FireFly network:

A Decentralized Substrate for AI Data & Services
Technical whitepaper

FireFly foundation

Version May 2021


The technology of the distributed ledger, the so-called blockchain, is expanding
beyond the creation of efficiencies in financial intermediation to include the management of consumption data originating from iot micro devices, the trading of
surplus renewable energy, and programmable transactions in the form of so-called
smart contracts. Although the majority of current applications involve payments of
some form, the key question going forward is whether the two technologies which
underpin bitcoin, namely a digital token and the blockchain, can serve as a basis for other use cases.
In this paper we examine the evolution and design of marketplaces for digital
economies. We first introduce a simple economic model which we use to understand
the dynamics of firm boundaries, and the organisation of economic activity more
generally. In examining how the contrasting forces of scale and scope economies,
together with the relative costs of transacting within firms and markets have facilitated the emergence of decentralised marketplaces, we make use of a number of
core economic principles. These include the economics of transaction costs, ownership and control, the principal-agent problem, bounded rationality, information
asymmetry and trust relations. We also consider the technological antecedents of
blockchain technology, including the Internet protocols and the fundamental distinction between a distributed database and ledger.
Against this backdrop we consider the motivation behind the FireFly protocol
which provides a set of tools for the dynamic creation of intelligent marketplaces
for an agent-based economy.




In this paper we provide a first overview of how we see the evolution and design of digital
economies. Given that decentralised markets are emerging, economists and computer
scientists are learning how to adapt the tools developed to build and understand the
dynamics of centralised marketplaces. As such we see this project as on-going and will
provide the necessary updates to this paper.





1 Introduction                                                                                                                                                1

2 Online Markets and The Internet                                                                                                        3

3 Firms, Markets and Platforms                                                                                                              4

3.1 Firms and Markets                                                                                                                                4

3.2 Market Places as Platforms                                                                                                                6

3.3 Decentralised Markets                                                                                                                        7

3.4 Peer Production                                                                                                                                     7

4 Ownership and Control in a Digital Economy                                                                                  8

4.1 Ownership and Control                                                                                                                       8

4.2 Cryptocurrencies                                                                                                                                    9

5 Mechanism Design for Agent-Based Decentralised Economies                                               10

5.1 The Distributed Database and The Distributed Ledger                                                            13

5.2 Scalability and Payment Systems                                                                                                    15

5.3 Trust and Reputation                                                                                                                           15

5.4 Machine Learning                                                                                                                                  17

5.5 Decentralized Applications                                                                                                                18

5.6 Digital Tokens                                                                                                                                         19

6 Emerging Markets                                                                                                                                     21

6.1 Granularisation of Digital Business                                                                                                 21

6.2 Managing Identity                                                                                                                                 22

6.3 Digital Content                                                                                                                                       22

6.4 Decentralised Energy Markets                                                                                                         23

6.5 Machine-to-Machine Markets                                                                                                         23

7 FireFly Network                                                                                                                                         24

7.1 Combinatoric Innovation                                                                                                                    24

8 Conclusion                                                                                                                                                   26




Blockchain is not only useful in moving money, it’s useful in moving any
asset – whether that’s a unit of energy or a unit of computing power – in a
very transparent and reliable way.
David Bartlett, Chief Technology Officer for General Electric’s
digital power services business.

“Blockchain is to Bitcoin, what the internet is to email. A big electronic
system, on top of which you can build applications. Currency is just one”.

Sally Davies, FT Technology Reporter.

1 Introduction

The emergence of distributed ledger technology has coincided with fundamental changes
in the organisation of economic activity, and in particular how economic agents interact
within markets. The traditional model of exchange has a firm producing goods and/or
services upstream and selling them to consumers downstream. However, with the proliferation of online marketplaces and e-commerce, we have observed the rise of alternative
forms of organising economic activity, including platform markets and more recently
peer-to-peer exchange. One example where the application of blockchain technology has significant potential to disrupt and create new markets is the energy sector. As Gugler et al (2017) note, prior to the introduction of liberalization and regulatory reforms in the European Electricity sector, vertical integration of upstream and downstream operations of an electricity utility was the predominant organizational form in order to benefit from scope economies.
More recently, the emergence of distributed energy resources (ders) in the form of solar and wind generation and energy storage, is changing the nature of energy markets.
Unidirectional flows of energy from transmission to end customers now coincide with a distributed system where reverse and peer-to-peer flows occur over a network where the size distribution of sellers is changing. In this context, a key challenge is how to manage, integrate, and monetize an increasingly distributed mix of energy supply and storage.

Increasingly, machine learning and artificial intelligence are being integrated with the
trading systems, facilitating the execution of agent-based transactions within a more
distributed network. At a recent financial technology conference at Michigan Law School,
it was estimated that computers are now generating around 50-70 per cent of trading in equity markets, 60 per cent of futures and more than 50 per cent of treasuries. The jp Morgan analyst Marko Kolanovic estimates that a mere 10 per cent of us equity market trading is actually now conducted by discretionary human traders; the rest is driven by various rules-based automatic investment systems, such as computerised high-speed trading programs.

Decentralised systems, based upon blockchain technology combined with autonomous agents, have the capacity to manage the flow of information emanating from the so called Internet of Things (iot) which operate and monitor systems that control electricity consumption. Platform markets such as Uber operate a model of exchange whereby the supply side is decentralised (by allowing any driver the opportunity to drive) with the platform providing a matching role. However, although the potential to buy services directly from individual providers gives the impression of a distributed peer-to-peer (p2p) network, as Tasca and Ulieru (2017) emphasise, “Uber runs on a ’smart’ phone via a quite ’dumb’ application which links into a centralized platform”.


In this regard blockchain as a technology has the capability to facilitate an alternative form of decentralised exchange, threatening the monopoly position of platform intermediaries in sectors such as transportation, e-commerce and financial services.

As McLeay et al (2014) point out, trust protocols such as banking systems and credit rating agencies that provide the basis for exchange between parties, face the emergence of new measures of trustworthiness. These measures, representing the codification of reputation and trustworthiness, provide a critical prerequisite for distributed interactions between buyers and sellers by allowing potential customers to rank supplier performance. In this respect a network of buyers and sellers has the potential to self regulate, circumventing the need for a centralised platform. In its original form, blockchain technology facilitates monetary exchange, as in the case of Bitcoin. However, as noted by Perez (2003), the digital ledger, the so-called blockchain, is expanding beyond the creation of efficiencies in financial intermediation to a technology that can facilitate different forms of exchange through smart contracts. Potential applications extend to a large number of areas, including the management of consumption data originating from the micro devices, the trading of surplus renewable energy, and programmable transactions in the form of smart contracts. As noted by Deloitte (2016), blockchains also promote vertical disintegration by changing the costs of supply chain management, streamlining processes spread across multiple parties and databases on a single shared ledger.

As emphasised in a report by the UK Government Chief Scientific adviser,2 new technologies come with a new vernacular, which in some cases can create problems for policy makers and those charged with making decisions around adoption. In the above quote by Sally Davies blockchain is being used generically to denote the related family of technologies and solutions. In contrast the use of blockchain as the indefinite article refers to the distributed ledger technology, where blocks of transaction made in bitcoin (or another cryptocurrency) are recorded chronologically.

In this paper we examine the evolution and design of marketplaces for digital economies. To do this we first need to understand the economic and technological antecedents of blockchain, and ultimately the emergence of decentralised marketplaces. In examining how the forces of scale and scope economies and the relative costs of transacting within firms and markets have facilitated the emergence of decentralised and agent-based marketplaces, we make use of a number of core economic principles, such as the economics of transaction costs, matching, the principal-agent problem, bounded rationality, information asymmetry and trust relations. We also consider the technological antecedents of blockchain technology, and in particular the fundamental distinction between a distributed database and ledger.

In Section 2 we consider blockchain technology in the broader context of the obstacles to competition in online marketplaces. In Section 3 we introduce a simple economic model which we use to understand the dynamics of firm boundaries, and the organisation of economic activity more generally. Section 4 provides a brief examination of the ownership and control dichotomy in a decentralised digital economy. In Section 5 we examine a number of mechanisms which facilitate the operation of decentralised markets including the distributed ledger and trust and reputation systems. In Section 6 we examine a number of what we refer to as “emerging markets” which exploit the availability of cheap storage and computer processing to deliver information and data. Against this backdrop, in Section 7 we take a brief look at the FireFly platform, which provides a set of tools for the dynamic creation of intelligent marketplaces for a future agent-based economy. Section 8 concludes.


