Apr 30, 2012

intentional ambiguity

Tenses are important:

abstract: This ethnography of nine internationally renowned avant-garde culinary groups in the US and Europe shows that they operate successfully despite having intentionally ambiguous (and hence uncertain and unstable) member roles and group goals. This is in contrast to the sizeable body of research suggesting that individuals and groups need clarity and stability to function effectively. I describe how this intentional ambiguity modifies group and interpersonal processes in these groups, then propose explanations for how these modifications might allow them to adapt to changing resources and environments and enable their members to teach and learn complex things such as house style. I conclude by showing how internal ambiguity supports adaptability and thus helps groups respond to external ambiguity.

TL;DR version: Ambiguity and uncertainty are not always bad. Wisely deployed, they can be beneficial for those in the business of innovation or working in unpredictable and rapidly changing environments. Which is nearly everyone working in a high-end restaurant, a startup, or an R&D group these days.

Just saying.

committee: Christopher Winship (co-chair), Amy Edmondson (co-chair), Jeffrey Polzer.

Apr 23, 2012

setting type

or, rather, TeX. or, actually, LaTeX. what does someone who doesn't write equations need latex for? the awful truth is: i switched to latex so that i could automatically set old-style numerals. also, it is always nice to separate content from representation and data from metadata (as anyone who has ever had to clean up a word document's formatting will tell you).

latex (built on tex) is a document production system. its chief virtue is in separating content from representation. there is no way to do this without a) thinking through the structure of the document, b) learning how to represent the structure abstractly (ie, structural markup), then c) writing documents separately from how they will appear in finished form (but marked up for structure).

every production system involves massive up-front investments of time (the so-called "learning curve") before the palmy days begin. latex was no different. i spent many tasty hours learning how to insert images, write tables, rotate tables, control linespacing, etc. then, for more typographical control, i added the fontspec package and xelatex. this broke some environmental parameters, which required debugging. did i mention how some typefaces don't have correctly coded glyphs and don't work as expected? (= more debugging). then i decided to go for broke and move my citation database to bibtex. this consumed many more tasty days. now my papers look great, i cite effortlessly, and it's usually all just peachy.

my latex distribution is miktex (2.9). my typeface of choice is minion pro (it handles oldstyle numbering, is readable in block text), and i use the article document class. to handle graphics, i use the graphicx package. for typography, the fontspec package. for tables, the tabulary package. for citations, the natbib package. other miscellaneous utility packages include rotating (to rotate things like tables and figures), booktabs (for high-quality table formatting), and endfloat (to automatically move all figures to the end of a document; endfloat requires a separate configuration file in the source document directory if it is used with rotating).

i renew the quote environment to change the linespacing inside blockquotes. i manage my citation database with jabref and add new citations either by hand (if my citations come single spies) or by importing from zotero (if they come in battalions). i compile by running xelatex, bibtex, then xelatex twice more. (it took a long time to figure out how to get the commands to call the right input files, and that the \bibliography command requires filenames with extensions on some operating systems and not on others.)

then, a dark moment. in the late stages of preparing a manuscript for submission last week, i ran into a problem: the journal calls for an idiosyncratic reference style and provides no style file. two options: rewrite a style file to suit, or generate a new one. let all of us who have been in a similar quandary but lack extensive experience with stack languages thank the creator of this custom reference style maker. you answer a string of questions, it generates a file that you can then compile into a style file (.bst). this took a bit (well, no, let's be honest: hours and hours) of tweaking to get 98% right. (remaining 2%: do you know how i can force a linebreak between the author block and the year block and a tab character between the year block and the rest of the citation? if yes, please email me.)

"triumph!" i thought. but wait, .pdf files are verboten for this journal's submission machinery. fortunately, .ps files are ok. a simple matter, you might think, to take the .xdv files from xelatex and use dvips to produce a .ps file. except dvips doesn't work with .xdv. xpdf's pdftops converter saves the day (though erratic, it is the least temperamental and the one that copes best with transparency in the source pdf).

there must be a better way.

