Max Hallinan

Silos of Subjectivity

Many peer-to-peer networks identify data by what it is and not where it is. This is called content addressing. Content-addressed data is not tied to a single host. Any node that downloads the data is capable of hosting it. Multiple nodes can host the same data simultaneously. If one host goes offline, the data is still accessible on other hosts. The data is available until all hosts go offline and the likelihood of this happening decreases as the number of hosts grow. Meanwhile, the address of the data remains the same. Links to the data do not break when the location of the data changes.

For these reasons among others, peer-to-peer networks have recently been celebrated as a remedy to the “information silos” and “walled gardens” of the contemporary world wide web. These terms broadly describe conditions on a client-server network where data is stored in a closed system and access to the data is controlled. Publicly-funded research published behind a paywall is a common example. Personal data that can’t be extracted from a social media platform is another example. In both cases, the data is not directly exposed to the public. The public must go through a controlling authority to access the data and access occurs under terms set by that authority. This control is possible because client-server networks use the location of a host to identify the data. Location addressing ties data to its host and whoever controls the host controls the data. This kind of control is harder to achieve on a peer-to-peer network because the data is not identified by or tied to a location. If a host tries to control access to the data, a client can route around this host and access the same data at another location. The intent is to embed the ideal of openness in the network architecture by making it much harder (perhaps impossible) to own publicly accessed data.

Better availability on peer-to-peer networks does not guarantee availability. Given a valid address, a client can wait eternally for a host to connect to the network. In my experience with Bittorrent, this is not uncommon. I rarely pirate movies because the movies available to download are often not the movies I want to watch. A torrent for a Canadian documentary is useless if I can only locate peers seeding the latest Hollywood blockbuster. What I actually watch is often a compromise between my tastes and what is available to download. That is because resources in greater demand are more likely to have multiple hosts. As the number of hosts grows in some rough proportion to the demand, the popular resources generally enjoy greater availability.

My primary concern is that this effect could distort the peer-to-peer web through a process that resembles natural selection: available resources will be those that have proven their fitness, either organically through propagation to a large number of peers or artificially, by location on a small number of resilient hosts, e.g. a VPS. Inevitably, the availability of data will be decided to some degree by patterns of selective downloading and hosting. Some data will be selected simply because it is the easiest to access, just as I have often chosen to pirate movies that are fast to download instead of movies that interest me. When I choose to seed these movies after downloading, as I routinely do, then that data has propagated to another host simply because it was the most available to begin with. As the pattern self-reinforces, the network can become weighted towards and amplify the bias of its users, especially the bias that informed the earliest selections. However this compares to the client-server web of today, it is inconsistent with the egalitarian ideal of free information that the peer-to-peer web is intended to manifest. Like the virulent echo chambers that have emerged on social networks, a web based on such a network could become more a tool for confirming bias than for deepening understanding. If the peer-to-peer web does not take steps to ensure general rather than selective availability, the silos of information will only be replaced with silos of subjectivity.