2 Online Markets and The Internet

Although the Internet has brought significant benefits by improving the efficiency of markets through the increase and speed of information flows, we have simultaneously observed the emergence of a new type of monopoly provider based upon the provision of internet-based services. In examining The Market Failures of Big Tech, Martin Sandbu highlights a number of obstacles to competition inherent in internet-based businesses, summarised below.

First and foremost, as the number of customers using platform applications and services increase, network externalities accrue creating economies of scale. The efficiency of the platform increases, as do entry costs for competitors. Second, independent of the scale effects that derive from network externalities, internet companies such as Amazon enjoy cost advantages due to the scale of operation. In addition, companies which assume a pure intermediation role by matching buyers and sellers, will accrue significant savings in unit cost following an increase in scale. Third, the fact that internet companies collect huge amounts of information on their customers, allows them to offer an increasing array of products and services. The resulting economies of scope, with average total costs falling with the increase in variety, provide further advantages over competitors. As Scott Leland notes

More than anyone, the Alphabet ceo Larry Page understands that if you first “organize the world’s information,” then second track and analyze most all of most everyone’s interaction with that information, one can then understand global supply and demand better than any mortal ….

Below we first examine the evolution of firms, markets, and platforms in terms of a changing dynamic between the returns to scale and scope, and the potential change in the costs of transactions, and ultimately market structure. In Section 5 we contrast the design of the Internet and blockchain protocols, and for the latter highlight the role of open source software, a distributed ledger, and a token mechanism, in providing alternate business models and the potential for a different market structure.


3 Firms, Markets and Platforms

“A firm has a role to play in the economic system if transactions can be

organised within the firm at less cost than if the same transaction were carried

out through the market. The limit to the size of the firm is reached when

organising additional transactions within the firm exceed the costs of carrying

out the same transactions through the market”.

Ronald Coase, The Theory of the Firm (1937).

The arrival and subsequent growth of the Internet has threatened to make one of the

principal tenets of neoclassical economics, perfect information, where consumers and producers are assumed to have full information on prices, quality and production methods,

a reality. Relative to off-line markets, search costs on the Internet might be expected

to be lower and online consumers to be more easily informed about prices. However, as

Don Tapscott,7 an author on the impact of digital technology on business and society,

notes that “With today’s internet of information you can’t store, move, transact value

without a powerful intermediary. And that’s what blockchains solve.” Although information technology has also lead to a fall in transaction costs for both buyers and sellers,

blockchain technology provides the potential to reduce the returns to scale and scope

through the cost reduction potential of decentralised consensus mechanisms, allowing

smaller entities to transact in the market.

In this section we utilise a simple economic model to understand how the tension between

centralised and distributed marketplaces gives rise to alternative forms of economic organization. We begin in Section 3.1 by focusing on the pivotal role of transaction costs

in shaping the structure of markets and ultimately the boundary of a firm. In Section

3.2 we briefly examine the economic model underlying platform markets. In Section 3.3

we examine the evolution of decentralised markets, an alternative form of organising

economic activity which represents a fundamental break from the classic Coasian firmmarket dichotomy. Finally in Section 3.4 we examine a number of key factors which

drive the economics behind peer production, including the impact of new suppliers on

incumbents, and the choice of pricing models.

At the outset it is important to be clear as to what we understand by the terms decentralised (dc) and distributed (d). In the blockchain literature these terms are often used

interchangeably. Although these terms are also context specific, in this paper we take

the position that d does not imply dc, as, for example, in the case of a distributed but

not decentralised database. However, dc does imply d in that a decentralised database

must be distributed.8

3.1 Firms and Markets

Throughout much of the 20th century the combination of the fall in both transaction

and communication costs, together with attendant economies of scale and scope, has

resulted in an increase in the division of labour within and beyond national boundaries,

with a significant effect on the size distribution of firms. Prior to the emergence of the

firm as a means to organise economic activity, goods and services were traded directly.

See Blockchain Revolution: How the Technology Behind Bitcoin Is Changing Money, Business, and

the World.

In this respect the two terms are not completely interchangeable.

Without middlemen; trade took place directly between individuals, with production on

a relatively small scale. With changes in the scale of production, facilitated by new

technology and changes in geographic distribution of the population, the relative costs

of organising economic activity through firms fell.

Here we introduce a simple economic model which relates the organisation of economic

activity and the structure of marketplaces according to the costs of transactions. Neoclassical economic theory at the time of Coase was based on an implicit assumption of

zero transaction costs. In an essay on The Nature of the Firm (1937), Coase highlighted

the role of the costs of transacting in markets, as the principal factor mediating economic

activity through the firm, as opposed to a multitude of independent traders who contract

in markets. In brief Coase viewed the market and the firm as “alternative methods of

coordinating production” (1937, p. 388).

A key component of transaction costs arise from the costs of coordinating buyers, sellers,

suppliers and labour. For example, in considering the size of firms, one of the most

widely cited stories on changes in firm boundaries is the case of Fisher Body (fb) and

General Motors (gm). In the first instance fb produced bodies for gm, with both

firms operating as separate firms, linked by contractual arrangements. During the 1920s

gm’s demand increased substantially, and fb refused to accommodate the increase in

supply. Subsequently gm bought out fb, creating an example of a vertical integrated

firm, effectively internalising trade. Arrow (1969) noted that “the existence of vertical

integration may suggest that the costs of operating competitive markets are not zero, as

is usually assumed in our theoretical analysis” (1969, p. 48).

In this context one perspective on the existence of a firm, and the one expounded by

Coase, is that it serves as a device for creating long-term contracts between labour,

suppliers and debt providers, when short-term contracts are too costly or time consuming.

However, Coase also pointed to the limits in the returns to scale and scope, and as

such the benefits of vertical integration, given that internal markets within firms create

transaction costs of their own.

Following in the footsteps of Coase, Williamson (2009) in his Nobel Prize lecture, highlights the key motivating question driving his work on transaction economics: “What

efficiency factors determine when a firm produces a good or service to its own needs

rather than outsource?” Williamson,9

together with a number of researchers including

Baumol (1959), sought to clarify the specific nature of transactions costs, introducing

the behavioral elements of principal-agent conflicts as additional determinants of the

boundary between the market the firm.

Other notable contributions in this area include Gibbons (2005) who develops an integrated framework for examining the contracting theories of Coase and Williamson

and vertical integration. Based upon the notion that the theoretical underpinnings of

transaction cost economics represent a partial equilibrium view of the world, Bresnahan

and Levin (2012) extend the analysis beyond contracting theories to include the role of

economies of scale and scope, alongside network effects.

9 See Williamson (1975, 2002).

3.2 Market Places as Platforms

A predominant feature of online commerce is the presence of two-sided markets or platforms, where platform operators derive revenue from the role of an intermediary. Martens

(2016) notes that platform markets represent a new “generation of online business models where firms organise markets rather than behave like a vertically integrated firm”.

Platform markets exploit information technology to provide a coordination mechanism,

facilitating interactions between buyers and sellers. By using various market design

mechanism platforms allow buyers to locate and trade with trustworthy sellers. Einav

et al (2016) provide an excellent review of the academic research in this area, and also

examine the design of Internet markets along with the economics of peer production.

As Martens (2016) notes, the emergence of online platforms has coincided with the dramatic decline in information costs, changing the balance between the benefits of internal

(firm-based) and external markets, enabling a volume of exchange that was not feasible

in online markets. In this sense information technology shifts the boundary between the

firm and the market.

Rochet and Tirole (2004) define two-sided markets as “markets in which one or several

platforms enable interactions between end-users (i.e. buyers and sellers, advertisers),

and try to get the two sides on board by appropriately charging each side”. In order

to capture network externalities, platform owners face the so-called ”chicken and egg”

problem, where the value to each side is dependent on the use of the other side. To

overcome this problem platforms face the choice of a pricing structure which reflects the

relative price sensitivities of the groups at both sides of the platform. One example,

Ramsey Pricing, sets individual prices above marginal cost in accordance with each

service’s price elasticity of demand. Services with more inelastic (elastic) demand face

prices with a higher (lower) mark-up above marginal cost, effectively cross-subsidising

the other side.10

Business models vary according to the relative strength of each side. In reviewing the

form of competition in two-sided markets, Rochet and Tirole (2003) provide a number

of examples of pricing strategies. In the case of video game platforms operated by Sony,

developers are charged to access the platform in the form of fixed fees for development

kits, and also per-unit royalties on games, with the gamers side of the platform treated

as a loss leader. Other operators derive revenue from charging for the service (Uber), or

subjecting the consumer to advertisements (Facebook or Google).

