Applying the conception of HCI engineering to the design of economic systems

Applying the conception of HCI engineering to the design of economic systems

Ian K. Salter

JP Morgan Chase & Co.,1 60 Victoria Embankment, London EC4Y 0JP, United Kingdom

John Long's Comment 1 on this Paper

Abstract

The Long and Dowell conception for HCI design (Long and Dowell, 1989) outlined the general design problem for HCI and contrasted between applied science and engineering disciplines of HCI. Salter (1995) sought to clarify the applied science conception through the application of Kuhn’s conception of science. Salter also built upon the work of Long and Dowell to produce a generic conception of engineering design. As part of this work the notion of preference was formalized. Building upon the generic conception a set of criterion for an engineering discipline is established. A general design problem for economics is outlined in order to apply the generic conception to the field of economics. Roth’s (2002) implicit conception of economic engineering is analyzed against the criterion and the formalized notion of preferences and found to be a consistent but not complete example of an engineering discipline. The general problem of economic design, based upon Long and Dowell’s approach, is employed to analyze a regulatory response (Turner, 2009) to the global financial crisis of 2007+ and develop a design-based solution to the problems. It is argued that the current applied science based responses to the economic crisis are insufficient and that a multi disciplinary engineering approach is necessary. This approach includes consideration of how economic participants interact with computers as part of the financial system.

Additionally, if you need help with a fairly unique way through which brands can achieve very high visibility in a considerably short time, Check out the train advertising cost.

Reprinted by permission of JPMorgan Chase & Co. 2009. JPMorgan Chase & Co. All rights reserved.

Acknowledgement: This paper would not have been possible without the aid of Professor John Long, who spent many hours of his time reading and discussing drafts. The errors contained within are however the author’s alone.

E-mail addresses: ianksalter@gmail.com, ian.salter@jpmorgan.com ((The author is not necessarily representing the views or opinions of JPMorgan Chase & Co.))

1. Introduction

In 1989 John Long and John Dowell (Long and Dowell, 1989) outlined different conceptions for HCI disciplines based upon the general problem the discipline seeks to address, and the consequent knowledge and practices that arise. Engineering disciplines have problems of design whereas scientific disciplines have problems of understanding and prediction. They conclude that in order to address the design problems of HCI an engineering discipline of HCI is required. The Long and Dowell conception took the form of a set of concepts that were used to structure work on HCI design.

Comment 2

The economist Alvin Roth, with respect to the field of economics, expresses similar thoughts:

‘Economists have lately been called upon not only to analyze markets, but to design them. Market design involves a responsibility for detail, a need to deal with all of a market’s complications, not just its principle features. Designers therefore cannot work only with the simple conceptual models used for theoretical insights into the general working of markets. Instead, market design calls for an engineering approach’ (Roth, 2002).

Roth outlines an implicit conception of economic engineering through the analysis of the redesign of the entry-level labor market for American doctors.

Salter (1995) developed a generic conception for engineering design, based on the Long and Dowell approach. In what follows this conception is extended through the addition of a number of criteria that a discipline should fulfill to claim it is an engineering discipline in the sense of Long and Dowell. The generic conception is then instantiated for economics through the postulation of a general design problem for economic engineering.

Through consideration of the American doctors labor market redesign, Roth’s implicit conception is analyzed with respect to the postulated general problem and the generic engineering conception. The aims of this analysis are:

• To validate the postulated general problem of economic design.
• To assess the consistency and completeness of Roth’s implicit conception with respect to the generic engineering conception.

Roth’s conception for economic design is restricted to the consideration of microeconomics and the design of individual markets. The continuing global financial crisis that began in 2007, and continued into 2008 causing considerable financial upheaval, has pushed the discipline of Economics to the forefront of public debate. It seems clear to many that some form of macro-economic redesign of the global financial system is necessary to ensure both economic revival and that such a crisis cannot happen again.

The UK Financial Services Authority (FSA) proposes a number of changes to the financial system through the Turner review (Turner, 2009). The postulated general design problem for economic engineering is used to analyze the Turner review. A simple redesign of the global financial system is proposed. The aims of this analysis and redesign are:

To consider the value of the use of the postulated general design problem in specific instances of design, even when the knowledge and practice of an engineering discipline are unavailable.

To illustrate that a design focused approach is of value in addressing the current crisis.
By considering these micro and macro instances of economic redesign it is argued that an discipline of economic engineering, of the type envisage by Long and Dowell for HCI, is required in order to provide solutions to economic design problems. It is further argued that economic engineering would draw practitioners from other disciplines and that those with a background in HCI will have a key part to play.

2. The conception of HCI design

Long and Dowell (1989) considers different possible conceptions for the discipline of HCI. Each type of discipline is characterized by three components:

• Knowledge.
• Practice.
• General problem.

Knowledge is used to support practice aimed at solving the general problem of a discipline. For the discipline of HCI the scope of the general problem is identified as:

‘humans and computers interacting to perform work effectively’ (Long and Dowell, 1989).

The scope of the general problem (see Fig. 1) is extended in another paper (Dowell and Long, 1989). The human and computer interacting together are thought of as an interactive worksystem. The concept of effective work is captured by the notion of desired worksystem performance, which is expressed in terms of both the desired quality of work and the costs of the human and computer components of the worksystem that are incurred in doing the work. Interactive worksystems exhibit actual performance, which is a function of the actual quality of work done by the worksystem and the actual costs incurred.

Fig. 1. The general problem for human computer interaction.

Comment 3

The Long and Dowell paper (Long and Dowell, 1989) outlined three different conceptions of the discipline of HCI, distinguished by the nature of their knowledge and practices. These conceptions are:

• Craft.
• Applied science.
• Enginering.

In what follows, the applied science and engineering conceptions from Long and Dowell (1989) are considered and illustrated with figures from Salter (1995). The nature of the knowledge and practices that correspond to each conception are considered in terms of their definition, operationalization, testability, and generalization.

Comment 4

The term definition is employed to mean the explicit definitions of the knowledge and practices. Operationalization is the transformation of the definitions of the knowledge and practices into a form that can be used and tested. The testing of the knowledge and practices is aimed at determining how well they support the general problem. Finally, knowledge and practices are general if they can be applied to more than one instance of the general problem. that correspond to each conception.

Comment 5

2.1. Applied science conception

The conception of an applied science design discipline describes a practice in the form of ‘specify and implement and test’ and knowledge in the form of ‘guidelines’. In the applied science conception of design, artifacts are still designed by a process of construction and evaluation and reconstruction. However, knowledge in the form of guidelines, derived from scientific knowledge, is used to guide the process. The conception of HCI as an applied science discipline is represented in Fig. 2.

