1 Introduction

The advent of cryptocurrencies has shaken the world of finance, the monetary system, and other economic institutions at their roots. Although the technical innovations accompanying cryptocurrencies are sometimes hard to grasp, not so are the reasons why cryptocurrencies have been introduced, which Satoshi Nakamoto, Bitcoin’s celebrated (and pseudonymous) founding father, took pains to clarify from the very outset. Most, if not all, of the origins of cryptocurrencies have to do with trust issues.

The root problem with conventional currency is all the trust that’s required to make it work. The central bank must be trusted not to debase the currency, but the history of fiat currencies is full of breaches of that trust. Banks must be trusted to hold our money and transfer it electronically, but they lend it out in waves of credit bubbles with barely a fraction in reserve. We have to trust them with our privacy, trust them not to let identity thieves drain our accounts. (Nakamoto 2009)

What is needed to fix these continued trust breaches, Nakamoto says, “is an electronic payment system based on cryptographic proof instead of trust, allowing any two willing parties to transact directly with each other without the need for a trusted third party” (Nakamoto 2008; p. 1).Footnote 1 Since the elimination of trust represents a core principle of cryptocurrencies, it is somewhat surprising that scholarly discussions have tended not only to remain unclear about what trust means (Jacobs 2021), but perhaps more strikingly to diverge from Nakamoto’s take on the concept. While the phrase “trusted third party” suggests that Nakamoto sees trust as an interpersonal (or organizational or institutional)Footnote 2 relationship – in other words, he considers banks and other financial intermediaries to be untrustworthy subjects – the relational dimension of the concept has been lost on many. The common claim that “cryptocurrencies like Bitcoin purport to replace trust with technology” (Bratspies 2018; p. 3), a broadly correct description of Nakamoto’s goal, unfortunately lays itself open to the misunderstanding that cryptocurrencies replace trust in people and organizations with trust in technology.Footnote 3

With the conceptual distinction between “trust” and “confidence,” De Filippi et al. (2020) marked a turning point in discussions about the fiduciary nature of cryptocurrencies and the blockchain. Since trust is regarded in philosophy mostly as an interpersonal concept, any fiduciary relationship held with a technology or a mechanism must fall under a separate category, which De Filippi and colleagues call “confidence” and the philosophical literature often calls “reliability” (Hawley 2014) or “reliance” (Baier 1986; see McLeod 2021). This conceptual distinction represents a powerful tool for assessing the diverse criticisms of cryptocurrencies’ claimed trustlessness (see Bratspies 2018). In light of this distinction, critics can no longer meaningfully blame Bitcoin for failing to abolish trust by pointing to users’ reliance on blockchain technology, simply because reliance, however strongly felt, does not imply trust. On the contrary, critics could rightly blame Bitcoin for not living up to its promises if it is proved that genuine forms of interpersonal trust lurk somewhere in the crypto ecosystem. The trust-reliance distinction has the potential to inform, more generally, any assessment, of both descriptive and normative variety, of cryptocurrencies’ trust profile. There seems to be no better way to study not only, as descriptive studies do, whether the current organization of cryptocurrencies involves genuine trust, but also, on the normative side, whether trust can be completely eliminated, even if only in principle, or whether the complete elimination of trust is, if possible, desirable.

To address these issues, this paper considers the trust-reliance distinction as a necessary but still not sufficient condition. The attempt to sift trust from reliance in the domain of cryptocurrencies would be impracticable, we argue, if not accompanied by a corresponding understanding of what cryptocurrencies are. To provide this understanding we read cryptocurrencies (and the related blockchains) – especially the two most popular, Bitcoin and Ethereum – through the lens of the philosophical hypothesis of the “extended mind” (see Menary 2010), hence seeing them as institutions that extend individuals’ cognitive processes. Two models of institutional cognitive extension will be compared, one introduced under the heading of the “extended mind” hypothesis (Clark 1997; Clark and Chalmers 1998), and the second the enactive model of the “socially extended mind” (Gallagher 2013). As we’ll show, each institutional model of cognitive extension is related to a prevalent fiduciary model, respectively reliance-based and trust-based. To assess which institutional model cryptocurrencies align with, we need to discuss which fiduciary model they are established upon. The bottom line of our exercise is to show that if trust represents cryptocurrencies’ fundamental fiduciary model, then they are best understood as socially extended “cognitive institutions” (Gallagher and Crisafi 2009). Putting it in a way that connects the descriptive and normative sides of our inquiry, the unavoidability and desirability of trust suggest categorizing cryptocurrencies as instances of socially extended cognitive institutions.

Why is it important to understand cryptocurrencies within the framework of extended cognition? The first reason, we think, is to emphasize the genuinely cognitive function(s) played by cryptocurrencies. Individual users cannot perform some crypto-related cognitive processes on their own, hence the very possibility of those processes depends on the existence and working of cryptocurrencies. This fact would suffice to characterize cryptocurrencies as mind-extending institutions. Within this argument, however, we take a further step concerning the specific form of this mind extension. Since technology usually takes the spotlight in discourses about cryptocurrencies, it is a short step to think that also mind extension is mostly due to technology. This is consistent indeed, we’ll see, with Andy Clark’s view of mind extension. What we’ll emphasize here is that the cognitive functions played by cryptocurrencies are ultimately dependent on specific patterns of interpersonal interactions that in turn depend on trust, and therefore that entire classes of cognitive processes would not be possible if there wasn’t trust. In effect, we’ll show that trust plays a cognitive role because it is necessary for the cognitive processes to take place. Seeing trust as a requirement for cryptocurrencies’ cognitive functions can be understood, moreover, in terms of seeing it as a lever in cryptocurrencies’ success. Establishing and promoting trust means, in other words, according to our framework, attending to the very possibility, and then to the effectiveness and sustainability, of the cognitive processes involved in the world of cryptocurrencies.

