Why Redesigns are So Damn Hard

Redesigning social functionality – that is, re-doing a social product already in use by some sort of entrenched audience – is sometimes tougher than designing from scratch. Very often you're fighting against organizational and institutional inertia; legacy issues that determine what's actually possible; entrenched user expectations; nervous stakeholders demanding guarantees; and a little understood, fickle market always on the verge of catastrophic "creative destruction".

But what interests me here, is that in my experience the difficulty of redesigning a product – both the subjective experience of effort and the objective expenditure of time and attention in completing a redesign, start to finish – seems to vary with the product's success, but not in a particularly straightforward manner. That is, how successful the product to be redesigned currently is affects the difficulty of the process in a complex, non-intuitive way. Why?

Well, first of all there are at least three big classes of factors impacting the difficulty of redesigning a site. Some of these are important in designing the site from scratch, too, but the way they interact with the current success of the product when embarking on a redesign is what interests me. The three factors are Market Information, User Entrenchment, and The Organization.

Market Information
When redesigning, we look to the market for clues, approach our users for tests, and effectively try to tease out guidance from the field. Ultimately, we're looking for the formula for design success. But this stuff is incredibly complicated, including such things as the design conventions determining current user expectations; the competitive landscape identifying areas of relevant differentiation; trend dynamics suggesting potential "innovations"; etc. But the most subtly difficult factors are the complex social dynamics of product acceptance. I've written about these before: the hard to discern workings of network effects, informational cascades, and the resultant path dependencies cause enormous confusion among stakeholders, greatly increasing the difficulty of redesigning social functionality. In our misguided search for a simple formula for redesign success, these factors, which decrease the role of product qualities (the stuff of design) in predicting product success, can really lead you astray.

In a nutshell, the externalities inherent in the social web (i.e. we initially evaluate a product based on others' actions and then get the most value from the products that everyone else uses) make predicting success from product qualities almost impossible. Success is usually a matter of context, timing and accidents of the product's acceptance history rather than objective design quality. So, beyond a set of user expectation-defining standards – the competitive baseline, which you should have already identified in the initial design round – market analysis and user research won't help you identify a redesign guaranteed to succeed.

Even worse, up to a certain point, the more relative success you have, the murkier the information derivable from testing users and combing the market becomes. Your variance from the competitive baseline tends to be smaller the more successful you are, and the true drivers of success are rarely clear cut product design issues beyond this competitive baseline. Very few stakeholders understand these subtle dynamics and insist that more market research, testing and "innovation" will crack the "guaranteed success" design code. As a result, redesigns of moderately successful products tend to bog down in obsessive, inconclusive research, the results of which are infinitely and inconsistently interpretable (especially if testing is public). Research is necessary, but will never provide a guarantee.

User Entrenchment
Perceived usability – as opposed to laboratory usability – is more about familiarity than objective human factors metrics. Users invest their time and attention in your product, amassing a sort of practical capital of product-specific know-how. When you start changing things, they view this as effectively theft of a precious resource: their investment in practical capital has been thrown out the window without their explicit blessing.

As anyone who has ever redesigned anything knows, users get angry when they feel they've wasted time and attention; they must reinvest to get back to the same level of practical capital. So, up to the point where nearly everybody's using your product and there's no real alternative (i.e. where the network effect is so strong and the average user's "sunk cost" so great they overcome practically all anger over perceived losses associated with change), redesigns get dicier the greater the success of the product. One consequence of this is that testing redesigns with current users will significantly skew your results toward the negative, regardless of "actual" quality. If these tests are public, you've just created a significant political battle and, thus, headaches.

The Organization
There are two related but distinct sources of organizational inertia. First, people within the organization simply get used to doing things one way and – being human – are reluctant to change. If the product's relatively successful, that felt reluctance can be rationalized by arguing that any redesign is a dangerous and unneeded rocking of the boat that endangers continuing success. Slowdowns, endless discussion, and frequent miscommunications are usually the result.

