Exploring the uses, forms, and speculative promises of self-discovering networks

The authors wish to thank Arjit Kapoor for extensive research support and Laura Lotti, Erik Bordeleau, Patrick Rawson, Nick Houde, Oliver Klingefjord, Anna Fasolato, and Marco Mattei for their critical insights and precious perspectives. Credits for the cover image go to Jean ‘Moebius’ Giraud.

Networks and Trust

“Knowledge is a transductive process: it is not a package passed from person to person, but a system which produces the nodes in the network – you and I – through the very process of circulation.”
Dominic Pettman, The Species Without Qualities: Critical Media Theory and the Posthumanities (1)

“Depersonalize trust. Make it impersonal.”
Brian Massumi, The Power at the End of the Economy (2)

Networks exist as a structural model to understand and represent the world. However, they do not exist purely as epistemological frames: they demonstrate an ontological consistency. In biological life, networks function as an intracellular interaction system (cellular network) or a human’s brain activity (neural network). Networks are present in social life too: the economy (trade/credit network), or the social formations on the web (social networks).

A qualitative network transformation happened with digital society: the availability of data fueled an epistemological shift. We now perceive networks everywhere because we have the tools to grasp them. The description and the understanding of network behavior is the object of the study of network science: a common methodology and measurements for a generalized study of networks (4). Networks are described as mathematical graphs in which nodes and/or edges have attributes (e.g. names), usable These attributes can represent “the global structures of the interactions within a system” (5). A network is a high-level abstraction that allows the description of different scenarios through a set of common, universal metrics.

An integral property in a network is the nodes’ capacity to communicate with one another. To do this, it is necessary to develop trust among the nodes. Trust is both an intuitive concept used in daily life, and a far more complex notion.

First of all, trust is one of the foundational facts of every human organization and social structure, from the State to the markets, from security issues to value systems: without it, human bonds progressively unlace and disperse or, in other cases, they are rendered redundant. According to the German sociologist Niklas Luhmann: “Trust occurs within a framework of interaction which is influenced by both psychic and social systems, and cannot be exclusively associated with either” (6). To trust is to expect one set of behaviors instead of another and to choose to be vulnerable to the opposite set of behaviors. Trust is different from rational economic expectations and it may even be the underlying force presupposing an expectation.

Trust is not only indivisible from such systems: it is indeed a sense-making tool to navigate them and reduce their always-increasing complexity. Trust acts on “the future horizon of the actual present”, increasing the “tolerance of ambiguity”. To trust someone is to expect one set of behaviors instead of another and to be vulnerable to the opposite set of behaviors. Trust is different from rational macroeconomic expectation and it may even be defined as the underlying force behind an expectation: trust comes first and whether trust is misguided, the set of expectations shifts in another direction.

For instance, If one trusts the European Central Bank’s power to ensure the Euro’s value, one will expect a consistent continuation of such action and not an abrupt change of policy overnight. Trust, therefore, acts as a filter towards the future as it constrains the infinite possibility space towards likely outcomes, thus relieving the agent of alertness wherever possible.

“Co-ordinated individual action open up trust, by reducing complexity, so revealing possibilities for action which would have remained improbable and unattractive without trust – which would not, in other words, have been pursued.” (7)

In the complex social scenario we inhabit, individual trusted relationships are not sufficient to navigate and act on the world: it is necessary to establish a scale-free model of trust which can encompass collective belief systems of various kinds. Communal trust is not an external quality of the network but an emergent property that grows organically from it over time as a result of (inter)individual relationships (8). Therefore, it is not separable from the context in which it emerges. The philosopher Mark Alfano defines trust as “the capacity of cultivating reliable dispositions according to the socio-technical context (9)”. In a networked society, this means that trust is shaped by the digital technologies that surround us. Consequently, it is crucial for these techno-systems to be reliable or even trustworthy by design (10): trust needs to be depersonalized and scaffolded onto a system.

The necessity of embodying trust in a network is even more urgent in a specific type of network, the open network. The open network works through a decentralized authority and aims to increment the individual user’s autonomy and independence, such as WWW, e-mail, or the BitTorrent protocols. Open Networks align with ideals of decentralization and shared democracy. Nonetheless, the lack of vertical control brings along new design challenges: open networks have an affinity towards attack vectors such as cartel formation and Sybil attacks.

How can we build a network that is resistant to these weaknesses? How can nodes trust other nodes when there is no hierarchical authority? It becomes necessary to grasp the trust circuits that are constantly generated through the interactions between nodes in a network. Yet, this is not easy: trust is something of an invisible force that often eludes quantitative measurements, prone to fluctuations across time. This result is achievable when an operable definition of trust is defined; we turn to Oram & Viega’s simple definition that “Trust is the mechanism we use to decide that a key is valid.” (11).

In a simple interaction space where nodes sign each other's addresses as attestations of trust, the aggregation of these signatures yields a binary trust graph that reveals the distribution of trust throughout the network. Binary trust is either present or absent and it has directionality, which means that trust can be mutual or unidirectional.

Other use-cases require more granularity while registering node signals. A cardinal example is the star rating system which is widely used on various platforms for aggregating user preferences for various services. In these digital spaces, the source of trust is the user’s judgment about the service’s characteristics. Such aggregations compile all inputs in order to generate a universal topology i.e graph state. The utility of a net score allows for applications to derive objective attributes like sovereign identity, spam, quality, etc. While useful, these applications do not accommodate any subjectivity beyond the ontology of the trust algorithm.

