Strategy, Tactics, and Heuristics for
A Structuralist Approach by Rob Tow
Grand Strategy, Strategies, and Tactics
Deciding "What to Do Next" (and How to Do it) is a problem I've faced repeatedly in my career. Over time, I've developed an approach that has a definite structure. Much of this is inspired by the thought of several historians of military history, combined with a wry appreciation of some of the valid points of deconstructionist philosophers. I combine these tropes into a method for conducting research for inventive design - that is, the design of that which is novel.
Liddell Hart, the military historian, wrote of a three-level top-down structure to the struggles between nations. He called the top level "grand strategy". It is the single overarching high level end goal- for example, to defeat the Axis, or the eradication of Carthage. An example in design would be Microsoft's desire to dominate the desktop. The next level down is that of "strategies". He described strategy as "the art of distributing and applying military means to fulfill the ends of the war." The bottom level is "tactics" (1).
A grand strategy will have more than one strategy (the second level down) that supports it. Strategies are the types of battles that are fought and how they are conducted. Examples in war are daylight strategic bombing, guerrilla warfare, and blitzkrieg combined-arms maneuver. In research, examples are Edison's strategy of methodically testing thousands of substances for the filament of the incandescent light-bulb (2), or Goodall's careful observations of chimp behavior (3). Both of these latter are "dialogues with Nature", where previous understandings lead to a choice of observations and measurements, with expectations that may be contradicted by what Nature actually does. In business, Microsoft's "embrace and enfold" is a powerful strategy. In design, think of Nokia's focus on the teen market.
Strategies are supported by tactics. These are the details in the small of battles, the day to day struggles. They change rapidly. Examples are the infantry assaults on pillboxes on Omaha Beach, and the use of MP3 as a format by Rio to sell music hardware. Nokia's downloadable ring tones and swappable covers are tactics supporting their strategic focus on the teen market.
Many workers in research and in design start their efforts by falling in love with a whizzy strategy - or worse yet a mere tactic - and then try to dream up something grand to do with it. This is like the Polish obsession with the glory of horse calvary tactics at the eve of WWII - doomed to failure in competition against Guderian's panzer divisions.
Let's take these three levels - grand strategy, strategies, and tactics - as a given structure, and observe that starting from the the top level and working down is the right way to proceed with an economy of effort. Nothing at the lower levels is accepted except that which supports the next level up.
Finding the Void
This observation sets a meta-requirement for strategic thinking - to find the void. I approach this in a structuralist manner - describing the essential qualities of an existing design space, or technology, or user experience - and then imagining inverting a small number of its specific qualities. In a way, this is akin to the deconstructionist creation of the "other" by knowing the "self", or to the inferring the nature of a lost text by inverting its triumphant Hegelian opponent (5). For example, consider a conversation between two people. In days ancient and now forgotten conversations were all face-to-face. The opposite (inversion) of this is to have bodies separated in space. Let's think of describing this structurally as a number line, with face-to-face at the left end at (0), a shouted conversation across a canyon a little to the right (0.3) - and a telephone conversation at the far right (1). We now have three points along an axis. One could imagine building something to populate the unoccupied point at (0.6) - perhaps a super-megaphone?
Can we imagine a new axis, at
angles (the "y" axis) to the first (call the first axis
"space"), and call this new axis "time"? What
would excursions along it be like? Well, face-to-face is at the
origin... and a recording is at the top of the y axis. We now have a
Cartesian grid, with tape recorders at (0,1), face-to-face
conversations at (0,0) and telephony at (1,0). And there is a void -
an unoccupied locus - on the grid at (1,1). Clearly, its "nature"
would be a recorded conversation between two people far apart in
space and separated in time.
Can we add a third axis? Perhaps that of other media types? This rapidly takes us into what is now the domain of instant messaging services – which are all populating new loci in “conversation space”. Or perhaps even a fourth axis – the number of people involved.
Notice that many of these loci exposed by such a structuralist analysis are dependent on having technologies to create capabilities to support their intent. Often such capabilities are logistical in nature, as in the telephone network - but they can be re-purposed to support a new intent revealed as part of a structuralist design process - just as the telephone network was re-purposed to be used by the first computer modems to transport digital data over analog voice lines by the Strategic Air Command's SAGE network in the 1950s (6). Cleverly re-purposing existing logistics is an important heuristic for design. But sometimes there is no way to create the capability that can populate the void - we call this unobtanium.
