A reply to Evgeny Morozov.

There are two ways to be wrong about the internet. One is to argue it doesn’t live up to its hype. Speculative futurism and unabashed mysticism have become commonplace in discussions of technological change, and it isn’t hard to find people ready to claim that the internet is a panacea heralding everything up to and including immortality. In such an environment, one need only be moderately critical about the internet to position oneself as a pariah standing against a swarm of naive technoidealists. Democracy doesn’t even work on Wikipedia, the argument goes, and so it is foolish to think that “liquid democracy” will change the form of legitimate governance (read: the nation-state) in any substantive way, hype be damned. The problem with such criticisms is that they treat the possibility of internet-generated change as all-or-nothing: either the internet meets the expectations of its most wide-eyed advocates, or it is a waste of time with all the sociopolitical importance of a video game. There’s no room in this view for registering the subtle cultural shifts that can change the practice of legitimate governance over time, or for understanding how the ideals of extremists can change the discourse even when their ideals are not achieved.

The other, more insidious way of being wrong about the internet is to accept that the internet changes things subtly, and proceed to argue that the old ways were better. That’s what I take Evgeny Morozov to be doing in this article, and it’s important to see how regressive his arguments (and the institutions they support) are. Just to be sure I have the argument right, I’ll try to charitably reconstruct its key points before blowing them to pieces.

Morozov’s core argument against Johnson’s “internet-centrism” is that it is shallow:

It’s not hard to see why: his Internet-centric theory of politics is shallow. Wikipedia, remember, is a site that anyone can edit! As a result, Johnson cannot account for the background power conditions and inequalities that structure the environment into which his bright reform ideas are introduced. Once those background conditions are factored in, it becomes far less obvious that increasing decentralization and participation is always desirable. Even Wikipedia tells us a more complex story about empowerment: yes, anyone can edit it, but not anyone can see their edits preserved for posterity. The latter depends, to a large extent, on the politics and the power struggles inside Wikipedia.

Not only are the naive politics of internet centrism underequipped to handle the realities of politics, Morozov argues that those realities are in fact best met by centralized, hierarchical structures. Thus, even in groups that  advocate decentralized politics we see an inevitable need to centralized and consolidate power in the form of traditional hierarchies:

The Porto Alegre experiment succeeded because there was a centralized effort to make it work. Centralization was the means through which the end of decentralization was achieved. Without well-organized, centralized, and hierarchical structures to push back against entrenched interests, attempts to make politics more participatory might stall, and further disempower the weak, and coopt members of the opposition into weak and toothless political settings. This was the case before the Internet, and, most likely, it will be the case long after.

Morozov repeats this theme for the rest of the article: decentralization is a poor method of organization, centralization and hierarchy is a political reality we must take seriously, decentralization is unprepared to meet that challenge. “Internet centrism”, while superficially advocating decentralization, is itself a centralizing, institutionalizing force, putting “severe intellectual limits” on the narrative surrounding politics in the digital age. More than just hype, the naive politics of internet centrism is a discursive nuisance. It distracts us from serious reflection on the activities of legitimate institutions of power.

At least, I take this to be Morozov’s position. The upshot is that we’d do better, from the perspective of the political discourse, to ignore “the internet” altogether, because treating it as a centralizing and totalizing political agent isn’t getting us anywhere other than up our own ass. Real political change comes from working with the realities of institutional politics today, and internet centrists “lack curiosity about how the world works”, worrying more about their gadgets and sci-fi fantasies than any serious reform.

There’s places to quibble about the details, but I hope it’s a generous enough reading of the article that the following comments have some impact. I haven’t read Johnson’s book and I don’t mean to defend it; my goal is to straightforwardly attack Morozov’s critique of it. Or rather, I want to give a defense of networked politics against the hierarchical institutions that Morozov treats as inevitable. Morozov’s critique ultimately misunderstands the role of networked analysis in politics, and profoundly misrepresents the advantages it has over other forms of political analysis. Perhaps this misrepresentation is Johnson’s fault, but Morozov should know better quite independently of Johnson’s popularizing text.

The error Morozov makes throughout the essay is the use of “centralization” as a primary axis of political analysis, ranging from the quite centralized operation of (apparently) successful hierarchical institutions (like political parties), to the decentralized, horizontal structures of “internet-centrist” institutions like Occupy. Morozov’s view is quite simply as follows: centralization is good for organizational consistency and sustainability, and decentralization is bad. He writes:

If one assumes that political reform is long, slow, and painful, hierarchies and centralizing strategies can be productive. After all, they can keep the movement on target and give it some coherent shape. Ideas on their own do not change the world; ideas that are coupled with smart institutions might… But Johnson is completely blind to the virtues of centralization. 