The next section provides a brief introduction to the two views of extended cognition and the alternative institutional models they support. Section 3 provides a brief introduction to cryptocurrencies sufficient to guide the reader throughout the discussion. Section 4 offers a preliminary attempt to describe cryptocurrencies in light of the institutional models of the extended mind, clarifies which cognitive processes cryptocurrencies extend, and discusses the role trust and reliance play in the extension of these processes. The ensuing section, Sect. 5, attempts to answer a first normative question: could cryptocurrencies extend the identified cognitive process trustlessly, even if only in principle? Section 6 then poses a second normative question: would it even be desirable, if possible, to extend cryptocurrency-related cognitive processes trustlessly? To answer this latter question, the focus will switch from Bitcoin to Ethereum, a blockchain that goes in the direction of regulating algorithmically a broader range of social relationships within a community.

2 Two models of the extended mind and institutions

The extended mind is a hypothesis proposed by philosophers Andy Clark and David Chalmers (1998) concerning the possible realization of an individual’s cognition outside the traditional (biological) locus of the brain. Consider Clark and Chalmers’ famous example of Otto who suffers from mild cognitive impairment and uses a pocket notebook as an extension of his memory. Under certain conditions, the most important of which is that the external resource (in the example, the notebook) plays the same function (memory storage) as internal resources (neurons), the functionally equivalent external resource can be said to extend an individual’s mind. This is based on what has come to be known as the “parity principle”: “If, as we confront some task, a part of the world functions as a process which, were it done in the head, we would have no hesitation in recognizing as part of the cognitive process, then that part of the world is (so we claim) part of the cognitive process” (Clark and Chalmers 1998; p. 8).

Two things interest us here about this hypothesis. First, following a concession made in the paper with Chalmers, Clark (1997; 1998; Ch. 9) maintains that institutions can be “vehicles” of mind extension exactly like notebooks insofar as they serve to externalize specific cognitive functions. If we consider again the case of memory, an example of institutional externalization of memory is the institution of national holidays (such as the Fourth of July in the US), which helps remind citizens of important historical facts. Clark calls these forms of institutional mind extension “scaffolding institutions”. Second, in an attempt to fence off criticisms about the possible inflation of what is deemed cognitive, Clark (2008) introduced three more restrictive criteria for mind extension.Footnote 4 They are:

  1. 1.

    That the external resource be reliably available and typically invoked. […]

  2. 2.

    That any information thus retrieved be more-or-less automatically endorsed. It should not usually be subject to critical scrutiny (unlike the opinions of other people, for example). […]

  3. 3.

    That information contained in the resource should be easily accessible as and when required. (p. 79)

The second criterion has been thought to involve the trustworthiness of the resource, which is why the three criteria are together known in the literature by the name “trust and glue,” a phrase alluding not only to physical proximity but also to the fiduciary relationship an agent needs to entertain with external resources for them to be mind-extending. But is “trust” the right category here? Since Clark demonstrates a predilection for examples that involve artifacts like notebooks and because people typically appear in his framework as cogs in larger mind-extending machines or mechanisms (we’ll come back to this important point below), reliance is probably the more philosophically fitting category.

As an alternative to Clark’s framework, Gallagher (2013) proposed the idea of the socially extended mind, which, among other substantial differences, considers institutional mind extension not merely a possible but a ubiquitous feature of cognition. The institutional model associated with it is called a “mental” or “cognitive” institution (Gallagher and Crisafi 2009). According to the definition, cognitive institutions are not just those that allow agents “to perform certain cognitive processes in the social domain but, more importantly, without which some of the agents’ cognitive processes would not exist or even be possible” (Petracca and Gallagher 2020; p. 747). The underlying idea is, in brief, that cognitive institutions make cognitively possible what would be otherwise not possible. Consider, for instance, markets as forms of cognitive institutions (Gallagher et al. 2019; Petracca and Gallagher 2020), hence not just as places that help or enable individuals to satisfy their needs but where these very needs are created, or where new knowledge is created (typically through prices) which would not be available otherwise (see also Dekker 2022).

At this point it is important to clarify that, in contrast to Clark’s model of institutions as mind-extending technologies or mechanisms, cognitive institutions are conceived primarily as forms of social or interpersonal relationships. This is not to deny that socially extended cognition can often involve mechanical and technology-based operations (for example, relying on communicative technologies), but even in such cases interpersonal relationships hold a conceptual primacy, for at least two reasons. First, in the case of socially extended cognition, relevant technologies do not count primarily as instrumental tools, but rather as vehicles that can both shape and express social relationships.Footnote 5 Second, purely mechanistic mind extension without human/social intervention is unsustainable in the long term because it can lead to social pathologies, like the progressive erosion of relational potentials (Gallagher and Petracca 2024). The fiduciary model behind cognitive institutions is, in a nutshell, based on (relational) trust.