Nonetheless, we now have a powerful analytic tool for design. We describe experiences and affordances, creating a Cartesian multispace, and note where existing things cluster - and then observe where the unpopulated loci are - and describe these latter's nature in terms of experience and affordance. Then we analyze the technology needed to support their existence... and proceed to actual engineering.
Problems will arise in the engineering, from resource constraints, and social problems. I'll examine these and illustrate them with some examples from my own experience.
Let's look at resource constraints next.
The Russian Sleigh Ride
Years ago I saw a painting of a horse-drawn sleigh racing across the Russian steppes, chased by wolves. The people in the sleigh were throwing cargo out, lightening the load to escape. In my memory a woman was holding a baby.. and looked to be desperately thinking of tossing the infant to the wolves.
This painting provides a metaphor for what to do when you have a design with many features - a lot of cargo - and you are being chased by constraints in resources - time, money, staffing, etc. You have to lighten the load.
In what order do you throw things out?
This problem arose in a project that I worked on with two brilliant designers - Brenda Laurel and Rachel Strickland - the Placeholder virtual reality project at the Banff Centre for Arts. We had planned an elegant and ambitious design for a multi-person and multi-world virtual reality installation, with 3D spatialized audio, manipulable virtual objects, field recording of visual and audio real-world locations, elaborate avatars as petroglyphic spirit animals and transformations of bodily appearance, sensory experience, and transformations of the audio of spoken voice. Due to a scheduling collision at the Banff Centre, we found ourselves with six weeks to do six months worth of work. The chief programmer had a nervous breakdown... and the head of the program panicked and demanded that we build something much simpler - one person, one world, no spatialized audio, no fancy avatars.
I thought hard about this, and for the first time did an explicit structural analysis of a design. He wanted one person; we wanted a social experience. He wanted one world; we wanted multiple worlds. He wanted no avatars; we wanted "smart costumes" where people were embodied as Spider, Fish, Snake, and Crow. We wanted to let people record snippets of narrative to leave behind in the virtual worlds, and for people to fly like a bird when they were embodied as Crow... he wanted a simple walk past a landscape, with no shared history.
I made my goal to make a plan that preserved as large a space - a volume - within the multiple axis' of social experience, multiple worlds, affordances, and sensory experience as possible. I tossed out elements that did not increase the enveloped volume within the design space - for example, we got rid of the buzzing mosquito that I had programmed that pestered users by buzzing around their heads - and we abandoned an effort to map bodily movements to "wriggles" in the bit-maps of the petroglyphic avatars. We did not implement sensory transformation models for the special qualities of vision for the Crow avatar, or for the Spider. Working with the programming and design staff, a plan was made that conserved the excursions along the structural axis', preserving the volume in the hyper-space of the design elements. We tossed a lot of stuff out of the sleigh.
We met our performance date. It was a Russian Sleigh Ride exercise of desperate order that preserved the baby. And yes, people did fly through the Three Worlds when they were Crow (7).
Some of the problems in design, and in expressing design in engineering, are social. Action can be hard to achieve.
The fighter pilot, designer, and strategist John Boyd did the first thermodynamic analysis of air-to-air combat. He translated his mathematics into the design of the F-16 and the A-10 "Warthog" - machines designed for very different roles. He also originated a cognitive and social model called the OODA loop - "Observation-Orientation-Decision-Action". Individuals (fighter pilots) and organizations (companies) both iterate this loop in their behavior. Boyd observed the importance of actually performing each step of this loop, and in shortening its standard delay. "Getting inside your opponents loop" (e.g., acting faster on similar data) is an important heuristic for a fighter pilot - or a design bureau (8).
An organization has to be able to actually and really perform each step of this loop. Sometimes the hardest part is the "decision" part - particularly in organizations that profess to operate by consensus. Committees have a hard time making decisions. When decisions can't be made explicitly, according to the social contract, they will be made covertly, and the confidence of the group in its process - and therefore its will to succeed - will be lost.
This problem also arose in Placeholder. There were two co-directors, with very different artistic esthetics. There were weekly design meetings. At the time of the crisis, there was no agreement between the co-directors. A coup d'etat eventuated, in which the technical staff (the sargeants) threw its support behind one of the directors (a captain), resulting in fast action on a unified design. The problem was the original structure of command and control - it was not unitary.
Designing the social structure of a design team should be done in such a way as to make the OODA loop work. This is best done by a hierarchical tree, with local autonomy residing in one person at every level. This should funnel to one person at the top. Discussion and feedback, both public and Japanese-style back-channel to preserve face, is important - but so is the ability for decisions to actually be made and acted on. The wise leader listens to her subordinates, and also is not afraid to do what she decides is correct - and to expect to be followed.