Morozov doesn’t elaborate on how centralized institutions might better challenge the existing power structures without recapitulating their crimes, but reiterates (with no argument) that they “might”:

 Challenging power requires a strategy that in many circumstances might favor centralization. 

Morozov’s position is purely observational: look at how decentralized movements like Occupy routinely (!) fail, and look in contrast at how many and powerful their centralized competitors are. If we are going for  lasting and effective institutions, it would seem clear that a hierarchy has the best chance. Right? If we were to take Morozov seriously, we might be persuaded to think that centralized hierarchy was reliably correlated with organizational power, such that the more tightly structured the organization, the more successful, controlled, and organized it becomes. Presumably the military is an institutional ideal on this view, one that political dissidents should emulate? If this seems absurd, we get no better suggestions from Morozov, who fails to acknowledge any of the vices of centralized control or to pin the failures of our institutions on their structure or theoretical grounding.

On his view it’s not so much that hierarchies are good, but that the evidence shows them outperforming decentralized alternatives. Given the existing order of things, it is a curious kind of pragmatism that defends the status quo. One wonders how Morozov explains the dreadful state of our existing institutions, if our equipment was already well-poised to challenge the powers that be? Instead of seeing the adoption of alternative institutional structures as a move born of desperation by those for whom the existing systems have failed, Morozov sees decentralization as a leisurely, unnecessary, and incompetent social experiment by people unwilling or unable to engage in the serious work of reform. I suspect such attitudes say more about Morozov’s social circle than the subjects he claims to study. Moreover, he’s convinced that the trial runs we’ve seen  by the groups identified as “internet centrist”, like Occupy and the Pirate Party, is already sufficient evidence for condemning the whole approach. He’s quick to link the innovative strategies of these radical upstarts with collectivist thought and media analysis from decades before the Internet, to further suggest that the strategy has had plenty of time to mature and succeed, and has utterly failed to do so. On his one-dimensional analysis, the jury is basically already in: decentralization is functionally incompatible with organizational success. Morozov is convinced that the barbarians can’t win; the only way to make Rome fall is to build a better Rome.

The observations and anecdotes he has collected here might be a convincing argument for perpetuating the existing order, if centralization were the only dimension available for analyzing the structure of political institutions. Thankfully, it isn’t. Organizational structure and success is not a product of hierarchy and centralization any more than it is a product of decentralization and horizontalism. Organization and centrality are entirely distinct measures of a network. Successful organizations find ways to effectively execute sustaining functions by any means, and this requires both concentrating and distributing power in a variety of ways. Dogmatically clinging to any degree of centrality for an institution is meaningless without some explanation of the institution’s functional role and methods for executing that role.

Network and systems analysis allow the study of these power relations– and their organizational capacities– beyond the arcane limits static, top-down hierarchies of nominal control. Organizations can flourish within a wide range of institutional control, but can just as easily fail. What matters are the organizational dynamics within those structures; these dynamics emerge from the actions of agents at many scales of analysis, not from the artificial bureaucracies that have been tasked to manage them. Hierarchical control imposes constrains on the system’s ability to organize. Some constraints may be necessary or even desirable, but it is a grave misunderstanding to mistake the constraints for the organization itself; it is something like thinking the pot causes the plant to grow. Morozov confusion over constraints is nowhere more apparent when he accuses Johnson’s internet evangelism with banal libertarianism. Network theory, which worships the same gods as Morozov’s disdained “internet-centrists”, does not consist in the dogmatic advocacy of decentralization or the arbitrary smashing of pots.  It is the study of organizational dynamics across all forms of institutional and environmental control. If that analysis recommends the smashing of pots, it’s because some pots really do hinder the development of the network. Digital politics cares about discovering when and how development occurs, not about arguing over whether or not we need more pots.

Consider, for instance, the dynamics of the following business:

This is a networked visualization of the organization. There is still quite obviously a hierarchy here (managers are depicted differently from their subordinates), and centralization (the CEO is depicted at the middle). Still,  to focus on these features of the network alone is to profoundly miss its organizational dynamics of the system over time. The “official” hierarchy of the business at any one time is preserved in this representation, but it isn’t necessarily the most interesting or important way of looking at the network, and there’s no reason to think that the success of the organization requires us to restrict our view to only those aspects of this network. In fact, we’d expect any description of the power relations within this institution that focused purely on hierarchical channels of power would miss all of the most exciting dynamics over the course of this time frame.

I want to give an example of how the dynamics captured by networked analysis can be missed by a traditional hierarchical analysis, but it’s important to first understand just a bit more about how organizations develop: they modularize. A module is a differentiated subunit that can execute a function for the sake of the overall organization. We see many such modules in the above business model; they are the little trees that seem to be floating somewhat independently from the bulk of the network. Occasionally, these modules break off from the main network (perhaps a group project within the company that took off on its own), eventually exploding out of the business altogether. Other times, the modules were acquired from outside the business and slowly become integrated into the rest of the network. These changes seem to culminate in a few large restructuring efforts across the business. It’s a very dynamic process!