Comment 6

By considering examples of HCI as an applied science discipline, Long and Dowell conclude that the knowledge and practices of HCI as an applied science are derived from scientific theories that are operationalized, tested and generalized. However, the knowledge and practices themselves are not operationalized, tested and generalized with respect to the general problem of HCI. The limited account given of the relationship between scientific knowledge and applied science design in Long and Dowell (1989) was not well understood at the time. Salter (1995) attempted to clarify what was meant through consideration of Thomas Kuhn’s (1970) conception of science (see Fig. 3).

Fig. 3. Kuhn and applied science.

In Kuhn’s view, a scientific discipline may be conceived of as a framework that contains the following elements:

• Paradigm.
• Disciplinary matrix.
• Shared exemplars.

For Kuhn, the history of a scientific discipline cycles through two distinct stages.

The shortest phase is the crisis period. During this period the symbolic generalizations, metaphysical assumptions and system of values that form the disciplinary matrix are in question. Rival positions abound until one begins to dominate.

At this point, the discipline moves into a period of normal science. During normal science the scientific community holds a consensus view concerning the disciplinary matrix. Scientists may solve scientific problems within the paradigm of shared exemplars. The shared exemplars, consisting of theory predictions, are developed until such time as it becomes increasingly difficult to develop exemplars whose theory predictions accord with perceived phenomena. At this point a period of crisis ensues.

The relationship between Kuhn’s conception of science and Long and Dowell’s conception of applied science is presented in Fig. 3.

Disciplines of different types attempt to solve different design problems. Scientific disciplines attempt to solve the problems of understanding, whereas engineering disciplines attempt to solve the problem of design. Transforming discipline knowledge and practices that have been operationalized, tested and generalized with respect to the scientific problem of understanding, does not necessarily lead to knowledge that can be operationalized, tested or generalized with respect to the engineering problem of design.

2.2. Engineering conception

The conception of an engineering design discipline describes a practice in the form of ‘specify then implement’ and knowledge in the form of ‘principles’. In the engineering conception of design, artifacts are designed by a process of specification, followed by a process of implementation. The process is supported by knowledge in the form of principles. The application of principles to the design process provides a guarantee that the artifact will satisfy the client’s requirements. The conception of HCI as an engineering discipline is represented in Fig. 4.

Comment 7

Fig. 4. The engineering conception of HCI.

The knowledge and practices of an engineering discipline are defined, operationalized, testable and generalizable.

The Long and Dowell (1989) conception assumed that HCI was essentially in a Kuhnian crisis period and the 1989 paper can be viewed as an attempt to establish disciplinary matrix, albeit one based on an Engineering approach. Salter (1995) took this work further by establishing a generic conception of an engineering discipline.

3. A generic conception of engineering design

Salter (1995) built upon the work Dowell and Long as well as Kuhn to build a generic conception of engineering design. In what follows this generic conception has been extended to include a set of criteria to which any specific conception should correspond in order to satisfy the generic conception of design.

3.1. Disciplinary matrix and scope of the general problem

Recall that Kuhn’s notion of a disciplinary matrix consists of symbolic generalizations, metaphysical assumptions and systems of values of the discipline. By introducing notions of desired performance and interactive worksystems exhibiting actual performance, Long and Dowell are explicitly outlining some of the metaphysical assumptions and systems of values of the discipline of HCI. Thus, describing the scope of a discipline problem amounts to providing two of the three components of a disciplinary matrix.

Comment 8

Kuhn asserts that the value of knowledge is in its usefulness in solving problems. According to Salter (1995) design problems have two key components: the requirements component and the artifact component. The requirements component is the ‘what’ of the design problem. In Long and Dowell’s terms the requirements component is the desired performance of the interactive worksystem. The artifact component represents the ‘how’ of the design problem. In Long and Dowell’s terms, the artifact component is the interactive worksystem together with the actual performance it exhibits.

Comment 9

Any conception of an engineering discipline of design must describe the scope of the general problem. This leads to the first criterion.

Criterion 1: The description of the general problem should describe the requirements component and the artifact component of the problem and the nature of the relationship between them.

3.2. The phenomena of design

The shared exemplars of Kuhn’s framework are grounded in perceived phenomena. ((The use of the word phenomena here is used to refer to an observable occurrence.So Kuhn was stating that shared exemplars must be consistent with observable occurrences. Salter (1995) was claiming that there are a set of observable occurrences associated with client requirements and a set of observable occurrences associated with an artifact.))

Salter (1995) distinguishes two types of phenomena associated with design problems, the phenomena associated with the requirements component and the phenomena associated with artifact component (see Fig. 5).

Fig. 5. Design phenomena.

The phenomena associated with the requirements component of the design problem are called the client requirements. A client here refers to an individual, or an organization whose requirements may consist of the, possibly conflicting, requirements of sub-organizations and individuals. The phenomena associated with the artifact component of the design are termed the artifact. For any design problem it is necessary that there be practices for determining whether artifacts fulfill or satisfy client requirements. These practices are termed empirical since they apply to phenomena. The term empirical derivation is used to describe a practice for developing an artifact from client requirements and the term empirical validation is used to describe a practice for determining if an artifact satisfies client requirements.

Figs. 1 and 5 can be combined to create a re-expression of the HCI design problem in terms of requirements and artifact ((Some commentators are unhappy with identifying the domain solely as part of client requirements, arguing that the artifact would also need to include the domain as the domain may be redesigned to improve performance. However such a modification of the domain can also be perceived as a modification of the requirements that is outside of the scope of the HCI problem.)) (see Fig. 6).

The success of any engineering discipline depends upon its ability to support the design of artifacts that satisfy client requirements. This leads to criterion 2:

Criterion 2: The conception of any discipline of design must have a set of empirical practices for establishing the relationship between artifact and client requirements.

Comment 10

3.3. Design practice exemplars

Lying between the disciplinary matrix and phenomena in Kuhn’s framework is the concept of the paradigm of shared exemplars. For a scientific discipline, the general problem is one of understanding by means of explanation and prediction. Thus Kuhn’s exemplars form understanding as explanations and predictions. Long and Dowell (1989) stated that for an engineering discipline, the general problem is one of design. Salter (1995) concluded from this that one form of shared exemplars are design examples representing abstractions of client requirements and artifacts and the relationships between these abstractions. (see Fig. 7).

Salter (1995) termed the abstraction of client requirements the specific requirements specification, whereas the abstraction of artifact was called the specific artifact specification. The term ‘Specific’ is used since these abstractions are specific to the particular design problem being considered. Salter (1995) stipulated that the relationship between specific requirements specification and specific artifact specification should be formal. This gives rise to criterion 3.