Next, as much as I hate calling simple regularities or covariances "laws" for comic effect (Moore's Law, Metcalfe's Law, etc.), Conway's Law, or something like it, seems to be pretty prevalent. In 1968, Melvin Conway observed that organizations produce products that mirror their structure or at least their internal communication patterns. He was referring specifically to the development of intercommunicating software systems produced by different design teams existing in some sort of institutional, social or communication structure. His claim is that this structure will manifest itself in the design of the interfaces between the systems.

Conway thought this was inevitable, which is way too strong by my lights. But stretching the idea a bit, some products, particularly websites, often mirror the structures of the organizations that created them. Furthermore, it seems to go the other way as well. Sometimes the organization and product have co-evolved in such a way that they're truly intermingled or even identified in many stakeholder's minds. Stakeholders view the product as a direct manifestation of the organization, and any change to the former necessarily impacts the latter.

Whatever the underlying mechanisms that generate this feeling, the upshot is that redesigning a product often has organizational implications, at least in terms of many stakeholders' perceptions of the project. Tampering with the product is often perceived as tampering with roles, responsibilities, and the delicately negotiated distribution of power within the organization. Understandably, this can lead to a significant amount of internal resistance, particularly when the product has been relatively successful and actual power and prestige have accumulated. The resulting inertia and sometimes downright sabotage can lead to massive struggles and delays.

These factors interact with the current success of the product (and with each other) to impact the difficulty of redesigning the product. Drawing from my highly subjective experience designing, redesigning, and "researching" social functionality, the situation seems to look something like this:


On the x axis is increasing success and on the y axis, increasing difficulty of redesigning the product. Some of the more interesting areas have been called out with letters. I'll discuss them briefly.

Why is a more difficult than b? After all, nobody's using the product, so redesign should be really easy. At this level of market failure, if you're asked to do a redesign, you should consider it carefully: this product should be completely abandoned and a different product built. But if you have to do a redesign, the whole organization probably understands and embraces the need to change and analyzing the market may help you identify appropriate conventions, standards and design directions. However, you are re-starting from a hole with little prospect of success (remember the importance of network externalities). Your potential users have already invested somewhere else (this a redesign… you had an initial chance and blew it. At this point your users have invested elsewhere and people burned earlier won't come back.). Finally, you have absolutely no user data on which to base your redesign recommendations and no prospect of launching then optimizing, bootstrapping your way to a successful design. After all, there's not enough activity to get a picture of where it's failing or succeeding; analyzing the market will suggest directions, but you can't really put it out there and optimize on the fly as no one's using your product.

As success modestly improves to b, the ease of redesign increases dramatically. At this point, your stakeholders are still ready for change. Analyzing the market could probably still help with design standards, conventions and trends. But now you have some data on what might work and what definitely won't. You have real users that you can gently test against, but not so many that changes will be met with a loud protest. You also have at least a foothold, which, with luck could be turned into something more by doing everything you can to generate an informational cascade through targeted design differentiation and smart positioning/marketing.

But things get progressively tougher from b to c. You're probably nearing the conventional designs as you're at least competitive at this level of success, so analyzing the market too much could lead to frustration and confusion among your stakeholders. Selling designs internally becomes more difficult as the product's modest success has been enough to spread a bit of influence and power throughout the organization. But, there are enough users to make switching costs to another product at least non-negligible to many users. However, the User Entrenchment process begins to become an issue at this point, as you've enough success for your users to have begun internalizing the system.

From c the difficulty quickly rises with success, slowing to a peak at d. This is the zone of incredible difficulty for a redesign; throughout this region all of the factors are potentially against you. Unless you're just adding hot new functionality pioneered by a competitor, culling the market for tips will be largely fruitless. You're most likely conventional – if not the leader – at this level of success, so most of the factors determining further success are those frustrating, inconsistently interpretable, and highly contingent social dynamics. Thus prolonged research and analysis in this situation actually pays little – and often costs much in terms of focus, morale and transactions. But most organizations in this range still get hung up searching the field and bugging their users in an obsessive search for a magic design formula when they should just be looking for suggestive trends to fuel experimentation. But you have to remember, users in this range become ever more entrenched, so the "right" changes are both difficult to identify and nearly impossible to justify through testing or interviews. Users want what they know now, not what they will want later, so at high levels of entrenchment asking them to evaluate a change (as opposed to just testing lab usability) often confuses the issue or leads to the agonizing death of bold new designs. Finally, organizationally, great success can make people scared to change for fear of losing it (people are, after all, far more risk averse than gain hungry). The organization has "rigidified." Success has pumped prestige and power into it's structure, creating significant vested interests in maintaining status quo. Projects in this range can quickly become nightmares.

