"Can We Reinvent the Internet?" by Viktor Mayer-Schönberger
Lee Kuan Yew School of Public Policy, National University of Singapore 259772, Singapore. E-mail: Viktor_MS@nus.edu.sg
Recently, researchers who support "network neutrality" have become worried that the Internet may lose its innovative edge. They are concerned that control could be shifting from the edges of the Internet toward the service providers at the center, which would allow the providers to have "gatekeeper" capacity and would contradict the Internet's "end-to-end" principle (1–3). This core tenet states that control over information flows should take place, to the extent possible, at the end points of the network (4). President Obama supported net neutrality during his campaign (4) and in recent statements (5), and the European Parliament has added net neutrality to its recent telecom bill (6). Taking the end-to-end principle from protocols to users, Jonathan Zittrain has called for maintaining the Internet's "generativity," the ability of users at the network's end points to create, distribute, and run whatever software code they choose (7). There are good reasons to preserve network neutrality and generativity, but it is unclear whether these are sufficient to ensure continued innovation. The larger issue is what policies are optimal to foster innovation on the Internet?
Innovation and the Internet
Given that the Internet connects general-purpose processors, most of the innovation happens in software code, which, once released, is rapidly distributed through the network. A new version of Apple's Safari achieved 11 million downloads in 3 days (8); many open-source software packages are compiled nightly to be ready for (automated) download the next morning. A combination of global Internet connectivity, automatic software updates, and centralized code management makes this rapid distribution possible. But these updates are mostly minor changes to software already installed on user machines. Although the ability to rapidly react to hacks and bugs by pushing down a software update is no doubt beneficial, it tells us little about the Internet as a system that enables code innovation.
In contrast, new software products or radically new versions of existing software often face a difficult time getting accepted in the marketplace. Perceived costs incurred when switching from one software product to another, e.g., the cost of data conversion and training, make users keep the software they already have installed (9). Costs are higher still if a code switch necessitates hardware changes, as well (especially when, owing to its complexity and specialization, the required hardware is expensive). Moreover, positive network effects ensure a pull toward the popular and established, but not necessarily the most innovative, software products. A superior new software product may attract relatively few users for a prolonged period of time, before finally, rapidly taking off (10). Conservative user choice, in turn, influences product development, pushing most vendors to alter their software titles only incrementally to avoid the risk of losing their user base. Microsoft's woes in converting users from Windows XP to Windows Vista offers a powerful case in point, so does the painfully slow switch from the network protocol IPv4 to IPv6 (11). This tilts development of new code toward incremental innovation.
The Social Network of Users and Coders
Peer production of software (facilitated by the open-source culture, new tools for managing distributed software development on the net, and global connectivity) has emerged as a powerful new mechanism of software creation (12). It has even extended beyond code creation (e.g., Wikipedia). In contrast to commercially developed, proprietary software, open-source development regularly involves users—not necessarily in writing code, but for reporting bugs, flagging areas of improvement, or partaking in online design discussions (13). Decisions are often driven by the need to establish a rough consensus of those participating (14, 15). Because peer production of software is organized through social networks, the Internet functions as both the technical infrastructure to diffuse code and the tool to maintain a social network, enabling communication and information-sharing among code writers and users. However, reliance on social networks may make open-source development just as conservative as commercial software production.
Figure 1: Structural holes Diversity through limited connectedness
Many open-source projects consist of a small number of active coders and a larger number of bug reporters and commentators, all connected with each other through the Net (13). This thicket of connections within the community of a particular project facilitates coordination and information exchange among its members, but, as a social network topology, it is not conducive to radical innovation. Burt has suggested that an individual node's value is not proportional to the number of its links, but whether these links are within one's existing community or connect two disparate groups (16). In the latter case, the two connecting nodes are valuable information brokers, spanning what Burt terms a "structural hole" see figure, left) (17). Brokerage opportunities between different communities, and the value gained from these, require the existence of such structural holes. Because of the way in which we have organized the development of software code through the Internet—in relatively well-connected online groups—open-source communities may suffer from a lack of such structural holes. An overabundance of connections over which information can travel too cheaply can reduce diversity, foster groupthink, and keep radical ideas from taking hold. This may explain, at least in part, why much peer-produced code rarely is altered in more than an incremental manner (18).
Figure 2: Groupthink Many or all nodes linked to each other
This is less problematic in a competitive, profit-driven setup. If a software vendor falls too far behind the innovation curve because of too-conservative incremental changes, a competitor can mount a successful challenge. Such a competitive setup, in essence, creates two disconnected communities of development. It may require significant initial capital expenditure and time to build momentum. But the prospect of future (if temporary) market dominance fuels risk-takers, as well as a sufficient inflow of venture capital.
The situation becomes problematic, however, when there is insufficient economic incentive for multiple products. The most prominent example is not commercial software code, but the Internet, or more precisely the protocols underlying this dominant network infrastructure. It is too costly and risky for a commercial competitor to create and market a set of radically improved, but incompatible protocols. This is true for the peer-producing, open-source community, as well.
The National Science Foundation supports the Network Science and Engineering (NetSE) program and the Global Environment for Network Innovations (GENI), two mechanisms aimed at enabling more radical network innovations (19, 20). Their work could be improved by incorporating the social network perspective on code creation I have outlined above. Given that they already have a social science component, such integration may not be too difficult. We need to create incentives to form network structures that stimulate and enable innovation.
How should policy-makers respond, especially when wanting to facilitate the creation and growth of the next-generation Internet?
If densely connected social networks, although they permit incremental innovation, are hampering radical innovation, two potential responses come to mind. The first focuses on the physical network, implying that it may be necessary to build a next-generation Internet outside of the existing infrastructure. This could free development of the legacy constraints that may hamper radical innovative changes. Any government money to build out the backbone of this future network could mandate that more radical innovations be implemented, with interoperability achieved through defined gateways rather than built right into the system.