Pollitt and Richter (2016) consider the choice of pricing structure for platform markets

in the context of the demand and supply of data. The principal question is whether on

one side of the platform the supplier of data (i.e. smart-meter) should pay to access

the platform, or be compensated given a high level of demand on the other side. The

authors utilise a stated preference methodology to examine the demand for electricity

service contracts in a smart grid context. In the presence of network externalities and

where the marginal value of a connection for current and future customers exceeds the

marginal cost of connecting an additional customer to the platform, the service provider

can compensate consumers and still generate positive profits. A main finding is that in

order to participate in automated demand response programs, customers are likely to

require significant compensation from the retailer to share their usage data.


3.3 Decentralised Markets

A natural progression of this form of market evolution is the removal of the platform

or in other words peer-to-peer transactions without an intermediary. We refer to this

type of exchange as decentralised in the sense that both supply and demand is geographically dispersed, with trade occurring outside of any centralised physical or online

marketplace.11 A notable example is openbazaar, a peer-to-peer commerce network

using bitcoin which in essence is a decentralized version of eBay. In the absence of intermediation there are no fees and no restrictions on what goods can be listed and sold.

Peer-to-peer transactions can also be enabled in distributed energy markets, with as

an example, small-scale transactions between individual producers of solar panels and

consumers. Power Ledger has built a decentralised energy trading market that allows

prospers to export energy to their peers in a trustless environment.

In this context blockchain technology can be viewed as the enabling technology by providing a platform for disintermediation. As we discuss further in Section 5, the mechanisms

underpinning market design in peer-to-peer decentralised market places are centred upon

a distributed ledger in conjunction with a consensus mechanism, providing a means for

authenticating the trustworthiness of buyers and sellers, and thereby facilitate exchange

in a permissionless network. This then provides the preconditions for an increase in

competition and less friction in transactions.

It is now well recognised that there exists a market for data that is created by individual

behaviour. Blockchain technology provides the mechanisms to create a market for individual consumers to sell data for monetary rewards, avoiding the high transaction cost of

fiat currency, given the existence of micropayments. Examples include, traffic congestion

data, electricity consumption on smart meters, and health. In many instances this type

of data is extracted from user appliances (which in some instances can occur without the

knowledge of the consumers) and aggregated through central points.

As emphasised by Coase, peer-to-peer marketplaces come at a cost, in that dependent on

the design of the marketplace, it is possible that the total cost of decentralisation exceeds

that of a comparable centralised marketplace. We expand on this point in Section 5,

where we examine the technology of the distributed ledger, along with a number of

scalability issues which have plagued the adoption of Bitcoin, and in particular the

problem of both settlement times and transaction costs.

3.4 Peer Production

In this section we briefly outline a number of features of what has been referred to as the

economics of peer production. Benkler (2002) argues that peer production represents a

“third mode of production in the digitally networked environment”, and best understood

as a category of economic organization that represents a fundamental break from the

traditional Coasian firm-marketplace dichotomy.

Einav et al (2016) develop a stylized model of peer production with two types of producers. Type A sellers are characterised by relatively large firms who operate in a market

with up-front investment and employ labour on a full-time basis. Type B sellers are

characterised by being smaller and as a result more able to respond to price signals.

11It is noteworthy that Einav et al (2016) use a different vernacular referring to eBay, Uber, and Airbnb

as peer-to-peer markets. However, the key characteristic is the focus on markets that are centralised in

that peer-to-peer trade is mediated by a central authority.

Both types of sellers face a cost of advertising, f, in order to become visible to sellers.

An entering seller can choose to operate off-platform and pay the direct advertising cost, or join the platform and pay a combination of fixed of fixed and variable fees. The

model assumes that buyers can purchase readily on or off the platform, such that sellers

can expect the same market price π independent of whether they join the platform.

One interesting output of their model is a stylised representation of the rise of peerto-peer markets and specifically the impact on prices and market structure. A number

of interesting questions can be addressed with this type of model. One particularly

important consideration for both incumbents and new entrants is the impact of the

structure of platform fees. For example, in the simple case where the platform charges

a fixed fee to all sellers, the platform will attract all (none) of the sellers dependent

on whether the fee exceeds (is less than) the direct advertising cost f. We believe that

future research might use this type of model as a basis for addressing similar questions for

decentralised peer-to-peer markets, isolating the role of alternative consensus mechanisms

and pricing using a (potentially) volatile cryptocurrency.

4 Ownership and Control in a Digital Economy

In this section we consider the fundamental dichotomy between ownership and control in

the context of decentralised markets, and examine how blockchain technology provides

the potential to change the form of the traditional principal-agent relationship. In addition, we examine the case for combining the digital technology of the distributed ledger

and smart contracts with a native token, or currency. We also examine the nature and

form of cryptocurrencies, and it what sense these currencies can be considered the digital

equivalent of money.

4.1 Ownership and Control

A fundamental prerequisite underlying the decision to enter a market to consume, invest, or trade, is the presence of safe institutions and attendant good governance. This

applies equally to sovereign and digital economies, although there are a number of notable differences. Fiat currencies are traditionally issued by sovereign governments, who

alongside central banks determine rules governing the supply of the currency. In these

economies banks hold the centralised digital record of transactions and are trusted to

ensure its validity. In a digital economy the distributed ledger contains the record of all


One of the central tenets in the theory of corporate governance is the potential conflict that emanates from the separation of ownership (shareholder/principal) and control

(manager/agent). The agency relationship is generally defined as a contract between

principal and agent whereby the agent acts on the principal’s behalf (See Jensen and

Meckling (1976)). However, as Brennan (1995) and Kaal (2017) have underlined, the

existence of both bounded rationality and information asymmetries means that it is not

possible for principals to contract for every possible action or inaction of the agent.

The technology of the distributed ledger, and in particular the presence of a consensus

mechanism and information that is consistent across the entire network, goes someway to

circumvent the conflicts which arise through the separation of ownership and control. As

noted by the Boston Consulting Group (2016), there is no need for a Bitcoin “account”,

with a separation between the account holder and the location of funds. The lack of


separation between individual ownership and control of a personal account is apparent

since Bitcoins are held in a personal wallet.

In the case of digital currencies, authenticity is provided by the digital ledger, which

is maintained by a network of computers. Governance is distributed throughout the

network as opposed to being allocated to special institutions. In the Bitcoin network,

governance is distributed in the sense of being controlled by special users known as ’miners’. Miners collect blocks of transactions and using cryptographic techniques compete

to determine whether the payer is the owner of the currency in question. As a reward

for successfully verifying the authenticity of a block of transactions, miners receive an

allocation of tokens and any transaction fees offered.

As security of the network increases, there are network effects as the demand for the token

increases, which then attracts more miners to provide additional verification. Miners

receive rewards in the form of new tokens, and dependent upon monetary rules, this can

create monetary inflation. As the currency appreciates in value, prices and transaction

fees within the token economy can be adjusted by pegging transaction fees to a fiat


4.2 Cryptocurrencies

The term cryptocurrency derives from the fact that the tokens are encrypted using cryptography techniques that secure and verify transactions, whilst preserving the anonymity

of users holdings of digital currency. However, the very notion of a cryptocurrency is, in

many cases a misnomer, given that the role of money is fulfilled to a limited extent. As

an example, Paul Krugman (2011) has noted that, based upon recent market conditions,

the short-run inflation of the dollar/pound value of cryptocurrency, can make these currencies a good store of value for investors. However, in the case of Bitcoin we have seen

the emergence of a scenario that blockchain decentralised marketplaces were designed to

overcome, namely high prices, driven in part by the design of underlying protocol, to the

detriment of use value.

Here and throughout this paper, we stress the importance of clarity in terms of the meaning of terms that are often used interchangeably. In this context, there is often confusion

as to the vernacular surrounding the terms digital token, currency, and cryptocurrency.