As noted, Clark offered his three “trust and glue” criteria as a way to address an objection that the extended mind hypothesis could lead to the boundless multiplication of cognitive phenomena, known as the “cognitive bloat” objection (Rupert 2004). Although Clark’s criteria might be problematic for a number of reasons (including the implied understanding of trust, for instance), they point to the idea that what creates mind extension is not the nature of the resource, a notebook or an institution (where one may seem a more plausible mind-extending resource than the other), but the relationship between the individual and the external resource, more specifically known as their coupling. An agent might not be in the right mind-extending coupling with a notebook when she uses it like a fan to cool off, but might be so with an institution when the institution helps her to think or to provide the cognitive wherewithal to solve a problem. Moreover, the coupling depends on an agent’s intentions and skills (an unskilled user’s cognitive processes might be diminished rather than extended by a hard-to-use resource).

Focusing on the coupling relationship is useful to respond to another possible misunderstanding about institutional mind extension that concerns the following question: how is it possible to distinguish an institution that is “cognitive” from one that is not? An initial way to answer this is to recall, as Hodgson (1988) pointed out, that institutions generally serve a “cognitive function” when individuals engage with them to make sense of their environment (see also Denzau and North 1994). Some institutions are especially designed to perform specific cognitive functions on behalf of individuals (an example is prediction markets, which help individuals to make predictions; see Gallagher et al. 2019). But the point is that any institution can become “cognitive” at some point, even those not designed to be so, when individuals engage with them for purposes of sense-making and problem-solving. This is a requirement for the institution to function as cognitive, and sometimes this is a requirement for an institution to function at all. Markets themselves have not been primarily designed to be cognitive institutions, but it turns out, as Hayek (1945) highlighted, that their cognitive component is a requirement for them to be pragmatic, that is, effective means to exchange goods and services.

3 A brief primer on cryptocurrencies

Cryptocurrencies are forms of digital currency, which means that they are digital mechanisms fulfilling the three functions typically assigned to money: medium of exchange, store of value, and unity of account.Footnote 6 Bitcoin, the first and currently largest cryptocurrency, possesses most of the features typically associated with cryptocurrencies, so we model the discussion around it (see Antonopoulos 2014) although cryptocurrencies may differ considerably from each other in important regards.

A ubiquitous feature of cryptocurrency is decentralization, since no centralized authority, called by Nakamoto “trusted third party,” either issues and regulates the currency or verifies the transactions taking place on the network. Transactions’ verification is assigned to peers, in Bitcoin called miners, who get rewarded with newly minted tokens and transaction fees for the service. For any new block of transactions, miners compete against each other to be the winning validators (in order to get the reward) by solving cryptographic puzzles (called proof-of-work) that require the allocation of considerable computing power, and therefore the expense of massive amounts of electrical energy.Footnote 7 The amount of such expense along with the reward in the same cryptocurrency set miners’ incentive not to cheat in their role as validators. Once a block of transactions has been validated, it gets added to an existing public ledger (called blockchain) that includes the entire transaction history and that, importantly, cannot be altered if not through the consensus of the majority of nodes. To provide an idea, among the many possible examples, of how cryptocurrencies may differ, we can mention the presence or absence of an upper limit for issuable tokens, the amount of that limit, the number of allowed transactions per unit of time, etc.

This short description refers mostly to a cryptocurrency’s architecture and operations (analyzable in terms of software, hardware, and organization), which are part of what we may call the “internal” functioning of cryptocurrencies. But like all forms of money, cryptocurrencies are also negotiated on the market, thus subject and vulnerable to a series of factors somewhat “external” to their protocol. Moreover, an informed assessment of cryptocurrencies should be cognizant that money is only one of the things that are implementable on a blockchain, so that the latter can be considered a more general disruptive institutional innovation on its own (Davidson et al. 2018). Although, as we mentioned, we think that all institutions may work as forms of extended cognition, we also agree with Frolov (2023) that the “wired and weird world of digital institutions” is particularly suited to “new ways of thinking [… suggesting] methodological experiments and new theoretical imaginaries” (p. ix). That’s what we’ll try to outline in what follows.

4 Which cognitive processes do cryptocurrencies extend?

Before deciding which extended institutional model cryptocurrencies align with, we first need to make sure that they are instances of extended cognition. The benefits of reading cryptocurrencies through the extended cognition framework concern not only finding what we think is a fitting ontological model but also providing a less speculative recognition of what cryptocurrencies really do (i.e., the extension of cognitive processes), and based on this recognition identifying levers (e.g., trust) to do it better.

Extended cognition allows us to look at two phenomena central to the cryptocurrency business as genuine cognitive processes:

  • verification against double-spending;

  • creation of an immutable transaction memory.