Apr 22, 2012

the argument for quality

Artists need to be in there from the start, making the argument for quality. The key to this thing is, for example, if you give an engineer a set of criteria which does not include a quality quotient, as it were—that is, if this sense of the quality, the character of the place, is not a part of his original motivation—he will then basically put the road straight down the middle. He has no reason to curve it. But if I can convince him that quality is absolutely a worthwhile thing and we can work out a way in which the road can be efficient and also wander down by the river, then we essentially have both: he provides quality in that the road works, I provide quality in that it passes by the river. In that way, art gets built into the criteria from the beginning rather than being added on afterward. In many cases, it may not cost any more, or just a bit more, for the road to wander by the river. As opposed to, say, giving the artist one per cent afterward—which is, hell, just tokenism to the nth degree. The best they’re going to be able to do is put in a few doodads. But if you affect this whole process from the beginning by putting in place some quality criteria and you look around and ask yourself who in this society are trained to make this argument for quality, the only ones that I can think of are artists. We’re the only ones with no real rationale for being except developing aesthetics, or quality—we have no other function. So our key role in our society right now—and what we’re really talking about here is translating values into dollars—is for artists to make winning arguments for why considerations of quality are absolutely necessary.
robert irwin
in lawrence weschler's new yorker article "in a desert of pure feeling" [PDF

Apr 20, 2012

gestural engineering

i'm reading, as i do every year, seeing is forgetting the name of the thing one sees, which is both a biography of robert irwin and an extended discussion of what it means to think about making a thing. the book is rich with interesting stuff for anyone willing to spend some time with it. it also represents a minor stream of thinking which currently appears scattered and diffused across many fields (computer science, organizational theory, design, manufacturing, and others): accounts of grasping not for mastery of a concrete thing but the abstract generative principle that produces concrete things of that class.

for instance, take arthur ganson. ganson is the MIT sculptor in residence, and the MIT museum has an ongoing exhibition of his work (note for cambridge residents: admission to the MIT museum is free today, 4/20, as part of the cambridge science festival). this is ganson's TED presentation (the embedded video will start at the beginning of the section transcribed below):

just as in robert irwin's work, there's a sense that there is a platonic ideal concept straining to emerge, that the process of making is a process of getting closer (but never close enough) to the realization in the real world of the idea in the mind:
i imagined a very simple gestural dance that would be between a machine and just a very simple chair. and when i'm making these pieces, i'm always trying to find a point where i'm saying something very clearly and it's very simple, but at the same time it's very ambiguous. i think that there's a point between simplicity and ambiguity, which can allow a viewer to perhaps take something from it. and that leads me to the thought that all of these pieces start off in my own mind, in my heart, and i do my best at finding ways to express them with materials. and it always feels really crude, it's always a struggle, but somehow i manage to get this thought out into an object. and then it's out there. it means nothing at all, this object just means nothing. but once it's perceived, and someone brings it into their own mind, then there's a cycle that's been completed. to me, that's the most important thing, because being a kid i wanted to communicate my passion and love and that means the complete cycle of coming from the inside, out to the physical, to someone perceiving it.
ganson has also posted footage of the performance of machine with chair, which includes the setup.

Apr 17, 2012


my photographs have been influenced by those of ben sherrill.

Apr 15, 2012

the myth of risk reduction

garry kasparov, considering the slowdown in technological development*, writes that

People have a sense that we should receive benefits from our investment, but need to reduce the uncertainty. Risk should be less, but the income should be the same. This creates a gap, because in a free society, in a market economy, there is a direct relationship between risk and return. If you want to avoid risk, but receive ten percent of of your annual income from your investment, you open the way for the so-called financial engineering. In fact this is all fake, not real income, because it is not based on real changes in the economy, because you do not create new and tangible assets. In the 1960s young boys dreamed of becoming aerospace engineers, now they want to be financial engineers, working in investment companies, which are the most attractive spheres for talent. This naturally affects the quality of the total scientific potential, because financial engineering creates nothing.
he's right, of course, though it would have been nice to distinguish between risk and uncertainty as frank knight (1921) did:
Uncertainty must be taken in a sense radically distinct from the familiar notion of Risk, from which it has never been properly separated.... The essential fact is that 'risk' means in some cases a quantity susceptible of measurement, while at other times it is something distinctly not of this character; and there are far-reaching and crucial differences in the bearings of the phenomena depending on which of the two is really present and operating.... It will appear that a measurable uncertainty, or 'risk' proper, as we shall use the term, is so far different from an unmeasurable one that it is not in effect an uncertainty at all.
the return on any investment (or, more generally, the outcome of any action) comprises three conceptually distinct components: a known part, a knowable unknown part, and an unknowable unknown part. the knowable unknown is risk, and risk can be "reduced" or "managed" by spending effort, time, and money in advance: gathering information and preparing defensive assets of a known nature and quantity.