Criterion 3: The conception of any discipline of design must have each of the following:

• a) A set of empirical practices for establishing the relationship between client requirements and specific requirements specification.
• b) A set of empirical practices for establishing the relationship between artifact and specific artifact specification.
• c) A set of formal practices for establishing the relationship between specific requirements specification and specific artifact specification.

Comment 11

3.4. Design research exemplars

For a scientific discipline, Long and Dowell (1989) believed that the practice of the discipline is the research that aims to construct and validate knowledge that supports understanding in the form of explanation and prediction. Salter (1995) claimed that for an engineering design discipline, whose knowledge is defined, operationalized, tested and generalized with respect to the problem of design, there is a distinction between practice and research. Engineering practice involves employing engineering knowledge to solve specific design problems, whereas engineering research involves the construction and validation of engineering knowledge.

Thus, for an engineering discipline, Salter (1995) claimed that there is an alternative but equivalent of Kuhn’s shared exemplars, that is, examples of engineering research. Since engineering research constructs and validates knowledge that is defined, operationalized, tested and generalized with respect to the problem of design, engineering research exemplars consist of abstractions of specific requirements specification and specific artifact specification and the relationships between them (see Fig. 8).

Salter (1995) termed the abstraction of the specific requirements specification the general requirements specification, whereas the abstraction of the specific artifact specification was called the general artifact specification. The term ‘General’ was used, since these abstractions are general to particular classes of design problems. The relations between all abstractions are formal. This gives rise to criterion 4.

Criterion 4: The conception of any discipline of design must have each of the following:

• a) A set of formal practices for establishing the relationship between specific requirements specification and general requirements specification.
• b) A set of formal practices for establishing the relationship between general requirements specification and general artifact specification.
• c) A set of formal practices for establishing the relationship between general artifact specification and specific artifact specification.

Fig. 6. The re-expression of the HCI design problem.

Comment 12

Fig. 7. Design practice exemplars.

Fig. 8. Design research exemplars.

Salter’s (1995) generic conception of an engineering discipline has now been outlined and extended with a set of criteria that any specific instance of an engineering discipline should meet. In order to instantiate the generic conception for engineering design for a discipline of economic engineering it is necessary to postulate the general design problem for economic systems.

Comment 13

4. The general design problem for economic systems

The general design problem for economic systems must have a clearly defined scope with the problem expressed in terms of requirements and artifact. Fig. 9 modifies Fig. 6 for application to the problem of the design of economic systems:

Fig. 9. The general design problem for economic systems.

Economic agents interacting together within a regime ((Regime here is used to denote a set of regulations. This set may be empty as in thecase of illegal markets such as the black economy or illicit drug dealing. Even foreconomic systems such as these there may however, be implicit regulation – it is notnormally considered good business for drug dealers to kill their clients en mass–eventhe death of one client is likely to attract unwanted attention of the authorities)) of regulation are thought of as an economic system, the artifact. The agents of the economic system are further distinguished into clients, whose requirements the system seeks to serve, and non-client agents. Agents may include human, organizations, software systems, making economic design a cross-disciplinary activity.

The concept of effective work is captured by the notion of desired economic system performance, which is expressed in terms of client preference. Client preferences are expressed with respect to a work transformation. Client preferences may be common across all clients of the economic system or may be specific to a particular client group.

The notion of economic systems doing work is somewhat novel but central to the notion of economic engineering being expounded here. It is illustrated in the macro and micro economic discussions in Sections 5 and 6.

A standard table format is used to describe economic requirements in terms of work transformation and client preferences. This format is given in Table 1.

A standard table format is also used to describe an economic system. This format is given in Table 2.

Domain
Work transformation Describes the work done by the economic system in terms of some work transformation of objects Client preferences Describes preferences of particular groups of clients Common preferences Describes preferences common to all clients

Table 1: Standard format for an economic design problem.

Economic system: sample system
Clients Describes the clients that form part of the economic system. There should be a correspondence between the clients listed here and the client preferencess Non-client agents Describes any non-client agents that are part of the system and the roles that they play Regulation Describes any regulation that is part of the economic system

Table 2: Standard format for an economic system.

Economic system worksystems exhibit actual performance, which is a function of how the work transformation achieves client preferences.

4.1. Formalizing client preferences

From the generic engineering conception the specific and general requirements specification for economic systems should be presented formally. In order to achieve this the notion of desired performance and hence client preferences must be expressed formally.

Traditionally in economics and game theory (Osbourne and Rubinstein, 1999) the preferences have been stated using a utility function or the slightly weaker notion of a linear ordering. Utility functions assign a real number to the occurrence that forms the actual performance of the worksystem. Liner orderings insist that all actual work system performances can be ranked along the real line, insisting therefore that it is possible to state for each actual performance, which is preferred. Work on bounded rationality (Simon, 1957) indicates that both of these notions are too strong. Clients are unlikely to be able to assign a number to actual worksystem performance and equally as unlikely to be able to rank all such performances.

Salter (Salter (1995)) introduced the notion of preference ordering, based upon mathematical pre-orders to address this issue. Linear orderings and utility functions are special cases of preorderings.

A pre-ordering hS, 6i is a set S together with a relation 6 over S such that for all a, b, and c in S:

• a 6 a (reflexivity)
• if a6b and b6c then a6c (transitivity)

A preference ordering hS, 6, Pi is a pre-ordering hS, 6i together with a subset of S, P of preferences, such that for any a and b in S:

• if aeP and a6b, then beP

Client preferences may be expressed as a preference ordering over the work transformations that are carried out by the economic system.

Work (Salter, 1995, 1993) indicates that Mathematical Category Theory (MacLane, 1998; Lawvere and Schanuel, 1987), of which the Theory of Orderings can be considered a part, may be an appropriate formalism for expressing some class of engineering design problems.

The generic conception of engineering design has been outlined and instantiated through the postulation of the general design problem from economic systems. Given this statement of the general problem and equipped with the tool of preference orderings it is possible to analyze the entry-level market for American doctors and Roth’s implicit conception of economics engineering.

Comment 14

5. The entry-level labor market for American doctors

Traditionally the role of the majority of economists has been to understand and predict aspects of economic systems, thus playing the role of scientists according to the conception of Long and Dowell. Increasingly however economists are becoming actively involved in the construction of economic systems including the design of markets for electric power (Wilson, 2009), airwave spectrum auctions (Milgrom, 2000) and labor markets for entry-level doctors (Roth, 2002; Roth and Peranson, 1997) Thus economists are increasingly playing the role of designer.

An example of the economic design of the entry-level labor market for American doctors reported by Roth (2002) is considered. The concept of residency programs (then called medical internships) began in the 1900s. Newly graduated doctor’s work as residents in hospitals to complete their training. The work of the economic system is to match doctors to residency programs within hospitals. The matching should take account of doctor’s preferences for residency programs and the residency program’s preference for doctors.