At d, however, things turn around a little. Though it's never as easy as it is when you've poor to moderate success, the network effect is so strong, that the specter of losing position is largely mooted. That effectively removes the rationalization for status quo from within the organization and greatly mitigates the sting of negative test results from entrenched users. Also, at this point, you lead the market and most of the tips you're looking for are in terms of competitive and trend analysis, not the elusive success guaranteeing design formula. You largely co-determine the conventions defining the completive baseline along with other highly successful players. Thus, redesign at this point becomes slightly easier, but there's still there's a lot of reluctance from those who might fear loss of position and entrenched users will definitely make a lot of noise, which is never pleasant.

Redesigns are necessary to stay competitive. But along with the excitement and product / organizational rejuvenation they generate, they can also be headaches for the designers involved. Hopefully I've managed to shed some light on why and how. As I said, all of this stems form my subjective observation that redesign difficulty seems somehow oddly related to product success. Whether or not your experience of the relation between the two mirrors mine, I hope you'll at least agree that there's some relationship and that it's impacted by the factors I've suggested.

Playing Nice For the Wrong Reasons

We were sold the story of being mainly self-interested, mainly rational actors interacting in market places. And the internet has shown that we have all these social, empathetic relationships with deep, authentic motivations that are nothing to do with selling and spending.

Clay Shirky

The Observer, Sunday 15 February 2009

One hears this sentiment a lot these days, particularly from American web pundits. It sounds like a lingering echo of the utopianist strains of early web propaganda; it's a rhetorical move positioning the web as a signal force in a new flourishing of hierarchy-smashing communitarianism against the old alienating and atomizing intellectual myths of “maximization” and “rationality.” The idea's intellectual sources include the recent widely touted corrections (e.g. behavioral economics) of some of the excesses of neo-classical economics and possibly the U.S. sociological tradition stemming from “functionalism.”

Anyway, I agree wholeheartedly that we often cooperate or collaborate on the web out of more or less benevolent – if not truly altruistic – impulses. But it’s also obvious that “deep, authentic motivations” aren’t the only incentives for acting in, or even just joining, groups. Significant “shallow” and “inauthentic” motivations impact social behavior as well. “Ignoble” motivations like esteem accumulation and status achievement – which are both distinct from reputation – significantly impact people’s behavior in social settings. Indeed, much of the functionality on the social web already has an esteem function baked into its primary or legitimate function. As a simple example, posting reviews is as much about displaying expertise or simple likes and dislikes as it is about helping others choose. As a rule, people genuinely like to play nice and help each other. But they’ll play nicer and help more if you also let them compete for group defined goods that are neither “authentic” nor “deep.”

Sticking with Clay Shirky, in his latest book he recognizes that there is a place for some sorts of less than noble rewards online. For example, he considers “vanity” and reputational benefit to be valid motivations for contribution. But the mention of reputation notwithstanding, he focuses on intrinsic, non-material motivations like self-esteem, the desire to produce something good and the need for communion. What he doesn’t talk about are the powerful extrinsic, non-material motivations involving the attention of and positive evaluation by others in your group. You contribute partly for the esteem your contribution can get you. And this holds true even if the esteeming group is minuscule. Indeed, it’s often the case that the smaller the esteeming group, the more valuable the esteem (“selling out” after all is the trading of esteem for popularity).