In the most broad sense, tokens are simply units of currency enabling transactions within

a particular environment, and may be exchanged for legal tender or other cryptocurrencies such as Bitcoin or Ethereum. However, in contrast to fiat currencies where the

value of the currency is generally fixed, cryptocurrency fixes the total quantity. This is

analogous to the gold standard, where the money supply is fixed rather than subject to

increase via a central monetary authority. Bitcoin, for example, has a clear monetary

policy: the supply of coins is fixed at 21 million. The rate of creation decreases by half

every four years.

In addition, digital currencies are distinct in a number of other notable ways. First,

whereas bank deposits represent a liability for the bank and an asset for the account

holder, digital currencies do not represent a claim on anybody. Second,12 although

digital currencies may be viewed as a commodity, their intangible nature make them

more akin to digital commodities.

As a point of reference, McLeay et al (2104) points to the following role of money in

See McLeay et al (2014).

society: a store of value with which to transfer purchasing power (the ability to buy

goods and services) from today to some future date; a medium of exchange with which

to make payments; and a unit of account with which to measure the value (i.e. price) of

any particular item that is for sale.

To the extent that digital currencies serve as money, the allocation of tokens to users that

contribute computing resources towards the verification of transactions on the network

is similar to seigniorage, in the sense that tokens have a net value given the cost of its


Kalla (2017) emphasises that it is important to recognise that cryptocurrencies are distinct, making reference to the following classification:

  • Cryptocurrencies built using Bitcoin’s open-sourced protocol are generally referred

to as Altcoins. As an example, the Litecoin token changed a number of parameters

including the mining algorithm and the total supply of coins.

  • Cryptocurrencies designed for a different purpose using a different blockchain. Examples include Ethereum and nxt, a proof-of-stake coin.
  • Cryptocurrencies designed for a specific application, and required to use that application. In this sense the generic term cryptocurrency is not appropriate here,

with digital (or use) token being a more appropriate moniker.

5 Mechanism Design for Agent-Based Decentralised Economies

.. Because computers are now cheap and ubiquitous, we can design “smart

markets” that combine the inputs of users in complex ways. Kidney exchange

is an example of a smart market. By running through every possible combination of donors and patients, it can arrange the highest possible number of


Al Roth (2007), Nobel Laureate.

In Section 3 we examined the role of transaction costs, alongside scale and scope economies,

in determining the organisation of economic activity, and specifically defining the boundary between the firm and the market. In this section we examine a number of tools that

are used to enable the operation of decentralised markets, and specifically how these tools

impact transaction costs. In this context it is instructive to view the Internet protcols,

aligned with web browsers and search engines, as a mechanism which provides efficient

and low costs matching between buyers and sellers. The design of market mechanisms for

online markets presents further opportunities to reduce the costs of entry and transaction costs, and therefore increase competition. However, there are very real challenges if

the ’automation of user-market interaction’ is to deliver the potential of truly automated

transactions over a wide range of sectors.14

In agent-based marketplaces the agent is a piece of software which in general terms works

on behalf of a user by implementing a series of instructions. Here we make the distinction

between simple agents and those which we will refer to as intelligent. Simple agents are

A key difference is that a payment is made to miners in return for the verification of transactions,

whereas seigniorage accrues to the government.

See Vulkan and Priest (2013).

Autonomous conditional on a set of instructions which exist in the form of a contract.

We can think of such a contract as complete in the sense that for the task at hand, the

instructions encompass the set of possible outcomes. For example, eBay’s bidding agent

autonomously bids on behalf of a user, conditional on an upper limit and pre-specified

increments. In contrast a number of protocols such as that being developed by FIreFly

(see Section 7), have combined machine learning with the distributed ledger, providing

intelligent agents that have the capabilities to learn from their environment and do far

more than execute simple instructions.

Although these tools physically exist in the form of computational features such as

Internet and blockchain protocols, distributed storage and databases, the rationale for

the underlying design rests firmly with a number of principles provided by a sub-field

of microeconomics, mechanism design, used by analysts and policymakers to design

institutions and markets where agents are incentivised to behave in a particular way.

We highlight the role of mechanisms that can be deployed to solve a number of key

problems in the design of markets which operate within permissionless networks. These

mechanisms are used in the determination of identity, matching buyers and sellers, the

authentication of transactions without a trusted intermediary, and the design of digital

tokens. For example, in the area of auction design and regulation of monopoly providers,

mechanisms can be put in place so that agents are incentivised to reveal true values in

a world of asymmetric information. In addition, in the case of online marketplaces, a

key design mechanism concerns how to build enough trust so that strangers can trade


Blockchain technology has the potential to reduce transaction costs, execution risk, and

information asymmetry, and in doing so increase the speed of transaction. The central

feature of mechanism design based upon blockchain technology is the use of decentralised

consensus which provides the authenticity of transactions within a trustless network.

That said, it is important to highlight significant variation in the form of blockchain

technology, manifest as different protocols which impact the way in which exchange is

mediated. For example, the limitations of the proof-of-work protocol that supports the

use of Bitcoin as a means of payment for goods and services, have been well documented.

Most recently Goldman Sachs in a note to clients15 underlined the potential of digital

currencies that leverage blockchain technology, including “ease of execution, lower transaction costs, reduction of corruption … and safety of ownership.” At the same time, the

authors noted that “Bitcoin does not provide any of these key advantages,” pointing to

the relatively long settlement times.

Below we examine the anatomy of the so-called “decentralised stack”, which represents

the totality of computing functions which constitute blockchain technology. This comprises storage, processing and communication protocols, combined with a distributed

ledger, a payment system, a trust and a token mechanism, machine learning and one or

more decentralised applications. We also contrast the design and associated incentive

mechanisms that underpin the decentralised stack of blockchain protocols and the stack

of the Internet protocols.

In Section 5.1 we look at the precursor to the distributed ledger, the distributed database.

Section 5.2 considers the role of payment systems within the decentralised stack, and

in particular how micro transactions require a different protocol than used in standard

blockchains. Section 5.3 examines the mechanisms by which trade is facilitated through

(Un) steady as She Goes. Outlook Report, January (2018).

Trust and reputation, and in Section 5.4 we examine the role of machine learning as a tool

to add value to a history of transactions recorded on the distributed ledger. Section 5.5

looks at decentralised applications, and Section 5.6 examines the role of digital tokens.

The Anatomy of The Decentralised Stack

In Figure 1, reproduced from Pon (2015), we present a schematic representation of the

layers of computing technology that comprise what some have referred to as the “decentralised stack”. The base layer comprises the computing functions of communication,

processing and storage on which applications can be built. In a number of protocols, decentralised storage has three components: a ledger which keeps track of all transactions

on the network, a file system, and a decentralised database. In this instance we have

represented decentralised processing using the Ethereum protocol.16

Figure 1: A Decentralised Stack: Computing, Platforms and Applications

Below we first list and then consider a number these components.

  • Decentralised Storage the base layer of the network comprised of:

a distributed ledger which provides consensus on transactions;

a file system;

a distributed database which stores information other than pure;

transaction data.

  • Trust and Reputation to provide assurance to users of the network.
  • Payment Systems which facilitate exchange of value between buyers and sellers.
  • Smart contracts to automate and streamline business processes.
  • Machine Learning to drive better decisions and detect network anomalies.
  • Decentralised Applications such as software which provides access to goods or


In Section 7 we will present the FireFly protocol and consider the advantages.

  • Digital Tokens providing digital representation of value.

5.1 The Distributed Database and The Distributed Ledger

In Section 3 we considered the evolution of decentralised marketplaces from the perspective of how changes in technology and the costs of transactions, determine the organisation of economic exchange and the structure of marketplaces. In a similar fashion it is

instructive to consider how the distributed ledger itself has evolved from a distributed


The Distributed Database

Why can’t companies wanting to share business logic and data just install a

distributed database? What is the essential difference between a distributed

database and a distributed ledger?

  1. Brown (2016).

The distributed database emerged, in part, with the development of the internet, and

the attendant need to process large quantities of structured and unstructured data which

could scale across networks. Databases are distributed when the storage devices are

spread across multiple physical locations or nodes. However, as emphasised in Section 3,

the physical distribution of a database (either in parts or in its entirety) over locations

within a network, does not say anything about the form of control. For example, given

that a database management system (ddbms) can integrate and manage distributed

data, we observe that although a distributed database is a necessary condition for the

decentralised control mechanism, distributed databases can coexist with a centralised

control mechanism.