Table 3 uses the standard table format for the domain to describe the economic design problem.

The domain described is, by necessity, a simplification ((The requirements are also simplified to aid exposition. For example medicalschools could have been included as participants in an attempt to capture the effortexpended by them in supporting medical student)) of requirements that have evolved over time. The rationale behind the common preferences aspects of the domain may not appear immediately clear. However a discussion of the history of the problem will provide this rationale.

The initial economic system operated a decentralized model with no major regulations (see Table 4).

This system worked well initially but competition amongst hospitals for medical students led to appointments of doctors to positions up to 2 years prior to taking up the residency position. This was not considered satisfactory, leading to the requirement given by the first common preference that the matching should occur in the doctor’s final undergraduate year. Thus justifying common preference (a). This led to a revised economic system (Table 5) in which additional regulations were imposed upon passing student information and date of appointments.

Domain

Work transformationTo match newly qualified doctors to resident programs. Where resident programs consist of a number of available residency positionsClient preferences Doctor For residency programs Residency programs Preferences for doctor Common preferences a) The matching between doctors to programs should not happen earlier than the final year of the doctors undergraduate program b) The matching should occur in a reasonable time, say tbeforeResidency, before the doctor is due to start the residency c) The matching should be simple for the doctors, not occupying more than, say rdoctors, of the doctor’s resources d) The matching should be simple for the residency programs, not occupying more than, say rresidencyProgram, of residency program’s resources e) The matching should occur consistently throughout the year, with not more than say m matches occurring in any one month f) The match should be stable in the sense that it should not be possible to produce a different match that better meets both doctor’s and residency program’s preferences

Table 3: The design problem for the entry-level doctors market.

Economic system: decentralized model (1900)
Clients Doctors applying for advertised positions Resident Programs advertising positions and selecting doctors through a waiting list system Non-client agents No major ones except for advertising outlets, etc. Regulation None

Table 4: The doctor matching system 1900s–1940s.

Economic system: decentralized model (1945)
Clients Doctors applying for advertised positions Resident programs advertising positions and selecting doctors through a waiting list system Non-client agents No major ones except for advertising outlets, etc. Regulation Dates references and transcripts could be sent restricted Contracts cannot be signed till last year of study

Table 5: The doctor matching system 1945.

The new system led to further problems as doctors receiving an offer from one residency program held off accepting until they had heard from a preferred program. This led to slow movement of waiting lists and lots of last minute matching which was also considered unacceptable. This was not considered satisfactory thus justifying common preferences (b)–(e). This led to the development of a centralized clearing system in 1951 (see Table 6).

The centralized clearing system worked well until the 1970s but actual performance degraded until the 1990s. Many issues were raised, in particular as to whether the matches produced were considered optimal. Thus leading to common preference (f). To address these needs, the National Resident Matching Program (NRMP) was developed using a clearing system algorithm by Roth and Peranson (1997) (see Table 7).

Economic system: centralized clearing (1951)
Clients Doctors submit preferences for positions Resident programs submit preferences for doctor Non-client agents A centralized clearing process Regulation None

Table 6: The doctor matching system 1951.

Economic system: US National Resident Matching Program – NRMP (1998)
Clients Doctors submit preferences for positions Resident programs submit preferences for doctors Non-client agents Engineered centralized clearing process Regulation None

Table 7: The NRMP doctor matching system.

Roth considers the redesign to be an example of an economic design discipline:

‘These developments suggest the shape of an emerging discipline of design economics, the part of economics intended to further the design and maintenance of markets and other economic institutions’ (Roth, 2002).

Roth considers the role of the economist as one of engineer, and implicitly outlines a conception of economics engineering (Roth, 2002). In what follows, this conception is made explicit and analyzed against the generic engineering conception outlined above.

For the NRMP design the general requirements and artifact specifications are partly described in terms of Mathematical Game Theory. The following is taken from Roth (2002) but presented in the style of preference orderings outlined above. This is done in order to illustrate that Roth’s work is consistent with the notion of preference orderings and to facilitate the analysis of the NRMP general requirements against the generic engineering conception: The NRMP requirements are as follows:

• There are two disjoint sets of F 1⁄4 ff1;…fng of firms and W 1⁄4 fw1;…wpg of workers where each firm fi seeks up to qi workers and each worker seeks exactly one firm.
• A match is a functionl : F ! }ðWÞ that maps firms to sets of workers. The set of all such functions is called M.
• For each worker wieW there is a worker preference ordering hF; wi ; Pwi i representing the worker’s preferences for firms such thatforallfi;fj 2Pwi; fi <fj; fi >fj orfi=fj.In other words all of the preferences are linearly ordered and can thus be represented as a sequence hfi, fj, . . .i.
• For each firm fi e F there is a firm preference ordering hW; fi ; Pwi i representing the firm’s preferences for workers such that for all wi;wj 2Pfi, wi <wj, wi >wj or wi =wj. In other words, all of the preferences are linearly ordered and can thus be represented as a sequence hwi, wj, . . .i.
• A match preference ordering hM,6,Pi can be specified on the setof matches M. The set of preferences is defined as P1⁄4fl2MklðfÞjq^ðw2lðfÞ)w2Pf ^f 2Pwg. In other words, for a matching to be a preference, no firm should be matched against more workers than it requires and all workers and firms should be matched against their preferences. The ordering is specified with two conditions. The first condition states that if |l(f)| < q and l0 is identical to l except that l0 (f) = l(f) [ {w}then l < l0 whenever wePf and fePw In other words a matching that adds an additional worker to a firm that is not at capacity is preferred, as long as the worker and firm are in each other’s preferences. The second condition states that if w0 < f w, f0 < w f, wel(f0),w0el(f)andwel0 (f)inlandl0 thatareotherwise identical then l < l0. In other words a mapping is preferred if it provides a better match for any worker and firm pair.
• Using the terminology of Roth a match l is considered Stable if there is no match l0 such that l < l0. In other words, a stable match cannot be improved upon by better satisfying user’s preferences. This formalizes common preference (f).

The postulated general problem of economic design has been used to describe the client requirements and artifacts for the redesign of the entry-level labor market for American doctors. Further, the mathematical formalization of preference orderings has been shown capable of capturing the formal aspects of the matching problem. Thus the general problem of economic design has been validated against the labor market redesign example.

The consistency and completeness of Roth’s implicit conception is now analyzed against the generic engineering conception. Criterion 4 is considered first.

There is a general requirements specification given in terms of the sets F and W and there are associated preference orderings. A general artifact specification is an algorithm that will always produce a stable match. Such an algorithm is documented by Gale and Shapley (1962) who go further and prove a theorem that given the general requirements there will always be a stable match and that a stable match will be produced by their algorithm. Thus the formal relationship between general requirements specification and general artifact specification has been satisfied. Thus Roth’s conception is consistent and complete with respect to part (a) of Criterion 4.