the unknowable unknown is uncertainty and is not reducible by advance action. the only way to "manage" uncertainty is to be ready to respond to conditions as they change unpredictably: this is costly in a different way since it requires creating a potentially enormous reservoir of excess capacity with which to respond. maintaining this excess capacity is expensive and may not pay off in the short or even the medium term.

responding to risk and responding to uncertainty require different approaches and different processes. all innovation requires both types of response. innovation in rapidly and unpredictably changing environments requires more preparation for uncertainty than risk. surprise! this will be my dissertation.

* found, since i read chessbase only infrequently, through tyler cowen's marginal revolution.

Apr 14, 2012



Apr 6, 2012

what is design?

design is not about making things pretty, nor is it all about problem-solving (though some people think so).
  1. design is problem-finding first, and only then problem-solving.
  2. if design is problem-finding first, then "designer" isn't necessarily a specific role played by specific people on a team. design thinking (see below) is a way to approach the world that every constituency and individual should possess, so that they can figure out the right problems to solve.
  3. sometimes, finding the right problem requires lots of search and research. but this isn't just reading papers and essays. research is also (maybe more about) being in the world, exploring it, and understanding it.
  4. most importantly, if design thinking is deciding what the right problems are to solve, then design at base is an issue of figuring out what values we have, what makes a problem the right one to put effort behind. values aren't objectively evaluatable, they cannot be commensurated. all you can do with values is state what values you have and be open to changing your mind.
So, what is design? (originally posted 9 september, 2009)
Though it's tempting to think of design in terms of final output, say a chair or piece of software, it is more useful to think of design as a process that leads to an outcome intended to solve a problem. The kinds of problems to be solved can be in nearly any domain—physical objects, processes, interactions, foods, intangible systems, data representations, institutions, organizations, public policy all can be developed using a design process, though the concrete details of the design process will vary depending on the domain.

Design as a process
While it is possible for, say, the design of a more efficient dialysis pump or a better system for mobilizing non-voters during an election year to materialize fully-formed from a designer's head, this is a rare occurrence. More often, the design process iterates through several stages. You can begin this process at any one of the stages (and go in any direction). The process ends when the designer is satisfied with the outcome (or loses patience).

Defining the need (or problem-finding) often involves identifying problems to solve, problems to not solve, and criteria for judging solutions. Problems can be defined at many levels of abstraction and frequently interact with each other; the outcome of a design process reflects the constellation of problems driving the process. For example, Gmail (the webmail service) resulted from a process driven by numerous problems ranging from relatively abstract ones like "How can email service be improved?" to concrete ones like "How can webpage programming languages like HTML and Javascript be modified to make a web application feel like a client application to the user?" or "Folders are silly. What do we replace them with?" Often, research, prototyping, and testing will indicate either that that the problem needs to be redefined, or that solutions to the original problems posed generate new problems.

For the designer trying to solve a given set of problems, the universe of possible solutions to a given set of problems is infinitely large and insufficiently understood. Research helps begin the process of increasing knowledge of possible solutions while also eliminating impossible, unfeasible, inelegant, or problematic solutions. Conducting research takes different forms depending on the kinds of problems faced and the degree of design iteration that has already occurred. Some examples of research include market studies, patent searches, product testing, conference attendance, and field studies. The objective of research in the design process is to leave the designer with a smaller, more probably viable, set of possible solutions to dissect, recombine, and evaluate.