At the banquet at the end of the
Placeholder project I rose to my feet, and toasted the triumphant
director with the ancient salute of the Roman legions to a commander
they felt was worthy to lead them into battle - hail Imperator!
I am indebted to Brenda Laurel for the following insight concerning strategy and grand strategy.
Brenda is a Star Trek fan of great enthusiasm. Her favorite character is James T. Kirk. Kirk is famous in the grand narrative of Star Trek for finding "the third alternative" in win-lose situations. Brenda has been heard to challenge her staff to "find me that third alternative!".
A problem facing many designers is an ethical one. How can you work for a company that has a grand strategy that is problematic for you? Brenda's insight is that a designer can sign up to work on a specific strategy that can serve two different grand strategies - one grand strategy that the company desires, and a different one that the designer desires.
Consider a feminist working for Nike. Nike wishes to sell more shoes than anyone else. The feminist wishes for certain oppressed groups to increase their well-being in the world. These are different grand strategies. However, the strategy of designing an ad campaign to enhance selling running shoes to women can be enthusiastically shared by both - because it sells more shoes, and because it increases the physical well being of women to run.
I call Brenda's insight the Laurel Maneuver.
Action in the Polis (3) - Sanjuro... the ronin leaves town
My first really good design inversion came to me when I was working at Xerox PARC in the late 1980s. I had spent several years working on color printing, digital halftoning, image processing... pushing pixels. One day at lunchtime I was sitting in the cafeteria, reading Vallee's book on the sociology of belief in UFO's as one of my co-researchers excitedly exclaimed "...and we can hyper-link all the pages by putting a barcode on the bottom of every page!"
I had designed digital typefaces. Good type and the esthetics of page
design are important to me. So I looked up from my book, and said:
"That would be fugly!!!"
The insight came from my knowledge that halftoned pictures are made up of dots of various sizes - dots that are usually shaped as an ellipse, oriented at 45 degrees. This is done digitally in a way that apes the way it has been done photographically for 150 years. The key was that in the analog world the dots all line up in the same direction - but in the digital world, with control over every dot in ways not possible with an analog halftone screen, one could make the ellipses tile to the left and denote a zero - and tilt to the right to denote a one. This could embed information within an ordinary picture, at rather high densities, in a two dimensional grid (9).
The technology was the easy part. Getting a gigacorp to actually use it was the hard part. I also discovered that a senior researcher was stealing credit for my idea. Fortunately, I had documented my insight, and got the basic patents nailed down. But I grew dispirited. I was in love with the beauty and elegance of the idea, and struggled hard to find marketing reasons why Xerox should use it (this was my “Polish calvary” experience).
Finally I left.
But what I soon discovered was that it was not my last good idea - and that there were other places where the sword of my mind was valued. I joined Interval Research, where we were challenged to "do the next thing".
The lesson is that like Kurasowa's wandering ronin Sanjuro, you can leave town, and it's OK to do that. It's important to be passionate about your work, and true to yourself - and sometimes the best way to do that is to walk away from an employer. You always have to have a customer for your work - one that values it.
When Interval Research was founded the original 24 members of the research staff were challenged by David Liddle to "create something as different from the personal computer as it was from the mainframe".
I thought about this, in a structural manner, and concluded that the right way to proceed was by systematically inverting a number of the media theoretic elements of what a personal computer (or work-station) was (circa 1993).
Doing structural inversion on these elements, with the added heuristic to design to human qualities and abilities that had been "left out" of the user experience of computer use, leads to:
I properly noticed that this design exercise, carried through to its logical extreme, results in a crowd of unruly puppies. Since I didn't think Paul Allen wanted Interval Research to be a puppy breeding farm, I backed off a little, and decided to explore building mobile robots, that communicated emotionally with people and each other, and which existed in the same realm of the senses as people did. This was my entry into what is now known as affective computing.
I argued against simulation of such, because people would inevitably react differently to something on a screen as opposed to an entity with a body in space. Therefore building real physical robots was deemed essential in order to create a joined system where people were the environment for the robots, and vice versa, and where we could ask such questions of people as "what is the robot feeling"?
When the project was started, it was in uncharted territory - no one else was doing anything similar. The patent search for the background intellectual property uncovered only a few remotely relevant pieces of intellectual property - a flocking patent, a game patent, and a few others.