It’s easy to mistake these modules for yet another instance of traditional hierarchies. After all, modules are usually “headed” by a manager (a node with a slightly larger radius at the root of a tree), and the managers sometimes connect back to the main network hubs, so if someone is looking for confirmation of the importance of hierarchies they might take this to be a sympathetic example. But modular organizations aren’t structured as hierarchies, any more than your heart is hierarchically superior to your stomach. Hierarchical models are simply incapable of capturing the dynamic interdependence between the organs of an organism: each are vitally important, and each offers something unique to the overall well-being of the organism. This is the kind of situation that network analysis was designed to handle; it’s why we talk today about ecosystems and food webs instead of the “great chain of being” with humanity (or a CEO or anything else) at it’s apex. Hierarchy is as obsolete as Ptolemy.

This same shift in perspective that has long overcome the sciences is also represented in the political sphere by peer progressivism and other forms of “internet centrism”. What these new political models hope to achieve is not just decentralization, but self-organization, and this is something categorically distinct from the libertarian individualist ideal. Self-organization is not an internet-only phenomenon, but the internet has given not only the tools but also the language and culture required for so many people to participate in self-organized systems at global scales. Memes are an expression of culture, and while social movements can’t be won by culture alone it’s absurd to suggest that we don’t need a strong and vibrant culture to sustain our ranks. The internet also provides the most compelling and inspiring examples of self-organized success stories available in the common discourse. Facebook, for instance, is largely a self-organized system, and despite what the advertisers and new media journalists tell you its success is largely a product of what we’ve made happen as users, quite independent of whatever inane constraints its overpaid staff of executives and designers care to impose on the network. The fact that they are getting rich off the process isn’t a sign that they are doing something right, so much as a sign that we’ll find ways to organize in even the most incompetent of social structures. It is no wonder, then, that the Internet has inspired such lofty rhetoric among its many vocal advocates: it has also inspired real hope in the possibility of an alternative. It provides empirical grounding for the belief that if we had better tools for visualizing our networks, we really would be able to organize better.

So here’s the example: we know that human organizations tend to be right-skewed; this is true of social and professional networks, and of the structure of the internet as a whole, and of many other human organizations besides. Scale-free networks are pretty common throughout the rest of the natural world, but what matters for this example is that human activity likewise scales with size to fit these power law distributions. Right-skewed networks have a very important property: most nodes have low degree (or low “centrality” as a network theorist would use the term), but some nodes have very high degree. If we think about a network of friends, most people have a small group of friends, but a few very popular people have a whole lot of friends that span many smaller groups. What counts as a “small” or “large” group of friends depends on the population size, but this popularity distribution is consistent across all populations. Not everyone can be popular, or else everyone would have a lot more friends. But most people know someone that is popular, so nodes with high degree tend to be important for connecting all the smaller clusters and organizing the network as a whole. If we were playing the “Six degrees of separation” game, the popular nodes would be the ones that people tend to use more often for making connections between clusters. The same is true of the networked business model shown above: most nodes have low degree, but a few important nodes have very high degree, and are associated with many different modules.

That’s the example, and here’s the point: there’s no reason whatsoever to think that the traditional hierarchical methods for identifying network relations is best suited to identify and support these high-degree nodes in their organizational capacity. High-degree nodes represent one measure of centrality on the network, but hierarchical models may treat different nodes as “central” and authoritative (hence Morozov’s confused terminology), thereby misrepresenting the organizing relations in the network. In other words, the people we’ve labeled “boss” and “manager” and “worker” through traditional top-down methods might not correlate with the actual dynamics of power relations within the system, and might not be helpful in explaining how the system organizes– or how it might be improved. Instead, it merely introduces a new constraint on the system, which might have any number of consequences as to how effectively the resulting system will organize. A hierarchy introduces a new lever of control that can be wielded by the designated authorities, but not one that bears any necessary relation to the organizational goals of the system. Morozov’s apologetic for “centrality” doesn’t correct this problem; if anything, it exacerbates the situation simply by not understanding what’s going on.

My example, of high-degree nodes not correlating with hierarchical models, is meant to be a clear case where the difference between a networked analysis and the more traditional analysis that Morozov advocates actually makes a difference for the success of the organization, yet doesn’t just reduce to an ideological debate between greater or lesser degrees of centrality in a network. But perhaps that’s still too abstract to make the example live.