If we’re honest with ourselves, it seems pretty clear that some of our motivations for social interaction on the web (and elsewhere) are self-interested, operate by some sort of market principles and are neither authentic nor empathetic. We engage in behavior that is intended to "maximize" some in demand good – esteem – relative to the costs we're willing to bear. Other things equal, the greater the potential esteem the more cost we're willing to bear. Esteem seeking definitely isn't ideally authentic behavior: you can get it only to the extent that it’s not apparent – to yourself or others – that you’re actively seeking it. And although esteem seeking involves empathy – it assumes taking the point of view of others in order to determine the most estimable move – it’s not an “empathetic” motivation in the laudatory sense intended by Shirky. But, of course, this doesn’t mean that we don’t genuinely like helping and hanging out with others. It just means we have both sorts of motivations.

Shirky clearly recognizes this fact since he qualifies the whole thing with "mainly." Most likely, he just wants to make the point that we often act socially for the non-optimizing, genuinely pro-social reasons we say we act. Our actions are generally genuine and not cynical. But this is the tricky part: I don’t think that recognizing the importance of esteem to most contributors automatically commits us to cynicism. Furthermore I don't think that the distinction between Shirky's "good" motivations and my market-like esteem considerations is all that clear and easy to maintain in the first place. Esteem, like self-interest generally, is what Philip Pettit and Geoffrey Brennan call a standby or virtual cause of behavior. It’s not what’s directly sought from your actions, but if esteem wasn’t provided, you’d be less likely to behave that way. It’s a bias that steers us rather than an explicit principle that guides us. Esteem – the positive evaluations of others by the norm-based standards of whatever reference group you’re using – is the emotionally powerful implicit incentive within social groups that maintains conformity while allowing constant competitive evolution. Seeking it isn’t necessarily cynical, rather it's inextricably woven into group-focused behavior.

Why on earth are we so afraid of “rational motivations” that we have to banish them almost completely from talk of group action online? Beats me. Anyway, as someone who has to design interaction spaces online I think it’s dogmatic, maybe even superstitions, to think that esteem motivations are somehow less real or powerful because “inauthentic” and self-serving. After all, in public goods experiments, we are shown to be conditionally cooperative, meaning that we cooperate with a self-serving bias. If norms of cooperation or reciprocation aren’t sufficiently salient or aren’t otherwise maintained, we tend to stop playing nice and settle for getting all we can. The desire for esteem and status can actually get people to observe and stabilize cooperative norms for purely self-centered reasons; they’re self-serving incentives for pro-social behavior . We should design accordingly.

In the next couple of posts I’ll look more closely at esteem seeking, Shirky’s book and the popular bias against instrumental motivations online. I’ll use the example of the development of the Linux operating system to illustrate the distinction between two different versions of social capital as well as two different takes on the free rider problem. Of course, both of them will utilize the idea of esteem along with Bourdieu’s idea of the “economy of symbolic goods.”

Defining the Situation: Goffman, Social Functionality and Minicultures

When an individual enters into the presence of others, they commonly seek to acquire information about him or to bring into play information about him already possessed... Information about the individual helps to define the situation, enabling others to know in advance what he will expect of them and what they may expect of him. Informed in these ways, the others will know how best to act in order to call forth a desired response from him

That quote is from Erving Goffman’s classic, The Presentation of Self in Everyday Life. The book's basic claim is that human interactions are mediated by stagey appearances through which we imply – and from which others infer – our expectations, preferences, status, etc. We assume roles, staging presentations of ourselves, to ensure smooth, appropriately stable and beneficial interactions. The roles coordinate, structure and define our interactions as we negotiate a mutually agreeable definition of the situation.

Goffman was talking about face-to-face interactions, but I clearly agree with his main point as it pertains to digitally mediated interaction as well. When engaging with social functionality, we tend to throw out signals, intentionally and unintentionally, that indicate our expectations, preferences, etc. (i.e. "define the situation" in Goffman's terms) and coordinate our interactions with others. Display – a dimension of social functionality discussed in previous posts – is often about projecting a desirable position or estimable image, putting us in a position to get the most from our interactions. We then have to live up to or act consistently with that situation-defining image on pain of embarrassment, ridicule and situational discomfort.