Brown (2016) makes a distinction between a distributed database and a distributed

ledger. In the case of the former, the network is comprised of a group of computers

which is invariably under the control of a single organisation where “each node in the

system trusts the data that it receives from its peers, and nodes are trusted to look after

the data they have received from their peers.” As a result, in a world where trust and

governance can be reliably delegated to a central operator, the gains from blockchain

technology do not derive from the decentralised consensus mechanism.

Gideon (2015) in considering the extent to which blockchain can be considered both an

economic and a computer science innovation provides an excellent comparison of the

technology of the distributed database aligned with multiversion concurrency control

(mvcc) and the distributed ledger. To do this Gideon utilises the following example,

based upon a

… list of bank accounts, in which each row contains an account number along with

the balance of that account. Let’s say your account starts the day with a balance of

£900. Today an automatic mortgage payment of £750 is scheduled and you also

need to withdraw £400 from an atm. Unfortunately you do not have an overdraft

facility so one of these operations is set up to fail.

The processes for mortgage payments and atm withdrawals run on separate systems, both of which access this single account database. Let’s say that each process

works by reading your account’s balance, checking it is sufficient for the operation,

initiating that operation, verifying the operation completes, calculating the new

balance and then finally writing it into the database.

As long as the mortgage payment and atm withdrawal don’t overlap, this logic

will work fine. The first operation will execute successfully, and the second will

abort because your account has insufficient funds. Depending on the order, you’ll

get an angry phone call from the bank or a rude message on the atm screen.

Given that mvcc technology facilitates parallel transaction within a short time interval,

whilst preserving the integrity of the database, Gideon (2015) further notes that

If two transactions attempt to delete the same row version, then only one of these

transactions will ultimately be accepted. Multiversion concurrency control acts as

a unified mechanism to detect and prevent these conflicts within a database.

This example demonstrates that blockchain technology solves the same problem using

an alternative mechanism for synchronising distributed databases. The question that

naturally follows is how is the distributed ledger technology different from a distributed

database? We consider this question below.

The Distributed Ledger

Cheap computing in the form of processing power and storage is a fundamental precondition for a digital economy. The distributed ledger technology exploits the availability

of cheap storage, distributing copies of the complete record of transactions across the

network, thereby creating the conditions for a decentralised consensus mechanism.

As demonstrated above, both blockchain and mvcc enable concurrent transactions to

access a single database without conflict or a degradation in process time. However,

as Raval (2015) points out, blockchains have a number of additional important features

which are not available in today’s distributed databases. The most important of these in

the context of permissionless networks, is that the mechanism of the distributed ledger

solves a key problem with distributed databases in that it is not possible to facilitate

transactions between parties within a network that do not know each other. In the case

of blockchain technology, the consensus mechanism is itself distributed throughout the

network, operated by multiple computers or “nodes”. This technology has the potential

to be less costly than traditional databases because the distributed system regulates

itself and automates much of the overhead generally required for verification of various

transactions. The reduction in risk inherent in blockchain systems provides the potential

for financial institutions to settle transactions faster and with less scrutiny.

The technology of the distributed ledger may be considered an application in its own

right, and can be deployed without digital tokens. As an example, ibm has developed

a blockchain platform which among other uses is designed to reduce the cost of making

global payments for businesses and consumers.17 Moreover, companies that currently

own and operate centralised platforms, such as MasterCard, are looking into the use

of blockchain as a means to increase the speed and reduce the cost of transactions for

existing business models.



5.2 Scalability and Payment Systems

When individuals send payments in the form of cryptocurrency in the absence of a central

banking authority, a primary challenge is the design of a (payment) system where transactions can be verified and automatically updated without transactions being altered.

Stocker (2018) refers to the distributed ledger mechanism which underpins Bitcoin and

Ethereum, as a discrete blockchain given the way in which blocks of transactions that

are collected en masse, authenticated by a ’miner’ and committed to an existing chain

of transactions.

There are, however, a number of known problems with particular instances of the distributed ledger technology. One immediate problem with this type of payment system is

scalability in the presence of micro, time-dependent transactions. In the case of public

blockchain networks which rely on proof-of-work for mining (i.e. Bitcoin and Ethereum)

the payment throughput handling capacity is limited given the way in which transactions

are written to the ledger. This creates a constraint on the viability of micro-transactions

without paying large fees. As a reference point, Bitcoin can handle around four transactions per second, Ethereum 15 to 25, Ripple around 1,000, and Visa around 10,000.

The issue is particularly important in the context of markets that are emerging based

on the demand and supply of data from the so-called iot. Most current iot applications

connect devices with a common owner, so they only need to exchange information or instructions. When devices have different owners, then in the absence of a shared intermediary, transactions involving small monetary value may not be economically worthwhile.

This is particularly true on a large scale as the transaction delay becomes consequent

and thus not suitable for most applications. iota, formed in 2015, with a primary focus

on enabling automated iot applications, was designed to transfer crypto-tokens at zero

transaction fee.

5.2.1 Micro Payment Systems

In a world where machines are continuously transacting physical and digital values in

a peer-to-peer (p2p) network, the use of blockchain protocols which rely on discrete

payment methods (i.e. proof-of-work) are not an economically viable method to process

micro-transactions. In this instance the scalability requirement may be realised by using

a permissioned blockchain network, in conjunction with consensus algorithms such as

Proof-of-Authority or Proof-of-Stake. So-called payment channels provide high frequency

scalable micro payments with zero transaction fees. This type of payment system does

not require block confirmations on the main chain and as a result can be used for deviceto-device transaction or in any scenario where instant payments are needed. In contrast, a

smart contract discrete payment method such as Bitcoin / public Ethereum blockchains,

is not yet scalable and requires payment of transaction fees to public chain networks

which makes micro-payments impractical.

5.3 Trust and Reputation

Matching two strangers with each other and facilitating a transaction to completion is very similar to a blockchain facilitating peer-to-peer interaction between two (or more) parties that do not know each other.

William Mougayar.

A prerequisite for exchange in online markets between anonymous agents is the existence

of trust and reputation mechanisms. However, trust is not a problem that applies only

to transactions between a buyer and seller. It is central to the authenticity of data,

the terms of an agreement, and the notion of identity. In some markets, such as used

cars and furniture, trust is created by inspection or by external regulations. Although

in-store transactions allows a buyer to physically inspect the product before buying, this

is not possible with electronic commerce where in some cases the identity of the seller

may not be able to be verified.

As Nosko and Tadelis (2015) underline, trust and reputation mechanisms mitigate the

inefficiencies in markets with asymmetric information. In his seminal article The Market

for Lemons, Akerlof (1970) highlighted the role of information shared between buyers and

sellers in determining the efficient operation of markets. For example, if we do not know

the true quality of the car as well as the selling owner, we are exposed to the possibility

of exploitation. In return, the lack of full information is reflected in the price individuals

are willing to pay. As information asymmetry increases, the trade that does occur will

be selective, mostly for poor quality cars, or ’lemons’, where the cost of judgement error

is low.

The economics literature has formalised this problem in identifying two sources of uncertainty that hinder markets from operating efficiently. First, adverse selection occurs

when uncertainty is linked to hidden information that determines the quality of the good

or a service. For example, products for sale in online marketplaces such as Ebay may be

misrepresented. Second, moral hazard relates to the unknown actions of the seller that

determine the quality of the good or service. For example, the shipment of a product

without the necessary protection to ensure that it arrives undamaged.

In some markets trust has long since been a problem. Consider the example of the energy

sector. Speaking to Utility Week in August 2017, JoJo Hubbard, the chief operating

officer of the blockchain company Electron, cited figures published by the consumer

website, that 1.3 million energy customers have been overcharged by a

total of £102 million, the equivalent of approximately £79 each.

Although the success of online marketplaces such as eBay, Uber and AirBnB, can be

attributed to the efficiency in matching buyers and sellers, and being able to facilitate

trade between strangers, a number of studies have pointed to problems with the design

of trust and reputation mechanisms. Nosko and Tadelis (2015) highlight two issues.

First, a poor quality transaction may cause a buyer to update not only his prior beliefs

on a particular seller, but on all sellers, thereby generating a reputational externality

on the platform as a whole. Second, using eBay records the authors demonstrate that

recorded estimates of reputation measures are upward biased, originating from buyers

not providing feedback in response to a poor transaction.