Of course, these phenomena are closely related as two sides of the same coin, but we argue it is more useful to consider them separately, the first as an economic and the second as an epistemological phenomenon. Given the spatial ubiquity of electronic tokens, the main service an electronic payment system must offer is protection against accidental or fraudulent spending of the same token for multiple transactions. As Nakamoto (2008) puts it in a way that emphasizes the cognitive nature of the issue, what users of cryptocurrencies are in search of is “a way for the payee to know that the previous owners [of the token] did not sign any earlier transactions” (p. 2, emphasis added). While in traditional finance the guarantee of transactions is entrusted to financial intermediaries, in cryptocurrencies like Bitcoin this is accomplished, as indicated, by strongly incentivized miners and by the existence of a public ledger. The cognitive process they perform, verification, is inherently extended because if it were delegated to the autonomy of single payees, they would not have sufficient motivation to report correct information in the ledger. In other words, if verification were not institutionally extended, it wouldn’t be at all.

Once this is agreed upon, the difference between traditional financial intermediaries and miners is said to lie in the specific structure of their economic incentives. While miners’ fraud would destroy the source of their reward value, namely, security, thus wasting the energy employed in the verification process (Ammous 2018), fraud by traditional financial intermediaries does not typically affect a currency’s profile of safety. Bitcoin’s incentive structure looks, in a word, sound: everyone has an incentive to do what is good both for oneself and the network. Kroll et al. (2013) model Bitcoin as an economic game – called the “mining game” – in which they “explicitly assume that players will behave according to their incentives” (p. 7), an assumption whose soundness, they add, depends on the strength of those incentives (p. 2). In such a game-theoretic narrative, monetary incentives represent all that the fiduciary relationships in the network are based upon: Bitcoin’s users don’t trust miners to perform correct verifications, they rely on the incentive mechanism to induce miners to do so. This is consistent with Clark’s model of institutional mind extension where miners, although individuals, are not trusted qua individuals but are relied upon as cogs in a sound economic mechanism (see Clark 1997; p. 272).

Although this interpretation seems to place cryptocurrencies definitively in Clark’s category of scaffolding institutions, closer inspection reveals that genuine trust continues to be important. Consider the so-called “51% attack,” the best-known potentially fatal attack on the Bitcoin network (Antonopoulos 2014; Chap. 8). As soon as one miner or pool of miners obtains 51% of the network’s computational power (hashrate), they can change the consensus rules by which new blocks are added to the blockchain, thus double-spending and preventing other miners from adding new blocks at will. This is not just a theoretical possibility since in June 2014 a mining pool named GHash.io in reality exceeded the 51% threshold, although this didn’t lead to an attack on the network.Footnote 8 Concerning possible cases of misbehavior in the Bitcoin network, it is somewhat curious that when Nakamoto (2008) discusses them, he does so by using a morally connoted language: miners and other network nodes get distinguished into “honest” and “dishonest” ones.Footnote 9 Although in Nakamoto’s trustless world these concepts merely refer to the consistency or inconsistency of miners’ behavior with their economic incentives,Footnote 10 nothing that involves morals in any case, Nakamoto’s use of moral categories is quite indicative. In fact, miners may always deviate from the behavior dictated by economic incentives in case reasons external to Bitcoin’s incentive structure come into play. In what is called a “Goldfinger attack” (Kroll et al. 2013), the only aim of which is to destroy the Bitcoin network, the decision to attack is certainly influenced by economic constraints, as the attacker might trivially not be able to afford the attack, but it does not arguably reduce to them. A strong destructive drive, experience sadly tells us, may overcome or simply disregard economic reasons. This leads to a simple consideration: trust in miners’ lack of reasons to destroy the network runs deeper than any form of reliance on Bitcoin’s incentive structure.

Similar considerations apply to the epistemic merits of Bitcoin’s technology, the blockchain, described in the following terms by a Bitcoin enthusiast.

[I]t did occur to me once that the Bitcoin ledger of transactions might just be the only objective set of facts in the world. You could argue (as many philosophers do) that every fact is subjective and its truthfulness is based on the person stating or hearing it, but the Bitcoin ledger of transactions is created through converting electricity and processing power to truth without having to rely on the word of anyone. (Ammous 2018; p. 174, fn. 4)

The use of such serious concepts as truth and objectivity has led some to identify in Bitcoin a basis of “epistemological utopianism that goes beyond money” (Dodd 2018; pp. 48–49). In addition to the realistic possibility that successful 51% attackers may write, as winners usually do, their own history by manipulating blocks at will, it is the attribute “objective” used to connote blockchain’s regular working that raises particular concerns. “Converting electricity and processing power to truth without having to rely on the word of anyone” looks a lot like what, in a widely occurring metaphor of technology-driven objectivism, cameras are thought to do with the conversion of photons into objectivity. But what gets recorded by the blockchain can’t be objective in any reasonable sense, it can’t be the bare or impersonal truth (whatever that is) due to the fact that the blockchain works through a “consensus” principle that is inherently intersubjective.Footnote 11 Any new valid block, any new piece of the immutable chain must meet the criteria chosen collectively by the network’s nodes. Any change in consensus rules – which may cause the blockchain to dramatically “hard fork,” namely, split in two whenever a modification proposal is not accepted by a crushing majority of nodes – changes the very way in which truth is established.Footnote 12 This means, to come to implications for our discussion, that the blockchain does not work as an institutional technology that simply records objective states of affairs – or, to put it in equivalent terms, it does not work as an institutional technology simpliciter and therefore as a scaffolding institution. On the contrary, cryptocurrencies testify to the intersubjective and conventional nature of monetary institutions’ epistemic ambitions, which, because they are vulnerable to deception (not just defection), are fundamentally trust-based.