In the Roth paper the specific requirements specification is given semi-formally as an English language description of the particular matching problem in the case of doctors and residence programs. The sets F and W are matched to residency programs and doctors respectively. Thus the general requirements specification and the specific requirements specification are related, all be it semi-formally. In a more formal approach specific requirements specification might be stated in terms of a design language such as Universal Modeling Language (UML) (Fowler, 2003). For Software Engineers with some knowledge of set theory and orderings, a formal relationship could be constructed between the UML specification and the game theory definitions, the general requirements specification. In the case of the problem discussed above, this would, for example, map a class Doctor to the set W and the class Residency Program to the set F. Thus, Roth’s approach is consistent, but not complete with respect to part (b) of Criterion 4.

Roth outlines no specific artifact specification. From a software engineering perspective a specific artifact specification, in the form of a computer program, may be produced in a software programming language such as Java (Flanagan, 2005). For software engineers the relationship between an algorithm, the general artifact specification, and a software program implementing that algorithm, the specific artifact specification may be considered formal. It is therefore possible to complete Roth’s conception, which is consistent but not complete with respect to part (c) of Criterion 4.

Thus, Roth’s conception appears consistent but not complete with respect to Criterion 4. Criterion 3 is now considered.

Even though Roth’s description of the specific requirements specification, is only semi-formal, he lists involvement with representatives of both student doctors and resident programs at the beginning of the design, ongoing contact with stakeholders throughout the design and final agreement for the design. Thus establishing the empirical relationship between the phenomena of client requirements and the specific requirements specification, even if the latter is semi-formal. Thus Roth’s conception is consistent and complete with respect to part (a) of Criterion 3.

As mentioned above Roth exposes no notion of specific artifact specification. However, from a software engineering perspective, the Java program described above would be instantiated as a running process on some physical computation hardware establishing an empirical relationship between specific artifact specification and artifact. It is therefore possible to complete Roth’s conception, which is consistent but not complete with respect to part (b) of Criterion 3.

The relation between the UML specification and the Java program may be established using tools such as the Rational Unified Process (RUP) (Kroll and Kruchten, 2003; IBM Website, 2009), Java classes can be generated from the UML specification and the UML specification reversed engineered from the Java classes using tools such as Rational Rose Developer for Java (IBML Website (2009)). Thus a formal relationship between specific requirements specific and specific artifact specification can be established. It is therefore possible to complete Roth’s conception, which is consistent but not complete with respect to part (c) of Criterion 3.

Thus, Roth’s conception appears consistent but not complete with respect to Criterion 3. Criterion 2 is now considered.

Although not reported by Roth, it may be presumed that prior to operation user acceptance tests would have be performed to ensure that, artifact, the implemented process, met client requirements. Roth does report that NRMP system has been operating successfully for a number of years. To consider the importance of stability Roth, also considers a number of alternative matching systems in operation in the US, the UK and Canada. From a total of 16 systems, the nine systems that produce a stable match are all still in operation. Of the seven systems that do not produce a stable match, all but two have failed. The analysis of the operating systems was supplemented by laboratory experiments to ensure there were no other reasons for the failures. Thus the empirical relationship between the specific artifact discussed and client requirements is produced. Further to this, the relationship has been established for a number of different specific design cases. Thus Roth’s conception is complete and consistent with respect to Criterion 2. Consider Criterion

Roth does not explicitly describe a general problem of economic design. However, it has been possible to map the design problem considered to the postulated general problem for economic design. Thus Roth’s is implicit conception is consistent but not complete with respect to Criterion 1.

Summarizing across the criteria, it can be seen that, given the postulated problem of economic design, Roth’s conception is consistent with an engineering conception of economic design.

Further to this, the formal relationship outlined between general requirements specification and general artifact acts as a specification of engineering principle. This Match Stability Principle may be stated as follows:

In order to produce stable matches (see common preference f), the Gale and Shapley matching algorithm should be used.

The sources of the incompleteness of Roth’s implicit conception are worth considering.
In order to demonstrate the consistency of Roth’s conception to the generic engineering conception, aspects of the design were artificially completed using tools from the field of software engineering. This indicates that the knowledge and practices from this field may prove useful in building an engineering discipline of economic design.

It should also be noted that the design exemplar has focused on one of the performance criteria, common preference (f). Roth (2002) actually does focus on other issues including concerns about whether preference orderings are worker or firm optimal. The consideration of these has not been possible due to reasons of space. However there still remain issues that have not been addressed by Roth. Common preferences (d) and (e) that are related to resources expended by doctors and resident programs in using the system. It may be argued that the general design problem is an abstraction and these were not seen as being a problem for the design before NRMP. On the other hand, this is a significant cognitive design problem, which may have an impact on overall acceptance of the economic system.

Further cognitive issues arise in relation to assumptions as to how preferences are stated. The game theoretic model assumes preferences can be ordered. This is a standard assumption for game theory. It is a cognitive issue as to whether this is necessarily the case. The consideration of these issues would likely involve the application of knowledge and practices from the emerging discipline of cognitive engineering (Dowell and Long, 1998).

The analysis of the completeness issues of Roth’s conception indicates that knowledge and practices from other disciplines may form an important part of any discipline of economic engineering.

It is important to consider also that the design exemplar given above is a simplification of the actual problem for the market being designed. Married couples have preferences for residency programs in the same locale. Unfortunately game theoretic results state that if such interrelated preferences are added there is not guaranteed to be a stable match within the match preference ordering (Roth, 1984). This issue might seem to significantly reduce the effectiveness of the stability matching principle. However Roth and Peranson (1997) indicate through theoretical computations and analysis of actual matching data from the NRMP system that cases where stable matches are not produced are extremely rare in practice.

The way the emerging discipline of economic design seeks to address this type of issues is key to the type of discipline that emerges. The value Long and Dowell (1989) adds here is the distinction between applied science and engineering disciplines. If the discipline is happy to rest with simple theoretical models that offer guidance to the design of actual systems it will have the knowledge associated with an applied science discipline and the match stability principle will really be a match stability guideline. If, however, as Roth appears to advocate, the formal knowledge is elaborated to provide principles that, for example, establish under what preference correlation and numeric conditions stable matches are guaranteed, the discipline will have the knowledge of an engineering discipline.

Roth’s implicit conception that focuses on microeconomics and the design of individual markets has provided a contrast between applied science and engineering disciplines. In what follows the macro-economic design issues engendered by global financial crisis are considered from the perspective of the general design problem for economic engineering.

6. The late 2000s global financial crisis

A brief description of the global financial economic system pre2009 is provided.(Table 8).