However, situation definition in social interactions generally isn’t only determined by the people interacting. That is, there’s usually an externally imposed limit on the available “definitions.” Defining an interaction situation is sort of like declaring to each other what game we’re playing. I suggest, “Let’s play football” and you agree. We now roughly know what we can and can’t do. Our available strategies are bounded by the rules, but not wholly determined. We can still surprise each other and gain advantage within the game’s constitutive rules. Analogously, many interactions are about settling on the game to be played. But I claim the majority of interactions are about jockeying for advantage within an already determined meta-game. So, to sharpen the insight from Goffman's quote, I suggest that many of the broader aspects of most situations are already defined for us and the remaining definitional negotiations tend to be about fine-tuning or determining advantage. That is, we assume roles that are already defined and relatively definitive, we don't usually create the roles.

This is similar to what the philosopher of social science Don Ross suggests as the solution to the problem of game determination. In many social situations, culture handles this coordination: we more or less understand the expectations, obligations, etc. of the folks assuming the various roles in particular contexts because we’ve been conditioned to understand them. Understanding this stuff via conditioned expectations gets us past the cognitively intractable task of having to always figure out what game we’re playing first (i.e. what the best plan is given what others’ probable moves will be). In short, culture coordinates us on the arbitrary but stable games to be played while avoiding the destructive “state of nature” games in which we approach every interaction as if it were a mixed-motive game like the Prisoners Dilemma.

Designing as Defining

Social functionality, or the largely digital mechanisms that impact, amplify or alter our socialization online and off, are really weird if we think about them in terms of the ideas above. They’re generally anemic compared to face-to-face interaction spaces, providing only rudimentary means of interpersonal situation negotiation. Also, the cultural cues for defining situations – conventionally encoded in everything from architecture to tone of voice to uniforms and styles or dress, etc. – tend to be either missing or unconventionally signified. Social functionality tends to provide very limited and very strange means of “defining the situation” in the Goffman sense.

And even though recent public goods experiments show that people want to cooperate, it’s with a self-serving bias and conditional on others doing it in a way that can be observed. Reciprocation, cooperation and general decency tend to decrease on average over the long haul if not “structurally maintained”, i.e. if norms aren’t made salient. In situations involving social functionality, where there are few recognizable cultural institutions and coordination among individuals is tough, things can get messy quickly, devolving into potentially destructive mixed-motive situations where everyone acts like a selfish jerk.

Unlike the real world, though, systems incorporating social functionality have a third party in on the situation definition: the designer. Designers have a significant impact on the definition and – most importantly – maintenance of the situation. Design can help to stimulate the development of norms and the cobbling together of a simple “culture” within spaces employing social functionality. In particular, if the social functionality system isn’t a) structured for norm development or b) partially mapped onto an independently existing, institution embedding cultural group, it will often tear itself apart.

For example, in terms of structuring for norm development, Wikipedia’s use of Display inflected functionality coupled with iterative editing and negotiation mechanisms allows norms (in the sense of mutually held expectations of others’ expectations) to stabilize very quickly. This in turn keeps the majority of contentious entries from erupting into wholesale warfare. That is, the interactions are defined in terms of – bounded by – group developed norms of conduct and quality, the development of which was fostered by certain design decisions (essentially involving Display functionality, interestingly enough).

Wide open social spaces like MySpace and Facebook, on the other hand tend not to have this focused structure. Yet they also rely on a combination of positive and negative feedback mechanisms to help coordinate on norms. In particular, content can be tagged offensive or great; users can gain reputations and status; or they can get flagged and barred.

But stupid, antisocial behavior still arises because of the massive volume of users, the low join-and-drop overhead and the ease of anonymity. So, in terms of the second situation-defining design trick, I think we’ll find that the amount of disinhibition is inversely related to the closeness of the system’s mapping onto “real-world” networks. That is, if the system somehow extends or amplifies offline socialization, situation definitions from the real world can come to structure the interactions in the digital system. Situation definitions must be maintained if your real world friends are mixed in with your more tenuous “online only” friends. Similarly, but to a much lesser extent, systems that allow users to institute voluntary, displayable assortation – or self-grouping that can define some sort of in-group ethos – will also see significant norm development and stabilization.