So here’s a more accessible example of the same phenomenon. It is well known in urban planning that the flow of traffic along a given street depends almost entirely on the conditions on the road (including the behavior of other drivers), and is almost entirely independent of posted traffic signs. In fact, we know that you can raise or lower the posted speed limit by up to 20 mph and have a statistically insignificant effect on the average speed of traffic.  To put the point simply, speed limits have no effect on traffic. They do have an effect on how many people violate traffic laws, but that still doesn’t change the patterns of traffic flow.

The speed limits do impose a constraint on the system, and motorists who violate those constraints to risk being penalized by the system. But these constraints are completely ineffective at their apparent goal, which is to keep the traffic safe and organized. Politicians and misinformed advocacy groups might think that lowering speed limits would make the roads safer, and they’d be wrong. Of course, the traffic can be managed: through improving the roads, traffic signs and signals, and improving driver training and vehicle requirements, increased automation, and so on. These are the “better tools” that will legitimately help traffic flow improve, and they are all predicated on the assumption that the system will effectively self-organize when provided these tools. These all require planning in some fashion, so saying “speed limits don’t help!” isn’t an appeal to libertarian dogma. Speed limits, as constraints, aren’t necessary for the well-functioning of the system, not because constraints are bad but because these constraints put the levers of control in entirely the wrong place.

The traditional view is that the organization of the system is the product of well-planned and enforced laws through well-established, hierarchical channels of legislation and enforcement. On the traditional view, the only machinery the public has for enacting justice is to appeal to this static hierarchical institutions, for through them all justice in the system supposedly runs. This is a kind of imposed “centrality” that may not reflect the dynamics of the system, but rather merely the constraints our designated leaders have seen fit to impose. Many of our social institutions function like traffic laws and incompetent bosses: where the system continues to work, it works despite these constraints, not because of them. But in many increasingly tragic places, the system has failed to function at even minimal levels, and the institutional constraints prevent any users any means for providing feedback to better calibrate and improve the system. Instead, they are forced to wait until their designated representatives find the political motivation to heed their call. In such circumstances, the stalwart defense of the existing order and the cheap mockery of attempted alternatives is appalling.

Morozov visibly struggles with the emergence of organization in complex systems when he presents a dichotomy “between two mutually exclusive methods of inquiry” towards the end of the essay:

One—an outgrowth of Internet-centrism—is driven by the impulse to totalize and generalize; the other by the impulse to disaggregate and particularize. One has space for the Internet and little else; the other eschews any talk of “the Internet”—it deliberately puts it in scare quotes throughout—and engages with platforms and technologies on their own terms, as if they share no common logic. Instead of assuming that these technologies emerge from “the Internet,” this second approach assumes that it is “the Internet”—as an idea, if not as a technical network—that emerges out of those technologies.

In fact, the move towards holism and particularism are not incompatible methods of inquiry, but must be taken on together. Pluralistic approaches are fundamental to studying any complex system, and Morozov’s suggestion that this dichotomy is exclusive reveals how little he appreciates the developments in network science over the last few decades.

But his characterization of the latter horn of the dilemma– treating “the internet” as an idea that emerges out of our technologies– isn’t far from the truth. The internet is partly the protocols that allow our technologies to interact, but it largely the employment of those technologies by users (most of whom are themselves other machines). The pages we browse through are the traces of that use over time: who produced what content, how it was distributed and reproduced across the network, and how it all links to each other, as an active and continuously unfolding process. Digital culture really is defined in terms of the unfolding of memes in an economy of attention, where users deploy whatever means possible to ensure fertile ground for their development and reproduction. By using these tools, we have collectively produced the the most lasting monument to human knowledge ever constructed, one that can’t be burned down or raided by vandals. We have produced a rich ecosystem of social networks that self-organize to sort through the massive streams of information we produce and consume. We are capable of doing this because of our tools, because we want to, because we are really good at it. As a species, we’ve been doing this for a long, long time.

All that’s changed is the tools we have for doing this, but given the right tools this can be a qualitative change with enormous impact, and to discount the potential impact of the internet at this early stage is hasty to the point of regressive. We’re just starting to mature as a digital culture, and to see our labors start to take shape and blossom. It’s like learning how to walk, and we are toddlers in newly acquired digital legs. They are uncomfortable and we don’t yet know quite how to move them right; we haven’t properly trained the network. We still routinely stumble over ourselves as we try to coordinate abilities we haven’t yet developed skills for; Kony was perhaps a particularly notable stumble. But to suggest from these early cases that we’ll never learn to walk– or dance– is absolutely cruel. Our digital selves, our emergent digital culture, is now part of who we are as agents. That doesn’t guarantee that we will dance gracefully with these digital legs, but given that our inherited legs are corrupted and nonfunctional it’s completely reasonable to keep trying, to keep practicing.

Because whether or not it will work, this is the only way it can.