Moreover, some applications are attempting to introduce subjectivity mechanisms such as Personalization Vectors of PageRank or Strategies in TrustNet. These weighting algorithms convert a topology with an objective distribution into infinite possible topologies by means of user subjectivation. For instance, the weights on the graph can be adjusted based on a classification input from the user. This is where personalized search results, curated feeds, recommendation engines, and the like originate. In cases when these preference orderings are organized en masse, algorithms tend to reproduce echo chambers where only the content that has already been discovered and interacted with is presented. This tendency can have serious consequences:

“Excessive homophily, however, helps to spread misinformation, frequently resulting in homogeneous, polarized clusters reiterating emotionally charged and externally divisive content“ (12)

When the mega-recommendation systems are sovereign, they decide who you should trust according to their own (commercially-minded) metrics, leading to an inhibition of self-discovery practices. It is important to have platforms that also allow for the emergence of novel subjectivities. This idea is reflected in the ethos of digital gardens, where values of encounter and curiosity — even at the price of chaos — are privileged over optimization and order.

How is it possible to bond together self-discovery and subjective trust? How can a secure, open network expand without a central authority?

The Web of Trust

The Web of Trust, built on top of the platform of PGP (Pretty Good Privacy), is a cryptographic protocol designed by Phil Zimmerman in 1991 in the context of the Cypherpunk movement. The Web of Trust offers a way to bind identities with the corresponding public keys in the form of certificates without relying on a central authority. It is a discovery mechanism; in it, users sign each other’s certificates and this mechanism originates a directed trust graph in which edges represent signatures.

When a user needs to obtain information about a certificate issued by an unknown user, he/she has to check for the presence of one or more trusted parties in the list of signatures associated with that certificate. The result is a system where users sign certificates of other users and retrieve information about the trust level associated with a certificate. Through these dynamics, the network propagates subjective trust, as every user is verified by a unique set of introducers (trusted intermediaries).

Web of Trust can also be depicted as a free association of trusted parts, mimicking the idea of the free association of producers from the anarcho-communist tradition. Free association is a utopian envisioning of a liberated world where workers associate themselves freely and according to their own capacities on an egalitarian decision-making basis. From this, a democratic production network emerges that functions thanks to member cooperation. The Web of Trust is similar, not issuing trust to the network’s participants according to some centralized, pre-established criteria, but instead through free and spontaneous relationships (in the context of a specific architecture).

In the Cypherpunk tradition, the exchange of public keys usually takes place during special meetings, named cryptoparties. In the early ‘10s, these convivial moments evolved into a nomadic institution, a global grassroots movement promoting P2P networks, end-to-end cryptography, and other privacy-enabling tools. In this context, trust was conceived according to an algorithmic, functional definition as well as a human component of personal interaction and shared conviviality. Through cryptoparties, the network grew all around the world without any centralized coordinator. A small operation — the key signing operation between people — can lead to a wider sociotechnical architecture and movement. As the physicist Guido Caldarelli argued:

“There must be some small-scale mechanism that, iterated through a great number of interactions, ends up generating a structure that is organized at the large-scale level.” (13)

The emergence of an order renders the network as something more than just the sum of its individual nodes. The network is an informational environment, an Umwelt [1], retaining trust indicators and coordinating users. This falls in line with the idea of stigmergy, defined by the P2P Wiki as a series of “environmental mechanisms to coordinate the work of independent actors”. It is also defined as local patterns that result from past constructions that provide cues for future architecture. Stigmergy is a method to inscribe the signs of collective intelligence into the network so that other agents can determine and incite subsequent actions according to those traces. For example, in OpenPGP WoT (a specific iteration of the protocol), trust is signaled in two ways:

‘Introducer trustworthiness’ refers to how much another user is trusted to apply care when verifying an identity. This value is determined and stored locally for every locally known user ID. ‘Public-key trustworthiness’. is the degree to which a user claims to be sure of a key-entity binding. Before using someone else’s public key, users must determine the key-entity binding and assess whether it’s likely to be correct. (14)

As the latter sentence reveals, the network environment, enriched with informational traces (who is the most trustable part, etc.), allows nodes to grow orientation and understanding.

From this perspective, Web of Trust constitutes a form of swarm intelligence, where individuals are collectively driven to a specific behavior, without relying on a leader. Group coordination is instead embodied in the network’s architecture, a socio-technical structure that facilitates the growth of the network through the repetition of an endogenous and simple dynamic (the key signing activity) based on a specific trust ontology. So, the cryptographic protocol combined allows for indirect, traceable communication of trust attribution (stigmergy) that fuels people into building this immanent structure that is the network. At this point, it comes naturally to recall Luhmann: an increase in trusted relationships doesn’t simply reduce the complexity of the system; on the contrary, it allows it to coordinate and exist, increasing the overall possibility space.

This text is the first part of a series. In the next iteration, we will investigate what the spatialization of Web of Trust could mean for peer-to-peer shared realities.

[1] The idea of Umwelt has been developed by the Baltic German biologist Jakob Johann von Uexküll and may be defined as the perceptual world in which an organism exists and acts as a subject