Economic system: global financial system (pre-2009)

Clients Consumers – – Investing – – Obtaining credit Firms and other legal entities – – Raising capital – – Investing Non-client agents Financial institutions – – Banks (retail and wholesale) – – Hedge funds Credit reference agencies – consumers (Equifax, Experian, etc.) Credit rating agencies – non consumer (Standard and Poor, Moody, etc.) Government – – Central banks – – Regulators Regulation Capital adequacy requirements National regulatory supervision Accounting regulation Deposit insurance

Table 8: The global financial system (Pre-2009).

Problems in the actual performance of the system emerged in 2007, as a result of a collapse of the US Housing bubble. In 2008 the global financial system failed in a catastrophic manner (Wikipedia, 2007–2009):

• A number of financial institutions found themselves in financial difficulties leading to effective take over by the government (Northern Rock – UK, Fannie Mae – US, Freddie Mac – US), take over by other institutions (Bradford & Bingley – UK, Bear Stearns – US, Merrill Lynch – US, Washington Mutual – US) and even bankruptcy (Lehman Brothers – US).
• In order for financial institutions to meet capital adequacy requirements the governments of the US and the UK had to inject $700 billion and £500 billion into financial institutions, respectively.
• Global equity markets collapsed wiping trillions of dollars off asset values.
  The failure led to a Credit Crunch that has made capital and credit significantly more difficult to obtain and has simultaneously reduced returns on investment, further degrading the actual performance of the system.

For now, a description of the domain of the global financial system will be forgone as a redesign, proposed by the UK Financial Services Authority (FSA) is considered. As might be imagined there are many other potential solutions including, for example, a return to Keynesian approach to economics that dominated between the 1940s and 1970s (Gailbraith, 2008; Stratton and Seager, 2008). In what follows only the FSA’s response is considered in order to illustrate the potential pitfalls of considering redesign without a clear understanding of the design problem.

6.1. The FSA response to the crisis

The Turner Review (Turner, 2009) proposes alterations to the global financial system to avoid a recurrence of recent catastrophic actual performance failures in the future. In order to simplify the presentation only some of the main points of the redesign are listed (see Table 9).

It is perceived that financial systems were not sufficiently capitalized going into the crisis so the FSA recommends that:

‘The quality and quantity of overall capital in the global banking system should be increased’ (Turner, 2009).

Unsurprisingly, the FSA proposes a number of extensions to regulatory supervision including extending regulation to hedge funds and regulation of offshore financial centers based upon globally agreed regulatory standards.

Economic system: global banking system (FSA recommendations)

Clients Consumers – – Investing – – Obtaining credit Firms and other legal entities – – Raising capital – – Investing Non-client agents Banks (retail and wholesale) Credit reference agencies Government Regulation Increased capital adequacy requirements Extended regulatory supervisio Accounting regulation to include off cycle reserves Regulation of credit rating agencies Increased deposit insurance Modified remuneration structure

Table 9: The FSA response to the crisis.

Domain

Work Transformation To match consumer’s and other legal entities investment and credit requirements To enable transactions between clients and to provide reporting of these transactions Client Preferences For investment and credit matches For transaction type For transaction report Common Transactions should be carried out between correctly identified clients Clients should be aware of the credit worthiness of other parties to transactions Client knowledge of the nature of transactions should be sufficient for them to be aware of the risks associated with different transactions

Table 10: The design problem of the global financial system.

The FSA believes that the risk management methodologies in use by financial institutions, in particular, the Value at Risk (VAR) (Gastineau and Kritzman, 1996) method did not sufficiently take into account the fluctuations in the economic cycle and thus the ‘Capital required against trading book activities should be increased significantly (e.g. several times)…’ And that published accounts should include an ‘Economic Cycle Reserve’.
The crisis has in part been attributed to failures of the credit rating agencies and so the report proposes that:

‘…they should be subject to registration and supervision to ensure good governance and management of conflicts of interest and to ensure that credit ratings can only be applied to securities for which a consistent rating is possible.’

The FSA recommends that retail deposit insurance is increased ((This measure has already been taken in the UK.)) and that depositors should be made aware of the extent of the insurance cover.

Finally, from the perspective of our analysis, the FSA proposes changes to the remuneration structures within financial institutions to ‘avoid incentives for undue risk taking’.
The recommendations of the Turner review should go a long way to preventing a recurrence of a similar crisis thus avoiding potentially catastrophic actual performance failures in the global financial system. However, the review does not specify a design problem for the global financial system. The closest it comes is to provide a list of ”What went wrong?” This implement and test approach is typical of a craft discipline as outlined by Long and Dowell. Without the consideration of a design problem that identifies desired performance, there is no way to judge, a priori, whether the Turner recommendations will, if adopted act to improve or degrade the ongoing overall actual performance of the global financial system. To attempt such analysis it is necessary to attempt to specify the general design problem for the global financial system.

6.2. A conception based response to the crisis

A simple domain model of the global banking system is postulated in Table 10.
The work transformation is seen as having two elements. The first of these is to match ((It may be argued that the work transformation should be to meet rather than just match consumer and legal entities investment requirements. The distinction between match and meet is a subtle one. Using the formalism of preference orderings, to meet requirements every client’s preference for investment and credit would need to be in the specified set P. This is unlikely to be possible unless some condition on the reasonableness of client preferences with respect to other participants. This seems too strong a condition to impose.)) investors to those requiring credit. This is taken to mean all financial transactions from simple deposits and loans, to complex derivative transactions and indeed any transaction that manages risk. The second element of the transformation is the services of banks that are a key component to the financial system, the provision of transaction management and reporting services to customers.

The client preferences capture preferences for investment and credit matches as well as the type of transactions that may be carried out and the nature of the reports on these transactions.

The common preferences capture key aspects required of the banking system: identity verification, credit reference and rating, and provision of financial advice.

Some of the FSA recommendations outlined above, in particular improved regulation of credit rating agencies and ensuring depositors are aware of the nature of deposit insurance should act to improve actual performance for this domain. However, other effects are likely to have a negative impact on ongoing performance. In order to increase capital and meet requirements for off cycle reserves, financial institutions will need to increase the spread between investment and credit thus providing less preferred investment-credit matches to clients. Increased regulation will also increase costs to financial institutions further increasing spreads. A more worrying aspect of increased regulation is the consequently increased barrier to entry for new financial institutions. This occurs on top of a radical financial institution consolidation that has already increased such barriers. The effect of raised barriers to entry is to reduce innovation in the financial institutions and reduce competition, thus leading to a further degradation of actual performance.