5.3.1 Trust and Reputation in Decentralised Markets

In a study written prior to the emergence of distributed ledger technology, Josang et al

(2006) review the state of the art in trust and reputation systems for online services. At

the time of publication it was felt that it was not possible to prevent reputation systems

from being manipulated. This conclusion is consistent with the general observation

that although the protocols underlying distributed communications architecture are well

established, the ability to manage trust without a centralised authority is only now

possible. Through a combination of a consensus mechanism, verification of identity

and measures of trust, blockchain provides the basis for a reputation system which can

determine the trustworthiness of users.

One obvious question when considering decentralised peer-to-markets is what are the

additional issues facing the design of trust and reputation mechanisms? One immediate

observation is that the design of trust and reputation mechanism in these markets needs

to be multilayered. For example, in the case of eBay, the problem of asymmetric information begins and ends with the buyers and sellers, with the platform assuming the role

of the trusted central authority. In decentralised markets, the role of authenticating and

monitoring transactions lies within the network. Although we have pointed to the benefits of organising economic activity in this way, there is an additional cost in monitoring

the trustworthiness of network nodes.

Just as the mechanism of the distributed ledger can be modified to account for different

types of transaction, trust mechanisms can also be designed according to the granularity

of the transaction. In this context one can consider a tradeoff between the level of assurance that a particular trust mechanism can be provide and the speed of the transaction.

We consider this issue further in Section 7 where we examine the features of the FireFly


5.4 Machine Learning

The realisation that a record of transactions across multiple users and over time can be

used to add significant value in future consumption is well known. Machine learning

techniques such as neural networks, support vector machines, and decision trees19 are

now widely recognised as set of powerful tools that can be employed to both understand

and predict many dimensions of consumer behaviour. Early successes were obtained

using supermarket scanner data (see, for example, Hendel and Nevo (2006)) allowing

analysts to generate individual level demand elasticities, facilitating better targeting of

price promotions. Data from search engines has been used by Choi and Varian (2009)

to predict activity in a number of markets including automobile, retail and trade. More

recently O’Neill and Weeks (2018) construct decision trees using recursive binary splitting

to predict demand response to the introduction of time-varying electricity prices.

As pointed out by McConaghy (2017), the characteristics of the blockchain in terms of

a decentralised ledger and immutability, has direct implications for any analysis that is

conducted. The decentralised structure of the blockchain coupled with the fact that no

single agent controls the data, facilitates data sharing where there exists some benefit.

For example, the existence of shared control facilitated by the distributed ledger makes it

easier to integrate data from different markets. Moreover, distributed ledgers with shared

control result in more data, shared control of training data, and (generally) better models.

The existence of immutability implies that data and models are more trustworthy.

In considering the full set of tools that are deployed in marketplaces for autonomous

economies, it is important to note that although each tool has a set of properties, the total

value added within the system critically depends on how the tools are deployed. Consider

the following two examples. First, the use of either supervised or unsupervised machine

learning methods for prediction depend critically on the availability of large amounts of

data. Second, the fact that computer processing in the form of machine learning tools sit

on top of the distributed ledger means that data, representing, for example, user choices

See Hastie et al (2009).

Across multiple markets, may be seamlessly integrated and processed.

5.5 Decentralized Applications

In this paper we have considered how the combination of distributed data, a decentralised

consensus mechanism, and a digital token, facilitates the operation of decentralised markets. An initial application of this protocol was the transfer of digital currency without

a central intermediary. To truly understand the incentive mechanisms of the blockchain

protocol, it is instructive to contrast this with the principal Internet protocols, ip/tcp,

http, and smtp, respectively the method for transmitting data between computers on

the internet, the method to display web pages in a browser, and the method for sending


Using ip/tcp as a reference point, and as Gajek (2018) points out, we can think of

the blockchain protocol as providing another layer on top of the network layer on which

decentralized applications, or dapps, can be built. This is apparent in Figure 2, reproduced from Gajek. In comparing the blockchain and Internet protocols, Union Square

Ventures (2016), characterise the latter as comprised of ’thin’ protocols and ’fat’ applications; with the reverse relationship holding for blockchain technology – namely ’fat’

protocols and ’thin’ applications. Here fat (thin) alludes to the existence of high (low)

returns to investment.

Figure 2: Extending the Internet Protocol Stack

As the authors emphasise, under the blockchain protocol the distributed nature of user

data coupled with open source software reduces barriers to entry and provides incentives. However, although there is general agreement that open source software, alongside a decentralised consensus mechanism, represents two key components of blockchain

technology, there remains the question as to how to incentivise and monetize both the

development of protocols and an open source decentralised application in permissionless networks.20

5.6 Digital Tokens

In the context of new and emerging decentralised markets, the additional component that

provides incentives for protocol development, which in turn incentivises competition at

the applications layer, is a digital token. Raval (2015) has referred to the digital token

as acting like an additional protocol layer for value transfer. The end game here is the

potential of the combined technologies of open source applications, the distributed ledger,

and a decentralised consensus mechanism coupled with a protocol token, to impact the

organisation of economic activity, and in particular the nature of rent extraction across

economic platforms.

In Figure 1 we observe the role and type of token varies across the levels of the decentralised stack. Soman (2017) enumerates the different types of tokens in monetizing

rewards for contributors. Unlike developers of the previous generation of Internet protocols, developers of blockchain protocols such as Bitcoin and Ethereum are rewarded as

demand for platform services increases, with the native currency (here either Bitcoin or

Ether) appreciating in value, acting as an incentive for the development of new protocols.

Developers of a new application may be rewarded with usage tokens (or appcoins), purchased using cryptocurrency (or fiat) to access digital services on a particular blockchain


In further considering the role of tokens in decentralised markets we pose the following


1) What characteristics of e-commerce based on blockchain technology influence the

decision to introduce platform-specific digital tokens?

2) What factors determine the specific design of digital currencies?

Mougyar21 notes that a major advantage of digital tokens is that “companies can be their

own payment processors without the cumbersome or costly aspects of traditional financial

settlement options.” As we have highlighted in this paper, a characteristic of number

of emerging markets is the granularity of transactions. Micropayments, combined with

off-chain payment channels, can be used to obtain access to a wide range of services and

commodities including apis, media content, bandwidth, computing power, storage, and

electricity. Protocols based on blockchain technology have the potential to enable small

payments which given the size of the transaction, could not be processed by a traditional

financial system. As an example, the Raiden Network22 is an off-chain scaling solution

for any erc20 compatible token, enables near-instant, low-fee and scalable payments.

The question of the design of digital currencies is taken up by Halaburda (2016), who

emphasises three main attributes: whether tokens can be i) bought and/or earnt; ii)

transferred between users; and iii) redeemed for fiat currency. Dependent on the particular business models, control over these design attributes will determine the demand

for platform services and also determine the extent of network effects. Although this

20Raval (2015) provides a detailed review of how the components of blockchain technology, combine

to provide a set of critical prerequisites for profitable decentralised software applications.


particular review focusses upon the variation in the design and use of tokens in centralised platforms, it is nonetheless informative. For example, Facebook’s (fb) Credits

and Amazon coins are testimony to the role of digital tokens in existing marketplaces.

As an example, fb Credits could be bought or earnt, with users unable to redeem tokens

for fiat currency.23 Amazon coins are provided to users who purchase the Kindle, but

expenditure of these coins is restricted to Kindle-specific apps, creating incentives for

developers to produce additional applications.

Given the decision by a large number of companies to provide products and services which

require digital tokens as a means of payment, the comments of Goldman Sachs on the

problems that derive from the plethora of alternative cryptocurrencies are pertinent.2425

Since the success of a particular platform can be measured by the demand for its own

tokens, the value accruing as a result of network externalities is constrained by the large

number of available tokens across fragmented markets. The question of interoperability

between different blockchain protocols naturally then follows. A related question is

why do we observe a proliferation of use tokens for purchasing alternative services and

products, when outside digital economies we use a single currency for all purchases. The

fact that currently there is no dominant internet or email provider, with emails exchanged

between individuals using different email providers, suggests that a world with multiple

blockchain networks will need to interoperate. In future work we plan to consider these

design issues in more detail.

In Section 5.3 we examined the design of trust and reputation mechanisms, noting the

differences in these mechanism across centralised and decentralised platforms. Although

Halaburda (2016) provides a useful framework for assessing the functionality of digital

tokens, there are once again very real differences across these two types of platforms.