This option, however, may be of limited value, if development of the next iteration of Internet breakthroughs, or even the next-generation network itself, is organized through the Internet. Unlike their brethren during the Manhattan Project, today's coders stay connected through their social networks and remain tethered to the context of the existing structures. Project requirements may permit them to consider radical solutions, but their connectedness to the thinking of the status quo through their social networks holds them back.
The second policy option, therefore, focuses less on the physical than the social networks among the research and coding community. When offering public funding for development, government could mandate that the mechanism for allocating funding breaks loose of existing social networks. For example, funding proposals could be encouraged to involve smaller, but competing, groups and for longer periods of time, reintroducing limited separation, somewhat akin to what in the business literature is referred to as modularity (21, 22). Proposals could be reviewed not by peers, as is traditional (who are likely from the same social network as the grant writer, which impedes heterogeneity of ideas), but through a panel of experts one step removed from the research community. Adding reviewers from more risky commercial startups could induce review panels to fund more radical and diverse proposals. The European Institute of Innovation and Technology, a funding initiative of the European Commission, has adopted a variant of this suggestion. Eight of its 18-member grant-awarding board come from the private sector (23). Announced well in advance, such measures signal to the research community that risky and radical proposals are particularly welcome.
Much of commercial development, owing to trade-secrecy concerns, is already less densely linked into a social network, with the result that less policy intervention may be required. Among the available policy tools for government are direct and indirect financial incentives for smaller startups. Government may also want to offer incentives for "skunkwork" projects—deliberately removed from the corporate structure and the connectedness this entails—within larger companies. The creation of the original IBM personal computer (PC), as well as the Apple Macintosh, are good examples of the kind of commercial behavior to incentivize.
Most of the deployment of the next-generation Internet—from the infrastructure all the way up to the application running on the network—will likely be privately funded. However, in areas where government steps in to fund network deployment—to overcome the digital divide, and/or to stimulate the economy—such funding could similarly shift from using efficient, but less innovative, mid-life-cycle mass-produced infrastructure equipment to more cutting edge technologies.
Connectedness of the current Internet constrains the ability of software coders to innovate and to build the code we need: messaging that is more resistant to spam, code to enable innovative controls of the personal information we share through social networking sites like Facebook, and measures to take open online collaboration to the next level, or to help us focus and tackle our society's most pressing problems. To enable innovations, especially nonincremental, discontinuous, and radical ones—which are needed, among other things, to launch successfully the next-generation Internet—may require unique policy intervention: reducing the social ties that link its coders.
References and Notes
* 1. W. Baumol et al., "Economists' statement on network neutrality policy" (AEI-Brookings Joint Center Working Paper No. RP07-08, American Enterprise Institute for Public Policy Research–Brookings Institution, Washington, DC, 2007).
* 2. B. Van Schewick, J. Telecommun. High Technol. Law 5, 329 (2007).
* 3. J. Saltzer, D. Reed, D. D. Clark, ACM Trans. Comput. Syst. 2, 277 (1984). [CrossRef]
* 4. MoveOn, Huffington Post, 29 October 2007; www.huffingtonpost.com/2007/10/29/obama-promises-toreinsta_n_70317.html.
* 5. Editor, Seattle Times, 8 June 2009; http://seattletimes.nwsource.com/html/editorials/2009315207_....
* 6. D. Meyer, ZDNet, 6 May 2009; http://news.zdnet.com/2100-9595_22-299414.html.
* 7. J. L. Zittrain, Harv. Law Rev. 119, 1974 (2006). [Web of Science]
* 8. J. Dalrymple, CNet, 12 June 2009; http://news.cnet.com/8301-13579_3-10263494-37.html.
* 9. H. Varian, C. Shapiro, Information Rules: A Strategic Guide to the Network Economy (Harvard Business School Press, Boston, 1998).
* 10. C. H. Loch, B. A. Huberman, Manage. Sci. 45, 160 (1999). [CrossRef]
* 11. A. Ely, Information Week, 20 December 2008; www.informationweek.com/news/infrastructure/ipv6/showArticle.jhtml?articleID=212501014.
* 12. Y. Benkler, The Wealth of Networks: How Social Production Transforms Markets and Freedom (Yale Univ. Press, New Haven, CT, 2006).
* 13. J. Lerner, J. Tirole, J. Ind. Econ. 50, 197 (2002). [Web of Science]
* 14. K. R. Lakhani, E. von Hippel, Res. Policy 32, 923 (2003). [CrossRef]
* 15. Steven Weber, The Success of Open Source (Harvard Univ. Press, Cambridge, MA, 2004).
* 16. R. Burt, Brokerage and Closure (Oxford Univ. Press, New York, 2005).
* 17. R. Burt, Structural Holes: The Social Structure of Competition (Harvard Univ. Press, Cambridge, MA, 1992).
* 18. S. E. Page, The Difference: How the Power of Diversity Creates Better Groups, Firms, Schools, and Societies (Princeton Univ. Press, Princeton, NJ, 2007).
* 19. NetSE, www.nsf.gov/funding/pgm_summ.jsp?pims_id=503325.
* 20. GENI, www.geni.net.
* 21. C. Baldwin, K. Clark, Design Rules: The Power of Modularity (MIT Press, Cambridge, 2000).
* 22. P. Evans, B. Wolf, Harv. Bus. Rev. 83 7, 96 (2005). [Web of Science] [Medline]
* 23. European Institute of Innovation and Technology, http://eit.europa.eu/about-eit/governing-board.html.