Economic system: proposed redesign of global financial system

Clients Consumers – – Investing – – Obtaining credit Other legal entities – – Investing – – Raising capital Non-client agents Transaction management institutions with functions: – – Transaction management – – Transaction reporting & accounting – – Identity verification – – Credit rating and reference provision – – Financial advice provision Investment institutions Government – – Central banks – – Regulators Regulation Capital adequacy requirements for banks Global regulatory supervision Based upon a tiered structure of investment intuitions. With increased regulation for higher tiered institutions. Government guarantees pay- ments for all tier one institutions. Level of government guarantees tails off as you go down the investment institutional tiers Accounting regulation – enforced through transaction management insti- tutions, extended based upon investment institution tier

Table 11: The proposed redesign of the global financial system.

The problem is not that more regulation is not needed but rather that regulation needs to be considered as part of a redesign of other aspects of the financial system. The statement of a design problem opens up the possibility for the knowledge and practices of other disciplines to be employed towards the development of a solution. In Table 11 a design is presented that employs an approach normally used by software system architects:

‘The software architecture of a program or computing system is the structure or structures of the system, which comprise software elements, the externally visible properties of those elements, and the relationships among them.'(Bass et al., 2007).

Such an approach considers the structure and relations of the components of the system, in the economic system case, the non-client agents, and how these components may be reconfigured to improve actual performance. To justify the proposed redesign evidence is given that agents that perform some of the functions described are already emerging.

A key aspect of this redesign is a separation of concerns amongst financial institutions. Such an approach was made after the great depression of the 1930s in the US Glass-Steagall Act (1933), the partial repeal of which in the Leach-Bliley Act (1999) (Barth et al., 2000) has been partly blamed for the current crisis (Halligan, 2009; Housel, 2009). The separation proposed in this redesign, however, goes much further than the 1933 act. The primary separation is between institutions with the function of transaction management and institutions whose function is investment.

The transaction management institutions play no investment role and therefore are risk neutral with respect to investment risks. They fulfill client’s transaction requirements moving money between different client’s investment institutions. It can be argued that emerging organizations are already fulfilling this role (e.g. Paypal ((Paypal actually does hold deposits but this is to facilitate transactions, rather than for investment purposes.))

(Paypal, 2009). A key aspect of any transaction management system is the correct identification of the clients that are party to a transaction. Thus, identity verification is a key role of transaction management institutions. Other functions of transaction management institutions include transaction reporting and financial advice. Web-based organizations have emerged that play this function Wesabe (2000) and Geezeo (2009). These entities go further than merely reporting financial transactions, enabling users to analyze their expenditure by constructing accounts. Finally transaction management institutions, with full knowledge of client’s transactions (and even accounts), and independence from any investment incentive, are ideally placed to offer accurate credit reference and rating information.

The concept of separation of concerns is carried further into investment institutions. A comment from Nobel laureate Paul Krugman acts as motivation for this separation. He postulates the following simple rule:

‘. . . anything that does what a bank does, anything that has to be rescued in a crisis the way banks are, should be regulated like a bank'(Krugman, 2009).

The separation for investment institutions aims to tie the level of regulation to the level at which the institutions will be rescued by governments. At the highest tier might be simple deposit takers that make low credit risk loans. Such institutions would be rescued in full in the event of a collapse through a deposit insurance scheme. By necessity, these institutions would be highly regulated. There are already emerging institutions such as (Zopa (2009) fulfilling this function putting depositors and lenders together in a radically different fashion to that used by existing banking institutions. At the lowest tiers are lightly regulated institutions that are involved in highly risky investments but that under no circumstances would be rescued by government. Transaction management institutions, through their role as independent financial advisors to their clients, are responsible for ensuring that clients understand the difference between the different types of institutions.

The simple proposed design should act to improve actual performance of the global financial system.

Improved identity management should better ensure that transactions are carried out between correctly identified clients. Such identity management may also have applications beyond the financial sector helping to improve identity management across the internet (Cameron, 2005). Competition amongst transaction management institutions should ensure that client preferences for transaction types and transaction reports are met. The provision of accounting services and independent financial advice by transaction managers would act to improve client knowledge of financial transactions and better enable them to understand the risks associated with these transactions. Shiller (2008) cites poor financial advice in the provision of sub-prime mortgages as one of the causes of the housing bubble that led to the financial crisis. He further recommends that financial advice should be offered to individuals through government subsidy. The inclusion of credit reference and rating functions the transaction management institutions, would greatly improve the information upon which the credit worthiness assessments are made. This not only improves client’s knowledge of the other parties to their transactions but also, through the elimination of false negatives, leads to potentially improved investment-credit matches. The tiered structure of investment institutions, and its consequent separation of low risk and high-risk investments enable clients to better select investments, also acts to improve investment-credit matches.

The recommendations of the FSA outlined above can be incorporated into the tiered structure of investment institutions without having such a negative effect on the actual performance of the overall system. Further, the independent transaction management institutions can independently assess the risk profile of investment institutions as part of their credit reference and rating function, reporting to the regulatory authorities.

In contrast to microeconomics, there appears, as of yet, to be no emerging discipline of macro-economic engineering. ((Shiller’s (1993) discussion of the use of macro-markets for managing societies large risks being one possible exception.))

There is no general design knowledge, there are no putative principles to be analyzed. The value the Long and Dowell (1989) work adds here is the address of macro-economic issues within the context of a general design problem.

The simple system architecture solution above is not a complete solution to the current financial crisis. For this to be the case there would need to be elements such as an empirical validation of the problem statement, an implementation plan (for the empirical derivation) that migrated the current financial system to the new architecture, many other issues need to be addressed using the knowledge and practices of multiple disciplines. However the statement of the problem and the solution in this design-focused form does enable, a prior discussion as to the effectiveness of the solution. Just the problem in terms of desired performance opens possibilities of using the knowledge and practices from outside of mainstream economics.

7. Conclusion

Using a generic engineering conception derived from the work of Long and Dowell, and a postulation of the general design problem for economic systems, instances of both microeconomic and macro-economic design have been analyzed.

The general design problem for economic design has been considered in each analysis. Through Roth’s exposition of the redesign of the entry-level labor market for American doctors it was possible to describe the design problem and multiple different economic system solutions in terms of the general problem. The general problem was also used to provide a new perspective upon the redesign of the global financial system in response to the current crisis.

For the Roth work, the analysis against the generic engineering conception illustrated that:

• The implicit conception of an engineering discipline exposed through the Roth’s example of redesign is consistent with the generic engineering conception.
• Although Roth’s conception is not complete with respect to the generic conception, completeness might be sought through consideration of the knowledge and practices of other disciplines.

For the case of the global financial system, it has been illustrated that considering the issues within the context of a design problem, may lead to solutions that would otherwise not be considered. Here, too, given the complex nature of the problem, bringing in practitioners from other disciplines and considering the problem from a design-focused perspective may prove beneficial.