Critically, above we have focussed on the design of use tokens as a means of payment

for platform services. In the case of decentralised platforms, currency is also used as an

incentive which underpins the consensus mechanism. In this regard, payment systems

for decentralised platforms need to consider the design features of alternative types of

tokens, and the potential interaction between them.

At this juncture, and at the risk of stating the obvious, we emphasise that in designing mechanisms which can deliver the appropriate incentives for the functioning of

decentralised markets, it is imperative to understand the technology, and in particular

the critical role of the blockchain layer on top of the network protocol. In this regard,

although we currently observe a large number of papers dedicated to the design of mechanisms at various points of the decentralised stack, such as those designed to provide

trust and reveal reputation, there is in general a dearth of studies that consider the

design and interaction of the consensus and token mechanism at the base layer.

23fb Credits were phased out in 2013


25In January 2018 the Financial Times reported that there were 39 digital currencies with market

capitalisations of more than $1bn.


6 Emerging Markets

The Fourth Industrial Revolution will usher in an era of radical automation,

with billions of connected smart machines interacting with humans and with

each other and generating a growing share of global GDP. As the Internet of

Things converges with artificial intelligence, a growing population of smart,

connected devices will eventually operate autonomously, creating new business

models and new value in the process.

Commonwealth Bank of Australia.

In this section we review the characteristics of a number of emerging decentralised marketplaces, alongside the challenges for traditional market mechanisms, in particular the

viability of traditional payment systems based upon a fiat currency.

6.1 Granularisation of Digital Business

There are a number of notable characteristics of decentralised markets, including the

emergence of many more sellers, an increase in the number of transactions, and a change

in the size distribution of the value of transactions. Commensurate with these developments is a change in the scale of operations and what we refer to as an increase in

the granularity of consumption. Used in this context granularity refers to the trend for

consuming products in smaller chunks. As an example, consider the demand and supply of print medium. One of the consequences of the Internet was the creation of new

markets for online media, disrupting the traditional booksellers and newsprint markets.

Over time we have observed further changes with the emergence of new markets that

present content differentiated by size. These markets are able to satisfy variation in

consumer preferences: from a consumer who wishes to subscribe to all available content

of a particular supplier over a fixed-term, to a consumer who wishes to consume on a

pay-per-article basis.

Increasing granularity creates a world with high transaction volumes, and a commensurate emergence of micro-level transactions. Related, markets have become more time

dependent in the sense that in some instances the demand for the product is not separable from the time of purchase. These markets require changes in the design of certain

market mechanisms, including governance, the representation and management of trust,

process optimization, and payment systems. For example, it is likely that trust mechanisms will need to evolve, reflecting the realisation that the required level of assurance

of authenticity of a given transaction is falling with the value of the transaction. In

addition with time-dependent demand, payment and settlement will need to be faster

and intimately tied to the provision of services. In Section 7 we examine the protocol underlying the FireFly platform, which incorporates an innovative solution to this problem

of transaction authentication in markets with micro-payments.

Figure 3, reproduced from Stocker (2018), provides a useful taxonomy, distinguishing

between physical versus digital exchange and the discreteness of the transaction. In considering the design of a technology to facilitate the trade of goods and services in a digital

economy, instances where the transaction is both continuous and time-dependent will require the development of new payment systems based upon a decentralised blockchain



Figure 3: Transaction Types and the Blockchain

6.2 Managing Identity

A fundamental prerequisite for exchange to take place within decentralised markets with

no central authority is the existence of a verifiable identity. By owning and controlling

our identity and associated data, we may choose to be anonymous or reveal certain

truths, without having to transfer the data that proves it to a third party.26 Kim

Cameron, Chief Architect of Identity for Microsoft, observed that “The Internet was

created without an identity layer” or in other words that the Internet’s addressing system

is based on identifying physical endpoints (machines) on a network and as such has no

way to uniquely identify people.

As an example, Sovrin, an open source distributed ledger, that has been created in

conjunction with the Sovrin Token, to provide secure peer-to-peer interaction by both

protecting and revealing identity.2728

6.3 Digital Content

In today’s markets internet content is experiencing an increase in the demand for granularity in terms of the ability to consume small amounts of media content. Although

it is possible to record media consumption by the page or the minute, this demand has

been confronted with a supply constraint, largely driven by the problem of inappropriate

payment systems.

26This capability is based on the use of so-called “zero-knowledge proofs”.




6.4 Decentralised Energy Markets

The emerging landscape of the energy sector is an expanding network of smaller suppliers

and microgrids creating the potential for p2p transactions with supply surplus being

traded on a more micro scale. With a distributed ledger provided by blockchain, a

number of the problems that plague centralised energy networks can be addressed. These

problems include a single point of failure, low speed of transactions, high transaction

costs, billing discrepancies, and a general lack of transparency.

As a example, Power Ledger, an Australia-based startup, is developing a blockchainbased platform that can turn an apartment building into a microgrid based on a shared

system of solar panels and battery storage. In such a system producers are able to

trade energy peer-to-peer with consumers. In the u.k Electron has developed a meter

registration platform which allows users to identify which asset in a system you want to

trade and optimise, what their characteristics are and where they are. One possible use

of the Electron platform is to facilitate the ambition of the Gas and Electricity regulator,

Ofgem, to move to reliable next day switching of energy supplier without relying on a

monopoly service provider (intermediary) that would need to be regulated.

The question as to how blockchain technology can be used to replace centralised platforms with monopoly authorities, is particularly pertinent when considering the design

and management of a new repository for smart meter data, holding customer-level data

on identity, location, and consumption. Currently the licence to manage the data and

communications network is held by an intermediary. Given the use of mechanisms to

reveal identity, and both validate and provide trust in data, blockchain technology provides a decentralised alternative with the potential to enable individual households to

realise value in their own consumption data.

6.5 Machine-to-Machine Markets

Blockchain embeds a wallet into machines. As a consequence machines are

getting their own profit and loss statement and the ability to do transactions

with other (machine) entities in an automated way.

Dr Carson Stocker (2017).

Driven by the continued fall in the cost of computer chips, it is estimated that by 2020

between 20 and 50 billion connected devices, from mobile phones to consumer durable,

will be in use, representing an incremental spend29 on the Internet of Things (iot) of

e250 billion.30As the iot converges with artificial intelligence, smart connected devices

will facilitate exchange between micro devices, mediated by autonomous economic agents


The nature of these markers where billions of internet driven devices exchange currency

and data, will require the development of new data-driven business models and market infrastructure. In Section 5.2 we highlighted the importance of designing payment

systems that are able to scale in the face of a large number of micro, time dependent

transactions. Consider the case where two parties agree on the value of data in a particular data stream. The buyer of the data sends a nano-payment (in the form of tokens)


30Boston Consulting Group (2017).


to compensate the seller, who is able to sell the data at no credit risk.31

The Internet of Things Alliance (iota) has developed a blockchain architecture based on

a Directed Acyclic Graph structure (dag), providing the potential for a high throughput

of micro transactions with almost zero fees.32 However, there remain significant problems

to overcome before this type of exchange can be made profitable.

7 FireFly Network

To date we have examined how the various layers of blockchain technology, aligned

with the tools of mechanism design, can be combined to create a set of tools which

incentivises disintermediation exchange between economic agents. In this section we

examine a specific instance of a blockchain protocol in the form of FireFly.

Built using an extended version of distributed ledger technology, FireFly incorporates

artificial intelligence and machine learning technology, combined with advanced measures

of a trust protocol, to create a platform where autonomous software agents, working on

behalf of their stakeholder (i.e. human owners, organisations etc.) perform tasks such

as delivering data or providing services. Agents are rewarded with a digital currency for

their efforts – the FireFly token. As we explain below, a critical feature of the FireFly

protocol is the way in which the distributed ledger mechanism, machine learning and

trust, are combined to create an intelligence driven agent system.

7.1 Combinatoric Innovation

A consistent theme throughout this paper has been the role of technology in shaping the

relative costs of transacting within firms versus within markets. Blockchain technology,

provides the potential to reduce transaction costs, provide the service or good in a more

timely manner, and in doing so allows new markets to emerge. Existing business models

in physical markets typically depend on multiple layers of decision making, adding to

both the cost, complexity and time to deliver a service. The FireFly protocol is predicated

upon the realisation that, relative to existing digital infrastructures, there are a number

of essential enablers missing.