5 Normative question one: is trust eliminable?

As this paper advocates and was first suggested by De Filippi et al. (2020), sifting trust from reliance is a precondition for any assessment of cryptocurrencies’ fiduciary profile. Although a rigorous application of the distinction reveals a general over-attribution of trust by cryptocurrencies’ critics, who often see trust where there is in fact reliance, it helps nonetheless to prove that genuine trust lies at the core of cryptocurrencies. Bratspies (2018), probably the most comprehensive description of interpersonal trust in Bitcoin, provides a useful summary of where trust seems to lurk in the ecosystem.

With regard to the blockchain itself, users are (1) trusting developers to build secure software, (2) trusting miners not to collude or attack the blockchain, and (3) trusting the wider cryptocurrency governance process not to approve a malicious hardfork. With regard to using the currency, users are trusting (1) that markets are not being manipulated, (2) that wallets will generate secure keys,Footnote 13 and (3) that trading platforms are using best security practices. (pp. 19–20)

“That is an awful lot of trust for a trustless system,” Bratspies exclaims (ibid.). There is, however, a conceptual distinction to be made between these sources of trust, which follows a distinction introduced in Sect. 3 above. Bitcoin’s protocol cannot be held responsible for all of them, as some are proper of any financial asset traded on the market (Gandal et al. 2018). Any asset is indeed vulnerable to the “pump and dump” price manipulation, consisting first in creating artificial hype over the instrument and then profiting from the price rise. Typical of ICOs (Initial Coin Offerings, analogs of IPOs for stocks), pump and dump leverages precisely on people’s trust; but even a staunch crypto critic would recognize that this is not the trust cryptocurrencies have promised to get rid of, nor could they ever credibly promise that. In this regard, a tentative distinction between internal and external forms of trust, where internal forms concern cryptocurrencies’ protocol directly and external forms involve the adoption or exchange of cryptocurrencies, seems useful, although not easy to implement.

As for internal forms of trust, it must be remarked that not all philosophically informed analyses agree that trust lies anywhere in Bitcoin. Drawing on Katherine Hawley’s (2014) definition of trust as faith in someone’s commitment, Lipman (2023) argues that since no stakeholder in the Bitcoin’s ecosystem makes commitments akin to those made by traditional financial intermediaries (e.g., a commitment to an inflation target, as in the case of central banks), no trust is therein really involved. “Commitment implies intention and taking up a responsibility, but the sort of postulational conventions at stake in social construction may be unintentional and unconscious, and simply emerge from certain patterns of coordinated and intentional behaviour,” says Lipman (pp. 799–800), as if patterns of coordinated and intentional behavior were not enough to express stakeholders’ genuine intentionality, and hence commitment. To this it may be replied that implicit commitment can be as intentional and binding as explicit commitment, and sometimes even more so. Bitcoin’s promise as a truly distributed technology has been supported by Bitcoin developers’ commitment to never unilaterally change the network’s consensus rules, although this commitment has mostly been implicit. (Bitcoin developers’ standoffishness is even more remarkable if seen in the light of the second largest blockchain, Ethereum, whose development team is known for exercising a much more interventionist approach to consensus rules).Footnote 14 It is hard not to read this as a form of Bitcoin developers’ commitment and therefore, by Hawley’s standards, as a trust condition.

So far we have offered a descriptive account of Bitcoin’s trust, which means that we’ve shown how trust is currently present in the Bitcoin’s ecosystem. But could it be different? Would a more accurate design, more computing resources, new hardware, a newly minted cryptocurrency perhaps, fulfill Bitcoin’s promise of trustlessness? This is a normative question that concerns whether trust is an accidental or instead a necessary, unavoidable component of cryptocurrencies. If it were just accidental, there would still be space for seeing cryptocurrencies as scaffolding institutions; but if trust is necessary, as we think it is, there would be no alternative, from an extended cognition point of view, but to consider cryptocurrencies as cognitive institutions. To settle this central question, the literature in the sociology of money, and particularly the work of the late Nigel Dodd, can be of decisive help.

Dodd (2018) points out what he sees as a true paradox at the core of the Bitcoin enterprise, where the extreme techno-ideology would undermine the success of Bitcoin as money. In Dodd’s own words, worthy of being reported at length,

Bitcoin will succeed as money to the extent that it fails as an ideology. The currency relies on that which the ideology underpinning it seeks to deny, namely, the dependence of money upon social relations, and upon trust. Insofar as Bitcoin has been successful qua money, it is because of the community that has grown up around it. Ironically, however, this community is sustained by the commonly held belief that Bitcoin has replaced social relations – the trust on which all forms of money depend – with machine code. This belief is a fiction. Bitcoin has thrived despite, not because of, its reliance upon machines. (p. 37)