The value offered by the Long and Dowell is the conceptualization of the fundamental distinction between the general problems of scientific and engineering disciplines and the consequent distinction this engenders for knowledge and practice. It may be argued (Wikepedia, 2009) that in the past, a scientific discipline of economics has dealt with emergent problems through Kuhnian style paradigm shifts in macro-economics. The response to the great depression of the 1930s led to the dominance of Keynesianism (Keynes, 1936) This paradigm was in turn overthrown in response to the low growth and stagflation of the 1970s, ushering in an era of Monetarism (Freidman, 1962).

Solving the complex design problems of the economic systems of the 21st century may require more that these 20th century paradigm shifts of a scientific discipline.

Practitioners from different disciplines may need to be drawn together into an engineering discipline of economic design, of the type envisaged for HCI by Long and Dowell. Further as the interactions of human economic agents are increasingly mediated through software economic agents one would expect HCI to play a significant role in such a discipline.

References

Barth, J.R., Brumbaugh Jr., R.D., Wilcox, J.A., 2000. The repeal of glass-steagall and the advent of broad banking. Journal of Economic Perspectives 14 (2), 191–204.

Bass, L., Clements, P., 2007. R. Kazman Software Architecture in Practice. Addison Wesley.

Cameron, K., 2005. The Laws of Identity, The Identity Blog website, December 5, 2005. <http://www.identityblog.com/stories/2005/05/13/TheLawsOfIdentity. pdf>.

Dowell, J., Long, J.B., 1989. Towards a conception of an engineering discipline of human factors. Ergonomics 32 (11), 1513–1535.

Dowell, J., Long, J.B., 1998. Conception of the cognitive engineering design problem. Ergonomics 41 (2), 126–139.

Flanagan, D., 2005. Java in a Nutshell. O’Reilly.

Fowler, M., 2003. UML Distilled: A Brief Guide to the Standard Object ModelingLanguage. Addison-Wesley.

Freidman, M., 1962. Capitalism and Freedom. University of Chicago Press.

Gailbraith, J.K., 2008. The collapse of monetarism and the irrelevance of the New
Monetary consensus,25th Annual Milton Friedman Distinguished Lecture at Marietta College, Marietta, Ohio, March 31, 2008. <http://www.gov.utexas.edu/ papers/CollapseofMonetarismdelivered.pdf>.

Gale, D., Shapley, L., 1962. College admissions and the stability of marriage. American Mathematical Monthly 69, 9–15.
Gastineau, G.L., Kritzman, M.P., 1996. Dictionary of financial risk management. Frank Journal of Fabozzi Associates.

The Geezeo Website, 2009. <http://www.geezeo.com>.

Halligan, L., 2009. Outrage at bonuses won’t solve the mess we’re in, The Daily
Telegraph website, February 16, 2009. <http://www.telegraph.co.uk/finance/ comment/liamhalligan/4623601/Outrage-at-bonuses-wont-solve-the-mess- were-in.html>.

Housel, M., Barker, C., 2009. Who’s More to Blame: Wall Street or the Repealers of the Glass-Steagall Act? Notley Fool Website, April 6, 2009. <http:// www.fool.com/investing/general/2009/04/06/whos-more-to-blame-wall-street- or-the-repealers-of.aspx>.

IBM Website – Rational Unified Process, 2009. <http://www-01.ibm.com/software/ awdtools/rup/?S_TACT=105AGY59&S_CMP=WIKI&ca=dtl-08rupsite>.

IBML Website – Rational Rose Developer for Java, 2009. <http://www.01.ibm.com/ software/awdtools/developer/rose/java/index.html>.

Keynes, J.M., 1936. The General Theory of Employment. Interest, and Money. Kroll, P.,

Kruchten, P., 2003. The Rational Unified Process Made Easy: A Practitioners
Guide to the RUP. Addison-Wesley.

Krugman, P., 2009. The Return of Depression Economics and the Crisis of 2008.
W.W. Norton Company Limited.

Kuhn, T.S., 1970. The Structure of Scientific Revolutions. University of Chicago Press.

Lawvere, F.W., Schanuel, S.H., 1987. Conceptual Mathematics, A First Introduction
to Categories. Cambridge University Press.

Long, J.B., Dowell, J., 1989. Conceptions of the Discipline of HCI: Craft, Applied
Science and Engineering, People and Computers V. Cambridge University Press.

MacLane, S., 1998. Categories for the Working Mathematician, second ed. Springer.

Milgrom, P., 2000. Putting auction theory to work: the simultaneous ascending
auction. Journal of Political Economy 108 (200), pp. 245–272.

Osbourne, M.J., Rubinstein, A., 1999. A Course in Game Theory. MIT Press. The Paypal webs, 2009. <http://www.paypal.com>.

Roth, A.E., 1984. The evolution of the labor market for medical interns and
residents: a case study in game theory. Journal of Political Economy 92, 991–
1016.

Roth, A.E., 2002. The economist as engineer: game theory, experimentation, and
computation as tools for economic design. Econometrica 70 (4), 1341–1378.

Roth, A.E., Peranson, E., 1997. The effects of the change in the NRMP matching
algorithm. Journal of the American Medical Association 278, 729–732.

Salter, I.K., 1993. A framework for formally defining the syntax of visual languages. In: Proceedings of the IEEE Symposium on Visual Languages. IEEE Computer
Society Press, pp. 244–248.

Salter, I.K., 1995. The Design of Formal Languages, PhD Thesis, University College
London.

Shiller, R.J., 1993. Macro Markets: Creating Institutions for Managing Society’s
Largest Economic Risks, Clarendon Lectures in Economics. Oxford University
Press, Oxford.

Shiller, Robert J., 2008. The Subprime Solution. Princeton University Press.

Simon, H., 1957. A behavioral model of rational choice. In: Models of Man, Social
and Rational: Mathematical Essays on Rational Human Behavior in a Social
Setting. Wiley, New York. Stratton, A.

Seager, A., 2008. Darling invokes Keynes as he eases spending rules to
fight recession, The Guardian, October 20th 2008. <http://www.guardian.co.uk/
politics/2008/oct/20/economy-recession-treasury-energy-housing>.

Turner, J.A., 2009. The turner review (a regulatory response to the global banking
crisis). Financial Services Authority (3).

The Wesabe Website, 2000. <http://www.wesabe.com>.

Wikepedia – Paradigm Shift, 2009. <http://www.ikipedia.org/wiki/Paradigm_shift>.

Wikipedia – Financial crisis of 2007–2009. <http://www.en.wikipedia.org/wiki/
Financial_crisis_of_2007%E2%80%932009>.

Wilson, R.B., 2002. Architecture of power markets. Econometrica 70, 1299–1340.

The Zopa Website, 2009. <http://www.zopa.com>.