In Section 5, Mechanism Design for Agent-Based Decentralised Economies, we examined

the key components of blockchain technology. The uniqueness of the FireFly protocol and

a particular source of value, is the way in which the specific components of blockchain

technology are integrated within a unifying framework. In reducing both the costs of

transaction and contract costs, and decreasing delivery time, FireFly has the potential

to serve new and emerging markets, based on granularity and where a key attribute of

a service is the immediacy of the delivery time.

7.1.1 Key Features of the FireFly Decentralised Stack

In Figure 1 we presented a schematic view of a decentralised stack, with ethereum

included as the blockchain protocol. The details of how the FireFly decentralised stack

differs is covered in a technical summary paper.33 In particular this document provides

31This example is taken from


33See Simpson et al (2018).


a detailed review of the three layers of the FireFly architecture: Autonomous Economic

Agents, the Open Economic Framework, and the Open Economic Ledger.

Below we highlight three central features in the design of the FireFly decentralised stack:

payment systems, machine learning and intelligence, and a stochastic consensus mechanism.

Machine Learning and Intelligence

The FireFly ledger is designed to support economic agents that reside within its network

with information and guidance. A key element of this support is the manner in which the

ledger exploits the information provided by the history of transactions undertaken by the

agents, beyond that contained in the value and the identities of the transacting parties.

A history of transactions can be used to understand preferences, thereby differentiate

agents, measure reputation, target delivery of options and so on. For example, if we

knew the past history of transactions in a supermarket at the level of individuals, one

could design an optimal coupon strategy that doesn’t provide discounts to consumers

who (from their past transactions) would continue to purchase the good irrespective of

a price increase.

It is important to emphasise that the use of a time series of individual transactions and

related information is not in and of itself novel. The novelty exists in the way in which

the data is used to create a prediction model. Moreover, distributed ledger systems that

have been designed to date, have not integrated the use of these histories as part of

the consensus mechanism. Current designs have provided a mechanism to authenticate

transactions, with other processes built on top. A significant disadvantage of assembling

a system in separate layers is that users have to deploy their own method and resource

to create intelligent models which can predict actions, trades, correlations etc.

FireFly in contrast, is providing this feature inherently, e.g to predict which two agents

amongst thousands are more likely to trade based on such attributes as trust, transaction histories, and owner preferences. It is in this sense that FireFly is an intelligence

driven system, providing a toolkit for agents with a built-in prediction model. As an

example, imagine an agent A interacting with other agents to trade. In connecting to

a marketplace, Agent A is actively provided services, such as optimal trading strategies

provided by machine learning using transaction data recorded on the ledger. Although

agent A still has to make a decision, such a system is able to present relevant services

much faster and much more accurately.

In a non-intelligence driven system we would observe a number of agents advertising

and an agent, say A, searching for the service. Once the relevant service is discovered

a decision is to be made by A based on a more or less hard coded criteria (a smart

contract). Once A decides to accept service/data from one of the agents the transaction

is recorded in the database.

Market intelligence

Market intelligence is a particularly interesting aspect of FireFlies activity. Over time,

it learns increasingly more about what kind of markets interact with others, under what

conditions and which ones overlap with others. This data has previously been held in

proprietary silos by large on-line markets such as Amazon and eBay, but for the first

time will be available publicly. This hugely valuable information becomes accessible

to participants in the marketplace as well as enabling smart market structure and an


additional layer of information for agents to leverage in order to maximise opportunistic

value use and increase utilisation of data and services.

Payment Systems

In this paper we have highlighted the limitations of existing market infrastructure both

in terms of providing a basis for blockchain as a new technology, and also considering

variants of blockchain. A critical issue in the use value of a number of blockchain protocols, and the applications which sit on top, is scaling and payment systems. In Section

5.2 we alluded to one solution based on the use of payment channels. Systems such as

Lightning or Raiden provide scalability to blockchain solutions by taking large numbers

of small payments off-chain. In an off-chain system the transactions are aggregated and

then placed on the main blockchain. Although this system provides authentication as to

which transactions happened such that any change would result in a non-consensus, the

information as to which parties were involved in the transaction, and for what purpose is

lost. This information is critical for intelligence driven agent systems like FireFly since in

order to provide agents with information (i.e the tools to find the right parties to trade

with), this type of information is required.

Stochastic Consensus Mechanism

Similarly to Bitcoin, Ethereum and many other ledgers, the FireFly system is designed for

rapid validation of transactions, and includes a set of deterministic rules for reaching a

consensus. However, the FireFly ledger differs from other blockchain protocols in that the

machine learning system, integrated with the design of the distributed ledger, operates

as a secondary layer that enables transactions to be assigned a probability of being

incorporated into the consensus chain. This probability reaches 1 when the consensus is

agreed according to the deterministic algorithm.

The advantage of adding a probabilistic approach to the consensus mechanism is that

this enables transactors to tradeoff a small probability of being defrauded via a double

spend with an increase in transaction time. This is particularly valuable for iot contexts

with goods (such as energy) and/or low value transactions. These probabilities can be

combined with other information such as the identity of the goods provider and historical

records to allow agents to assess their trust in counter-parties and thereby transact

rapidly and efficiently with each other.

8 Conclusion

In this paper we have sought to understand both the evolution and requirements of marketplaces that are emerging in digital economies. As a point of departure we considered

the Theory of the Firm and in particular the role of transaction costs economics, alongside scale and scope economies, in determining the dynamics of the boundary between

the firm and the market.

The evolution of how economic activity is organised is a running theme throughout this

paper. In this context we highlighted the pivotal role of the Internet in changing the costs

of both disseminating and retrieving information, and moreover its role in facilitating a

large matching mechanism for buyers and sellers. The emergence of both centralised and


decentralised peer-to-peer marketplaces is testimony to both understanding the incentive

properties of these markets, and designing the appropriate tools.

In considering the design mechanism for markets where trade is peer-to-peer without a

central authority, we focussed on the pivotal role of the decentralised consensus mechanism and the distributed ledger, placing the evolution of this technology relative to

existing distributed databases. We also examined the design of decentralised applications and their protocols, smart contracts, the use of digital tokens, machine learning

and the importance of trust and reputation systems.

As with the evolution of any technology, there are many variants of blockchain technology, based upon different protocols. To a greater or lesser extent these protocols

will succeed or fail in their stated aim of providing the market infrastructure which

supports a particular decentralised marketplace. Similarly, and as already stated with

respect to existing centralised platform markets, platforms that succeed will need to add

value beyond the simple matching of buyers and sellers, and the ability to execute smart


In writing on the phenomenon of computer mediated transactions, Hal Varian, the Chief

Economist of Google, has underscored the role of ’combinatorial innovation’, where the

component parts of these technologies can be combined and recombined by innovators to

create new devices and applications. It is instructive to view the way in which the FireFly

platform has integrated the layers of the decentralised stack in this light. Although the

protocol underlying the FireFly platform incorporates a number of singular innovations

such as a means to handle transaction authentication in markets with micro-payments,

the combination of agent-based technology with a transaction-level database integrated

with machine learning tools, provides the potential to add value through real-time learning.

Going forward we see the need for further research in a number of areas. First, there exist

considerable benefits in taking the simple model of peer-to-peer production developed

by Einav et al (2016) and adapting this for decentralised platforms. How and to what

extent do decentralised markets reduce entry costs, facilitate entry of smaller suppliers

and affect market structure? In addition, research is required into the dynamic properties

of markets where new and old technologies will, at least in the short-run, co-exist. For

example, in the face of new entrants to these type of markets, how will incumbents

respond? In the u.k electricity sector the emergence of distributed energy resources

has coincided with a move away from traditional unidirectional flows of energy from

generation, distribution, and retail. This shift has prompted incumbents to create a

number of new innovations, including the emergence of virtual power plants enabling

the aggregation of household-level savings in electricity consumption and the related

innovations in so-called passive demand response.

In addition there is a large extant literature on mechanism design for agent-based marketplaces. As we have stated, the design of market mechanisms for online decentralised

markets presents further opportunities to reduce the costs of entry and transaction costs,

and therefore increase competition. Much of the existing work on platform competition

in two-sided markets, the economics of peer production and market design, has been

carried out for centralised platforms. Key questions such as how does price setting in

two-sided platforms interact with the i) alternative consensus protocols and ii) volatility

in cryptocurrency, indicate that there remains much work to be done in applying the

principles of mechanism design to decentralised marketplaces.

By continuing to use the site, you agree to the use of cookies. more information

The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this.