Lurkers Aren’t Necessarily Free-riders

These are pretty obvious observations and have been put forward elsewhere, although without the theoretical structure of situation definition in terms of system-specific mini “cultures.” What’s interesting is that this allows us to look at things like “lurking” in a totally new light. Situation definition isn’t all about curbing jerky behavior. It’s also about determining appropriateness of interactions in subtler ways.

A while ago, people who cared about this stuff considered lurking – that is, partaking of the content on social sites without contributing any yourself – to be a from of free-riding on a public good. Lurkers were reaping the benefit without contributing to upkeep. Recently, however, some writers have begun to think of lurking in terms of cultural capital. On this view, many people lurk because they’re accumulating cultural capital, i.e. the skills, codes, knowledge, etc. that a user has for interpreting cultural artifacts. A significant proportion of lurking is about the user trying to figure out the community norms.

In our terms, lurkers are uncertain of situation definitions and the areas available for further negotiation. That is, they’re uncertain of the expectations of others and the proper way of expressing their own expectations and preferences so that will be understandable by others within the system and thus potentially beneficial to themselves. They don’t know how to act to coordinate on situation definitions that will be mutually understandable, useful and beneficial... or at least won’t result in discomfort.

So, designing for the emergence, stabilization and display of norms within systems utilizing social functionality can both curb disinhibition and more quickly convert lurkers into actors. Now, getting designers to realize that this is part of their job in a more than just “do what’s worked in the past” way – getting them to realize that social functionality design is embedded in and impacts cultural as well as social systems – is the really hard part.

Why Some People Are So Pissed About the Facebook Redesign

For the last couple of months there has been some fuss and furor over the Facebook redesign. Some love it, most don’t care, but a vocal, petition-waving few really, really don’t like it. Frankly, I think it’s a pretty good redesign. Most importantly for me, it cuts through the haze of humorous/cutsey/unused applications, placing them under a noncommittally named “Boxes” tab. Good move, I think. The majority of Facebook Apps remind me of the bulky, holstered PDAs that dangled predictably at every MBA’s hip a couple of years ago: a clumsy accessory masquerading as an interesting device. Of course, that’s what Facebook Apps really are, but I guess it’s just not for me.

Anyway, what interests me is the vehemence of the reaction and what it says about the idea of usability and the nature of design lock-in. In particular, the furor highlights the inadequacy or partial irrelevance of “objective” standards of classical usability. Obviously a considerable number of people feel that increased effectiveness, efficiency, learnability, etc. – whatever items comprise your favored usability checklist – isn’t an adequate rationale for changing a design mid-use. Nor do the measurable usability improvements seem to add up to satisfaction. Usability as preached by the industry is clearly distinct from the preferences that arise within the dynamics of actual use.

That last sentence is a mouthful, but it’s really just an observation that “usability” in the real world (i.e. the perception of comfort, effectiveness and satisfaction with some system) is as much about familiarity with arbitrary, usually suboptimal conventions as it is with traditional human factors issues. Given a choice in a non-laboratory, not “controlled for” situation, people generally choose what they know and understand over what’s new but ostensibly optimal. No matter how much the new one latches on to the tested realities of the human perceptual/cognitive machinery, people prefer what they already know.

So, one must ask, what was the assumed gain behind redesigning an interface with a familiarity base of 100 million? Hard to say, but usability in the wild is obviously as much about familiarity as optimality. Just because something has been and continues to be done in a particular way – just because coordination of goals and means has been achieved and internalized in some fashion – people become invested in that way of doing it regardless of whether or not it’s the best way. They have an unreasoned emotional response to change even though the change they’re reacting against is ultimately beneficial. From their perspective, any change represents a move from comfortably non-reflective “know how” to a comparatively frustrating re-investment in “learn how”. It’s bitterness over the perceived loss (insofar as they couldn’t vote on the switch) of time and attention spent internalizing the old system. And as research and common sense show, we hate losses.