To bring the paradox to light, Dodd needs to overcome two common misconceptions at the root of Bitcoin’s ideology. The first one concerns the assumption that Bitcoin is entirely based on technology. This assumption, however, is easy to prove wrong, as other social scientists, economists for instance, have long been aware that “[n]one of the rules of Bitcoin are self-executing; any rule can be ignored by users. […] Cryptographic rules and other technical rules are like all other rules, in that they exist only as words on paper and therefore will be followed only to the extent that players have incentives to follow them” (Kroll et al. 2013; p. 9). Thus, while Bitcoin is certainly a technical infrastructure, it is before that (or perhaps because of that) a social infrastructure. But to the extent that economic incentives are thought of as the only driver for agents’ behavior in the Bitcoin ecosysyem, the genuine social dimension is obscured. That is, it is obscured, as seen above, by the (ideology of) techno-utopia according to which economic mechanisms come before the people who populate them. This is the second misconception that Dodd needs to overcome, and he does so by shifting the focus onto the role that community plays in the Bitcoin ecosystem. In this regard, a video released by Bitcoin activists on Youtube in 2014, entitled “The Declaration of Bitcoin’s Independence” is enlightening.Footnote 15 This milestone in cryptocurrency sociology shows Bitcoin activists enacting fragments of a common script voicing the community’s feeling of having been betrayed by traditional finance.

We’ve been cyclically betrayed, lied to, stolen from, extorted from, taxed, monopolized, spied on, inspected, assessed, authorized, registered, deceived, and reformed. We’ve been economically disarmed, disabled, held hostage, impoverished, exhausted, and enslaved [moment of suspense]….and then there is Bitcoin.Footnote 16

In the video, cyberpunks and crypto-anarchists are mentioned as frontline social movements in the Bitcoin project, and this is interesting for us since Clark’s version of the extended mind also has some roots in the 1990s cyberpunk milieu.Footnote 17 It is tempting, then, to find a common background for Bitcoin’s paradoxical attempt to remove sociality by means of engaged social movements, and the extended mind’s inclination to see institutions as pieces of technological equipment while, in reality, their mind-extending potential stems mostly from the people enacting those institutions. The video, which closes with the loo** phrases “Bitcoin is ours. Bitcoin is,” would have been more truthful to its spirit if it had concluded: “Bitcoin is us.”

The Bitcoin community hasn’t gone that far, namely, identifying itself with the currency, probably aware that that would reveal the project’s contradiction; but other post-2008 movements critical of the monetary system have not refrained from such identification. Dodd (2014) discusses a poster that appeared during a 2011 London bank protest reading “We are the true currency,” a message able by its immediacy to point directly to the essence of money: social relations. In his analysis, Dodd follows Georg Simmel’s (2004[1900]) seminal view about the origin of money (condensed in the famous motto “money is only a claim upon society”), radicalizing it a little. The London protest poster reveals, Dodd argues,

that money’s value is derived from social life. This idea would suggest that money is not simply a claim upon society, but—ideally— is social life, gaining its value not from the institutions that produce it but from the people who use it. (Dodd 2014; p. 9)

Another follower of Simmel’s message, Geoffrey Ingham, is even more explicit about the nature of currencies as entities constituted by social relations, which aligns with the enactive idea of “constitution” supported by cognitive institutions (De Jaegher et al. 2010). Social relations, in other words, are not for Ingham incidental or arbitrary features of money but they constitute the very core of money, in the sense that without them a currency would not exist or even be possible.

Obviously, money is socially produced in the sense that it does not occur naturally, and it also mediates and symbolizes social relations […] However, I wish to go further and argue that money itself is a social relation. By this I mean that “money” can only be sensibly seen as being constituted by social relations. (Ingham 1996; p. 510)

As an American pragmatist might say, the concept of money is cashed out in social relations. Would Bitcoin be the first case of money esca** this Simmelian argument? After all, wasn’t Nakamoto so much aware of money’s social character that he tried to build a version of it that was different exactly in this regard? According to Simmel’s followers (see also Paniagua 2018), however, this would simply be impossible: “all forms of money are social relations” (Ingham 1996; ibid.), which means that if something is money, it is also necessarily a social relation. Moreover, these are fully-fledged social relations in that they cannot be reduced to economic incentives: Bitcoin’s social relationships concern, Dodd remarks, “varying modes of shared existence and experience” (Dodd 2018; p. 52) that go much beyond the mere monetary aspect, and among which there is also shared trust. We’ve insisted on this Simmelian theme because it represents the theoretical support for the statement that Bitcoin’s attempt to remove social relations from the currency is a self-defeating, even conceptually impossible goal.

To this conceptual paradox in Bitcoin’s project correspond empirical paradoxes. Recently, Jalan et al. (2023) have shown that citizens’ interest in Bitcoin, along with their inclination to adopt it, is higher in societies where baseline interpersonal trust is also higher. What is the significance of this evidence? For one, it represents an empirical challenge to the idea that Bitcoin adoption is a response to absolute trust issues within society, a narrative fostered by Nakamoto himself who presented Bitcoin as a response to people’s distrust of financial institutions after the 2008 financial crisis. The evidence provided by Jalan et al. (2023) shows that Bitcoin is instead better thought of as a reaction to trust breaches within already trusting societies, thus concerning relative and not absolute trust levels. In any case, the very positive relation between trust and Bitcoin adoption found by Jalan and colleagues is challenging, if not entirely paradoxical, considering that even if the relation were inverse, going from higher Bitcoin adoption to higher trust,Footnote 18 as supposed for instance by Kowalski et al. (2021), we would not apriori expect that. If Bitcoin really removes the need to trust others as it claims, the most plausible expectation is of lower trust levels within societies that adopt it. This important implication, and more generally the implications of cryptocurrencies’ trustlessness will be the object of discussion in the next section.

