear way. Not perfect, but better than bullet points.
Also, selfishly, glad to discover this blog!
cvh