Looking at it this way, we see why most of the detractors’ gripes are couched in terms of usability, without forcing us into a battle of intuitions about what constitutes “usable”. The usability issue is most likely just a salient rationalization. By most “objectively” recognized usability standards the new design is superior. People are really griping about the fact that they’ll have to reinvest or re-learn something and that the standards that determined the value of the change weren’t the standards by which they evaluate the site’s functionality. In other words, something not clearly broken was fixed without explicit agreement of the stakeholders, in effect forcing a new investment without due consideration of past investment.

The Miscellany, Nominalism and The Essence of the Web

[NOTE: This version originally appeared on Armchair Media's blog. The more verbose version that was here has been removed... it was a drag.]

Given the tagged, user generated nature of the web today, is Information Architecture a dead practice that just won’t lay down? Are rigid ontologies a crime against the intrinsic nature of content on the web? Do outdated, hierarchical ways of ordering content on the web commit us to stifling and misleadingly authoritative structures that curtail innovation and invention?

David Weinberger seems to think so. He doesn’t like hierarchies, taxonomies or narrow instrumental classifications. His latest book, Everything is Miscellaneous, is devoted to drawing out just how inadequate these structures are when applied to the distributed, messy heap of content on the web. This isn’t a particularly new observation. Clay Shirky has been speaking to this point for years now. The basic idea is that the classificatory structures that we use to deal with the very real constraints of the physical world lend an artificial “essentialist” necessity to our task-based classificatory practices online. Ordering and making sense of the welter of stuff that populates our closets and sciences is necessary given the limitations of the physical world. That doesn’t mean they’re necessarily “true” in some big, extra-systemic sense, however. The need for strict, pre-determined structure is mitigated online because computers are really good at creating ad hoc order out of an undifferentiated mess. Online, not everything needs one and only one place. Acting and designing as if it did at best limits the democratic/utopianist promise of the web and at worst imposes or perpetuates illegitimate power structures. My interest is in what this does to the practice of web design.

First of all, I agree with most of what Weinberger and Shirky say about the changed face of order on the web. The majority of content falls into a miscellany. It’s just a mess of digital artifacts awaiting order. Anyone can throw stuff out there and it doesn’t really matter where it resides. If it has metadata on it or we can read its content, we can find it again and probably even put it to some use for which it was never intended. It's the potential for order that’s important. Designers have been working out the implications of these ideas for years now. Apparently, Weinberger’s book is intended for the decision makers near the top that may not have worked out the implications of ideas like tagging.

A key target of Weinberger and Shirky’s ire are ontologies or data structures intended to explicate the concepts in some domain of knowledge. Ontologies, in the information sciences sense, are the cornerstone of the Semantic Web. Shirky uses the term more broadly, thinking of it in philosophical terms as the sorts of things countenanced by some system, the sorts of things that some domain recognizes as constituents. The claim is that ontologies lead to a species of essentialism, or the world view that everything has some set of properties that uniquely identifies it in all possible situations. Both Shirky and Weinberger spend a lot of their time pointing out the fact that most ontologies are arbitrary and conventional, but that they can give the impression of necessity or self-evidence. For them, different ontologies are just competing webs of more or less useful concepts. In holding this view we can refer to them as nominalists.

My problem with most influential nominalists on this and other points is that, if not read or followed carefully, they can give the impression that there’s no place for local, designed order online at all. In other words, sometimes their fervent tone gives the impression that there’s absolutely no room or reason for islands of structure built from the welter of free form content. In their desire to rout “ontologizing” hierarchies simply on principle, they obscure the fact that a task-based local structure doesn’t automatically engender a totalizing, absolutism. The local structures on most sites are about sense-making and communication, not world-making and dictation. For instance, Flickr, Weinberger’s favorite example isn’t really the free-wheeling heap of free-range content he casts it as. Content enters Flickr neatly, if minimally, categorized (“category” isn’t coextensive with “hierarchy”) by user, camera type, etc. One could almost say that the user provides a loose ontology for Flickr. The user, a sort of default set, is the center of gravity that holds Flickr together; it’s the initial structuring principle that gives the content its potential for further order and defines possible transformations. It’s a wonderfully loose binding, but without this initial act of minimal categorization, the potential for further arbitrary re-ordering is lost. Though Weinberger and Shirky clearly know this, they should probably state it more explicitly as their management level audience may not be in the trenches enough to cut through the polemic and hyperbole.