6 Normative question two: is trustlessness desirable?

Although the eliminability of trust has been the main concern of normative discussions of cryptocurrencies, another important normative issue concerns whether trustlessness, if achievable, is desirable in the first place. Trustlessness’ desirability is of course implied every time crypto supporters enumerate reasons for not letting a trusted third party run monetary policy or validate transactions; but this kind of argument, as it is, does not concern the desirability of trustlessness per se which is what interests us here. To address this broader issue, it is necessary to go beyond the monetary aspect and regard trustlessness as a characteristic of any possible blockchain-based institution, hence bringing the discussion to whether trustlessness represents a desirable principle of social organization. Given the broadening scope of the inquiry, we will switch the focus of this section to Ethereum, a blockchain that compared to Bitcoin pursues a grander project of institutional organization based on (trustless) “smart contracts”.Footnote 19

Although this is rarely noticed, Ethereum’s way to trustlessness differs from Bitcoin’s in important regards. One is that, in striking contrast with Nakamoto’s standoffishness, Vitalik Buterin, Ethereum’s principal founder, has exercised a decidedly interventionist approach to the network which earned him the name of Ethereum’s “benevolent dictator” – with conspicuous implications, commentators have noticed, for the network’s perceived level of trust-dependence (see Ammous 2018). Buterin has also been much more careful than Nakamoto in handling the trust concept. “Blockchain applications are never fully trustless,” Buterin remarks, “but some applications are much closer to being trustless than others,” which is why he prefers to reason in terms of “trust minimization” and “degrees of trust” (Buterin 2022; Part 3, essay 7). Buterin has even offered a definition of trust as “the use of any assumptions about the behavior of other people” (ibid.), an interesting way of putting the trust concept which prompts two considerations. The first is that Buterin’s definition seems to come in the footsteps of economists’ definition of trust as a probabilistic assumption about others’ behavior (Gambetta 2000), an approach that doesn’t really distinguish between trust and reliance (see McLeod 2021).

The second consideration is that different factors can contribute to determining the assumptions about behavior. “When you run a piece of code written by someone else, you trust that they wrote the code honestly,” Buterin says, “whether due to their own sense of [moral] decency or due to an economic interest in maintaining their reputations” (Buterin 2022; Part 3, essay 7). Buterin proves in other words to be inclusive about the range of individuals’ motives, more than Nakamoto ever was.

Individual actors may be motivated by extra-protocol motives, they may get hacked, they may get kidnapped, or they may simply get drunk and decide to wreck the blockchain one day and to hell with the cost. Furthermore, on the bright side, individuals’ moral forbearances and communication inefficiencies will often raise the cost of an attack to levels much higher than the nominal protocol-defined value-at-loss. This is an advantage that we cannot rely on, but at the same time it is an advantage that we should not needlessly throw away. Hence, the best protocols are protocols that work well under a variety of models and assumptions. (Buterin 2022; Part 2, essay 2, emphasis added)

In the face of such pluralism about motives (or perhaps because of it), Buterin turns out to be a staunch advocate of economics and game theory for the design of institutions. “I consider economics and game theory to be a key part of cryptoeconomic protocol analysis, and consider the primary academic deficit of the cryptocurrency community to be not ignorance of advanced computer science, but rather of economics and philosophy” (Buterin 2022; Part 1, essay 4). But “Economics is not everything” (Buterin 2022; Part 2, essay 2), he is also eager to clarify.

In the penultimate sentence above, there is a keyword that recurs in Buterin’s writings: community. In a recent essay, Jackson (2023) reports the strident experience of reading Buterin’s collected essays where “trustlessness” is as frequent a word as “community”: but isn’t trustlessness, he wonders, what erodes communities in the first place? How is it even conceivable to establish a community upon trustlessness? Echoing Karl Polanyi, Jackson argues that by trying to bring such opposites together “Buterin and his comrades are an unreflexive demonstration of […] entirely marketized subjectivity” (Jackson 2023; p. 9). Certainly Jackson’s sort of criticism applies as well, and possibly even more, to Bitcoin, where, as one supporter put it, the protocol “operates under the working assumption that all computer nodes are hostile attackers” (Ammous 2018; p. 242). How can there be anything resembling a community where everyone acts under the working assumption that the other is a hostile attacker?Footnote 20 As one commentator remarked about the prospect of a financial world in which trustlessness is eventually achieved, “[t]here are no financial instruments that will protect you from a world where we no longer trust each other” (quoted in Bratspies 2018; p. 57).

Gallagher and Petracca (2024) have argued that genuine interpersonal relationships, hence trust-based relationships, are a requirement for cognitive institutions to work properly. While institutional technologies or market mechanisms may work well as mind-extending resources for a while, cutting out interpersonal relationships would at some point present the bill in terms of institutional pathologies. Technology and impersonal markets may cause cognitive and affective forms of atrophy, a disaccustom to interaction with others qua persons that can potentially affect any domain of social life. A certain degree of trust, therefore, is not only necessary, as discussed in the previous section, but desirable, as an institutional glue that helps hold institutions and citizens together and hence allows institutions to perform their cognitive functions better and more sustainably. We’ve already noticed that in Buterin’s writings the juxtaposition between community and trustlessness is not so strident as it may appear at first glance: Buterin is more aware than Nakamoto that community is a crucial factor in blockchain’s success and that complete trustlessness is self-defeating. In fact, Buterin’s true goal is to balance economic incentives and community spirit with the aim of building sustainable blockchain-based institutions.