My grand strategic goal for the project was to do a strong design statement, that would produce design heuristics that others could emulate. Such a strong design statement could be subsequently relaxed from its initial purity of expression to find use in other areas, such as adding affective communication into workstations - in other words, one would not have to always make a full up robot to take advantage of the understandings produced by first building an emotionally communicating robot.
We started by exploring what were
minimal elements needed to make a face that people would experience
as having expressive emotions. We decided not to go for realism, but
for what amounted to a line drawing in aluminum and steel. Mark
Scheeff built what we called the Mark One "Severed Head" -
a cube with six degrees of freedom for two lips, eyes, and eyelids -
each driven by computer controlled servos. The Interval "Aquarium"
staff (our experimental psychologists) did user testing with naive
subjects who were directed to use a computer keyboard to drive the
face to achieve various emotions - happy, sad, angry, afraid, etc. A
surprise occurred when we observed the test subjects miming the
facial expressions they created on the robot's faqce with their own
faces! We used the resulting settings for the servos to build a
database of emotional control points.
The end result, four years later, was a prototype robot that had an expressive face, a mobile body with a articulated neck and head; gestures that emulated the ballistic motions of animal gestures; a sense of touch that differentiated between a caress, a light contact, and a blow (using sub-modalities of an accelerometer and a capacitive sensor); stereo color vision that located peoples bodies in space and tracked them via 3D blob analysis; had male or female gendered emotionally expressive audio - and implemented an internal emotional state machine that emulated the seven basic expressive emotions described by Charles Darwin in in The Expression of the Emotions in Man and Animals.
By 1998 others had started to work in the same general area. Some of these poured far more resources into their engineering - notably Sony with the "Aibo" robotic dog. In the annual project review I argued that we should either spin the project out of Interval as a development effort aimed at real products in the toy market with enough resources to succeed, or we should kill the project. In the end, I killed my own dog - the project was ended.
Mark Scheeff, the member of the research staff who did the mechanical design for the robot, followed on with a six month "cremator project" He ripped all of the AI out of the system, and turned the robot into a tele-operated puppet, to explore what people perceived about the robot, not knowing what it was "under the hood" - i.e, what they projected onto the media surface. He exhibited it at the Tech Museum in San Jose, and members of the project under his leadership published a paper summarizing the results from this exploration of projective intelligence (10).
The other tangible result from the
project was a very broad patent (and a subsequent improvement patent
with the same title) covering emotional communication between real
robots with actual bodies with each other and with people... where
the major prior art cited was Charles Darwin (11).
his sword for
(1) "Strategy", by Basil Henry Liddell Hart. Publisher: Meridian Books; 2nd edition (April 1991). ISBN: 0452010713
(2) “Disenchanted Night: The Industrialization of Light in the Nineteenth Century”, by Wolfgang Schivelbusch. Publisher: University of California Press; (1995). ISBN: 0520203542
(3) “Chimpanzee Material Culture: Implications for Human Evolution”, by W. C. McGrew. Publisher:Cambridge University Press; (1992). ISBN: 0521413036
(4) "The Art of War", by Sun Tzu, translated by Brig. General Samuel Griffith. Forward by B. Liddle Hart. Publisher: Oxford University Press; (June 1971). ISBN: 0195014766
(5) "On Deconstruction: Theory and Criticism After Structuralism", by Jonathan Culler. Publisher: Cornell Univ Pr; Reprint edition (August 1983). ISBN: 0801492017
(6) "The Closed World: Computers and the Politics of Discourse in Cold War America (Inside Technology)", by Paul N. Edwards. Publisher: MIT Press; Reprint edition (August 1, 1997). ISBN: 0262550288
(7) "Placeholder: Landscape and Narrative In Virtual Environments", by Brenda Laurel, Rachel Strickland, and Rob Tow. Publisher: ACM Computer Graphics Quarterly - Volume 28 Number 2 May 1994. See also: <http://tauzero.com/Rob_Tow/Placeholder/Placeholder.html>
(8) "Boyd: The Fighter Pilot Who Changed the Art of War", by Robert Coram. Publisher: Little Brown & Company; (December 2002). ISBN: 0316881465
(9) "Methods and Means for Embedding Machine Readable Digital Data in Halftone Images". Tow, R., U.S. Patent no. 5,315,098, May 1994.
(10) "Experiences with Sparky, a Social Robot", by Mark Scheeff, John Pinto, Kris Rahardja, Scott Snibbe, Rob Tow. Proceedings of the Workshop on Interactive Robotics and Entertainment (WIRE-2000).
(11) "Affect-based Robot Communication Methods and Systems". Tow, R., U.S. Patent No 5,832,189, November, 1998.