There’s no such thing as order ex nihilo. Computers are really good at aggregating if they’ve access to some sort of potential ordering mechanism (tags, official metadata, etc.). If it’s text, this could be pure content. If not, we need to attach something, thus Flickr’s initial, user determined set. If we’ve no handle by which to grab the content, it might as well not be there (particularly with images and video). This is not to say that there’s no emergent order, which is a separate issue entirely.

As we’ve been stating for years, Information Architecture in the old fashioned tree-structure sense does seem pretty irrelevant. Definitely so for IA in the grand, Peter Morville sense, in which all sites are trees, everything has one and only one spot and there’s very little cross-linking of paths. Still any structure that gives sense a la Flickr’s implicit categorization is a form of architecture. It’s an ordering principle that somehow sensibly aggregates content on a local scale. So we’ve no longer Information Architecture in which all structures are made from custom built materials. Rather it’s more of an information shanty town in which personal structures are built from the miscellaneous heap of common materials. But the important point is that it’s still architecture, and still reproduces common elements like doors, windows, etc (to stretch the metaphor a little thin). We need pools of order with understandable principles of manipulation that briefly and locally coordinate elements from the miscellany. This point is often obscured in Weinberger’s book and we, as designers, will have to pay the price once it starts to circulate within marketing departments.

And though Weinberger (at least) seems to frown upon instrumental concerns on the web, we tend to have to design sites that people use to do things. At an abstract level, the nature of the miscellaneous mess of content means that it becomes more useful the more content there is and the more it’s tagged. For example, the more people upload and interact with content on Flickr, the better your chance of finding photos tagged “fruitcake” that are actually pictures of fruitcakes. Tagging after all is often more about the tagger than the tagged. Once enough people are doing it, however, the tags start to become useful content locators; at high volumes, idiosyncrasies and bad tags tend to filter out. Unfortunately, this suggests that loose structures, based entirely on tags tend to have an instrumental usefulness that varies with content volume and interaction. The moral: if you’re working on a project with
a smallish expected volume that needs to be useful out of the gate, you may still need to impose a strict categorization scheme in order to meet you goal. This, to Weinberger, is a sin, or at least a backwards looking crime against the essence of the web. At least that’s the impression one gets reading the book. (Shirky takes a more nuanced, “domain of discourse” based approached)

On the ground, designers realize that these quicksilver, local categorizations are the fundamental means by which we define our sites and some of our only means of communication online. They’re a part of the language of the web through which we communicate our clients’ messages. Without these structures, there would be no sites per se, just a grey fog of rootless content awaiting individual requests for order. This can’t be what Weinberger has in mind. Yes, Google is hugely popular. But it’s also a high-level general ordering mechanism and a liminal space intended to get people through it, not to it. The success of Google at what it does does not mean that we should all follow that example. Unless all we want to build from here on out are search engines.

So, we all agree that, in general, top down hierarchies are less and less relevant to a distributed, open access dumping ground like the web. And, of course, nobody really thinks that there’s only one relevant ontology. Ontologies are domain specific and the attributes assumed within them are never intended to be intrinsic to the content. Still, the need for order comes from two directions: the client and the user. Sites are doors into the mass of content, zones of order that communicate in part through the principles they impose on the miscellany. Users don’t yet want to get down to the level of the miscellany. Google, is the closest we like to get: a presentation of an ostensibly ordered set bound together by some minimal user-defined “intension.” The lower level domain-specific sites we generally concern ourselves with usually require a tighter sense of order than something high-level like Google. In general, as domain specificity increases so too the appropriate ordering principles (conventional content breaks, vocabulary, etc.). Users want a certain decrease in uncertainty as they become more specific in their searching behavior, but I agree that they never want its elimination. This is what the new ordering principles are all about. Allowing local order without stripping away the global properties of the miscellany.