Often, evaluation is easier and more effective when possible solutions are given a more concrete form that designers and testers can interact with—a prototype. Prototyping can range from developing a policy proposal, to making wireframes (mockups of how webpages or software applications will look and the order in which users will encounter them), to 3D and physical models of varying degrees of precision. Prototypes made early in the iteration process tend to be simpler, faster, and cheaper than those made later in the process. Building prototypes used to be expensive and slow, and required extensive equipment and/or specialized knowledge (hence the proliferation of proof printers, model shops, one-off precision milling shops, and the like). Fortunately, fast, inexpensive prototyping software, hardware, and facilities are becoming increasingly available. Fast, inexpensive prototyping accelerates the design iteration process significantly, since prototypes are frequently the source of rich insights for designers, particularly in fine-tuning elements of the product and in recognizing new problems that must be resolved.

Design prototypes are evaluated against the solution criteria established in the design process's problem-definition stage. Designers collect information about the design's performance and use this to evaluate the design; in other words, they try to understand if and how the design should be modified in the next iteration of the design process. Particularly in the case of software applications, it is becoming increasingly clear that designs cannot be considered complete or effective if they do not include methods for collecting information needed to evaluate the application's effectiveness in solving the design problem. As with prototyping, gathering information about design performance has become easier, less expensive, faster, and more comprehensive for many categories of products.

Apr 3, 2012

les jardins

a box of macarons from pierre hermé landed, unexpectedly, in my hands last night.* macarons are a bubble which has not yet popped. a few years ago, most people wouldn't know them from a macaroon. today, many places with aspirations to quality (and some without those aspirations) manufacture and sell them. for a while, mcdonalds in france sold macarons.

macarons are little sandwiches with cookie shells made of ground and blanched almonds held in a flavoured meringue, filled with a flavoured ganache or cream. it is easy to make a mediocre macaron.

the shell is notoriously finicky. the meringue must be robust enough to support the almond flour during baking, requiring the maker to detect and respond to very small gradations in the texture of the uncooked meringue. this in turn responds to very small gradations in ambient temperature and humidity and the hydration of the meringue. macaron-making thus features many arcane directives. some recipes call for aged egg whites (which have lower water content due to evaporation). many formulations also call for uncooked macaron shells to be set out for a while to dry further and develop a skin on the top surface. this skin constrains the cookie as it rises during baking, producing the desired smooth, glossy top and rough-edged foot.

fillings should have flavours and textures that complement the shells. overly aggressive flavours are unfortunately common, but texture seems to be the bigger hurdle. many fillings in inferior macarons are poorly matched to the texture and robustness of their shells. biting into a robust shell causes a too-fluid filling to squeeze out of the sides. conversely, a delicate shell collapses too quickly around a too-rigid filling. both scenarios are untasty. (not incidentally, the same principle of textural harmony applies to sandwich construction and design.)

inevitably, average macaron quality declines as supply of macarons increases. i am happy to leave most macarons on the shelf, for those who would exchange their hard-earned doubloons for dense (or over-airy and/or sticky) shells and tasteless (or overly flavourful and/or overly avant-garde and/or poorly textured) fillings.

on the other hand, it is a good day when life hands you a box of macarons from pierre hermé. their shells have perfect texture, density, and surface, and just enough understated filling of precisely the right texture (some traditionally flavoured, others quite unusual). naturally, he is Big in Japan. confiserie sprüngli, conveniently close to the main train station in zurich, is another place to get very correct macarons of a different style. (and different name. these luxemburgerli are smaller, with a more fluid filling and a more delicate shell, and more traditional flavours.)

a great macaron is usually a rare and costly experience.

* thx journeyman; welcome home.

Apr 2, 2012

a prescription for a grand cuisine

A prescription for a grand cuisine: Take a vigorous people, give them an empire, let them get accustomed to the spoils, and then take it all away again ... bit by bit. Palates accustomed to good things will fight—hard, then harder still—to keep them, or to find replacements. Luxuries now rationed will become distilled. The ultimate cuisine may not be as luxurious as it was in its "glory," but it will be more inspired, more various, and more complexly realized. Roman cooking was lavish; Italian cooking is great.
john thorne and matt lewis thorne, "la cuisine creole," in serious pig.