A few quotes can give a sense of Buterin’s balancing effort. On the one hand, incentives are deemed crucial in the short term to make people interested in the blockchain business.

While Ethereum’s financial applications may be what initially excites many people in the cryptocurrency community, the long-term promise is arguably in the ways that Ethereum can work together with other, non-financial peer-to-peer protocols. One of the main problems that non-financial peer-to-peer protocols have faced so far is the lack of incentive—that is to say, unlike centralized for-profit platforms, there is no financial reason to participate. (Buterin 2022; Part 1, essay 2)

But it is sociality that proves to be the true glue of blockchain institutions in the long run.

On medium-to-long time scales, humans are quite good at consensus. Even if an adversary had access to unlimited hashing power, and came out with a 51% attack of any major blockchain that reverted even the last month of history, convincing the community that this chain is legitimate is much harder than just outrunning the main chain’s hashpower. They would need to subvert block explorers, every trusted member in the community, the New York Times, archive.org, and many other sources on the internet; all in all, convincing the world that the new attack chain is the one that came first in the information-technology-dense twenty-first century is about as hard as convincing the world that the US moon landings never happened. These social considerations are what ultimately protect any blockchain in the long term, regardless of whether or not the blockchain’s community admits it. (Buterin 2022; Part 2, essay 2)

Although Buterin’s best energies and insights are devoted to fine-tuning economic incentives, because “a blockchain protected by social consensus alone would be far too inefficient and slow” (ibid.), the passages declarative of his philosophy do not avoid the relevance of trust. Trust is “a precious quality that underlies decent societies” (Buterin & Lanier, 2018, p. xxiv) and must be preserved. Among Ethereum’s “contradictions,” Nathan Schneider, the editor of Buterin’s writings, lists the jarring juxtaposition also noted by Jackson between a “vibrant community and relentless mistrust”: but in Buterin’s oeuvre such contradiction seems to find resolution, at least theoretically, in the distinction between short and long term: while incentives are designed to take care of the short term, trust would be the long-term resource of the network.Footnote 21 But for all of Buterin’s sophistication and shrewdness in acknowledging the importance of trust on paper, it must be conceded to Ethereum’s critics that the project still relies heavily on incentives and the rhetoric of trustlessness. Smart contracts are designed to influence and possibly take over all relevant aspects of sociality, much more than Bitcoin’s self-limitation to the monetary domain could ever do. If smart contracts really took hold and were to spread, no one could guarantee that they wouldn’t snowball into a pathological dishabituation to trusting leading to social atrophy. This represents a dramatic possibility that supporters of smart contracts need to be aware of, which no mere lip service to trust could prevent from happening. In other words, trust needs not only to be acknowledged but also promoted and practiced.

7 Concluding remarks

Trustlessness has been at the crossroads of all cryptocurrencies’ hopes and contradictions; what started as cryptocurrencies’ main selling point has turned out to be exactly what they cannot achieve, and even more strikingly given the resounding promises, what they shouldn’t desire to achieve. What to do with what sounds like real disillusionment? This paper has suggested that the disillusionment of trustlessness offers in return an important chance to explore the nature of cryptocurrencies and other blockchain-based forms of social organization. Cryptocurrencies, we’ve seen, can be thought of as institutions that extend individuals’ cognitive processes in line with the philosophical idea of extended cognition, and in addition we’ve argued that the extended-cognition model that fits cryptocurrencies best depends on the nature of the fiduciary relationship holding between individuals and institutions. If Clark’s model of scaffolding institution, with its emphasis on instrumental, mechanistic, and technological forms of mind extension based on reliance could be originally considered a natural model for cryptocurrencies, the newly acknowledged centrality of trust requires that we change the extended-mind model to that of socially extended cognitive institutions. In a cognitive institution framework, the variety of epistemic and pragmatic roles played by institutions for the people who engage with them can be carried out only in specific patterns of interpersonal relationships, and hence of trust.

Since Nakamoto’s resounding claims about the eliminability of trust, advocates and designers of cryptocurrencies have demonstrated increasing awareness and caution about this issue, at least on paper. Here we’ve tried to shift the analysis of trust in cryptocurrencies from a descriptive to a normative perspective, which means that if we are on board with the idea that trust is both necessary and desirable, this awareness must have some implications for the current management of the cryptocurrency enterprise. Nigel Dodd’s insight that Bitcoin has thrived despite, not because of, the technological utopia surrounding it, means that the future of cryptocurrencies will probably depend on how much the issue of trust will become explicit and the object of specific attention. The socially extended cognition framework has allowed us to see that a great part of what cryptocurrencies do is cognitive work, for both the individual and the system, and that trust is a necessary condition for that cognitive work to be done. Beyond that, the centrality of trust means that it can become a “policy lever” for making extended cognitive processes more efficient and sustainable. It is then in the interest of crypto designers, users of cryptocurrencies, and policy-makers that trust turns from being an embarrassing topic to an acknowledged and promoted factor of cryptocurrencies’ sustainability and success.