What's the good of a science platform?

1. Introduction

Science may be understood as a process of knowledge generation. A science platform is an infrastructure designed to generate knowledge cost-effectively. Platforms confer advantage through increased scientific opportunity, the standardization of best practice and economies of scale. Although platforms typically involve high ‘front-loaded’ financial investment, marginal costs ‘per datum’ reduce as the volume of data increases, and the benefit–cost ratio ‘per research question’ increases as the data are exploited. In these ways, platforms enable science at scale.

The efficiency of a system can only be judged in the context of a common purpose. Here, that purpose is to facilitate science (the generation of knowledge) through knowledge diversification, knowledge sharing and deepening interdependence. This goal would be well served by conditions that encourage the emergence of a highly collaborative, mutually stimulating and self-organizing scientific community.

2. The challenge of platform science

Platforms may offer potential benefits, but platforms also generate costs. These include the management of increasingly complex scientific activity, which is reflected in the structure of the science economy. This challenge may be specified more closely in terms of the ability of the science economy to reflect the value of scientific activity in platform science, i.e. that the differentiation of the science economy is commensurate with that of the activity which it supports.

Science environments may be understood as knowledge-based communities in which there is an ongoing process of knowledge diversification, technological advance and deepening interdependence. Science communities, however, are privileged in that they enjoy a degree of protection from the wider economic pressures of society. Although this protection has value, providing the relatively stable environment required for success, it also protects the science community from the creative process of having to adapt the science economy to reflect the activity of the science community.

3. Framing the problem

The underlying problem may be framed as how order emerges in dynamic social environments that grow in complexity [1]. This problem is approached from a value exchange perspective [2] in which:

1. Individual scientists and research organizations collaborate to benefit from value exchange opportunities, where value exchange is defined as giving recognition for the benefit realized.

2. Order emerges as individuals and research organizations seek to reduce the cost of obtaining benefit, i.e. the transaction cost.

3. Institutions, defined as systems of rules and sanctions for the exchange of value, emerge to reduce transaction costs by defining property rights, i.e. the right to benefit from a resource by use, income or disposal.

4. Unclear property rights constrain value exchange opportunities, hindering scientific activity.

(a). Value exchange

In the science economy, value stems from scientific benefit, but value cascades differently according to stakeholder group. In a data access context, value is created for investigators through scientific and technical opportunity. For research funders, it is recognition of multiple benefits from the initial investment. For research organizations and infrastructures it is increased profile through utilization. For all, it is being a valued member of the scientific community. An efficient value exchange mechanism is one which enables all of these stakeholder groups to receive recognition for the value provided. An efficient exchange mechanism presupposes that the science economy is sufficiently differentiated to reflect each of these contributions proportionately and rapidly.

(b). Emerging order

Order may be understood as the coordination of resources, presumably for benefit. Within the biomedical sciences, order emerges largely through a process of constrained spontaneity in that, within a constraining context provided by funding agencies and research governance, there is freedom for self-organization, leading to consensus regarding best practice. Although the balance of constraint and self-organization varies, the common purpose of generating knowledge acts as a guide to efficient resource coordination, i.e. the minimization of transaction costs.

(c). Institutions

Institutions exist to assign property rights. Within the science economy these rights are expressed in terms of who may control or claim credit for an activity. In the science economy the two main institutions for exchanging value are resource ownership (knowledge, technology, data, etc.) and academic attribution (authorship, acknowledgement). Resource ownership is used to facilitate or restrict resource access through transaction costs (governance constraints, access fees, etc.). Academic attribution is typically expressed in terms of authorship or acknowledgement. By convention, authorship typically provides value for investigators by recognizing intellectual activity, while acknowledgement typically provides value for infrastructure and funders by reflecting technical or management activity. Over time best practice becomes adopted as a convention, and conventions become formally expressed as institutions. For example, the ‘understood’ conventions surrounding academic authorship have become explicit rules in the criteria for authorship specified by the International Committee of Medical Journal Editors. As technologies change to improve the efficiency of scientific activity, so institutions that do not change lose efficiency in value exchange.

(d). Property rights

Property rights exist to facilitate value exchange, and incidentally reduce conflict. Property rights evolve (differentiate or combine) to reflect changes in scientific activity. Where property rights are unclear, i.e. do not reflect scientific activity, there is no incentive for value exchange; knowledge is not exploited and the science economy effectively stalls. For example, if an infrastructure is unable to obtain incremental value from increased use, there is no incentive towards improving access to investigators. Similarly, if there is no evidence of infrastructure utility, there is no incentive to funders for continued support. This chain of interdependent value exchange is particularly important for data collection infrastructures, as these provide the basic currency on which all else is built. From the perspective of knowledge generation, restrictive property rights, for example to obtain pecuniary or professional advantage, ultimately fail.

4. Science platforms as a public good

To provide context for the assignment of appropriate property rights, a view is required on the type of resource in question. In this case, the type of resource that a science platform is considered to provide.

In the UK, the academic science economy operates at two levels. For funding and career progression it operates as a common good, in that ‘anyone can play’, but there is competition for resources. This model serves well to encourage high-quality science, but tends to focus resources among relatively few individuals and may be self-limiting in terms of overall scientific output.

For knowledge dissemination, however, it operates (aspirationally at least) as a public good, in that ‘everyone can play’ as resources (findings) are to be shared without restriction or cost. This model serves well in facilitating the sharing of data and best practice. Of course, scientific resources are not public goods in the purist sense of the term as there will always be a transaction cost. However, by sufficiently reducing the cost and absorbing it elsewhere, the resource can be effectively a public good at point of use. An example is the National Institutes of Health PubMed resource, which is operated as a public good on behalf of the wider scientific community.

The tension between models appears to serve the contrasting goals of encouraging innovation through competition, and exploitation through collaboration. The extent to which these models are most efficient at achieving their respective goals is moot [3]. Prior to scientific platforms the opportunity for large-scale infrastructures and collaborations to generate innovation was extremely limited. It may be that innovation and exploitation are best generated in a scientific community where ‘everyone can play’.

An argument can be made that the public good model is implicit in all platform activity, including funding and career progression. The platform environment facilitates information sharing and self-organization, and generally lowers institutional (and institutionalized) barriers to collaboration, thus encouraging information-rich and scientifically diverse research communities. Scientific output is highly interdependent and multi-disciplinary, resulting from the free sharing of technologies, knowledge and ideas. A public good model also offers a framework for assessing ‘fairness’ in academic attribution as it reduces the economic and political incentives to restrict attribution, and increases the incentives to reward purely on the basis of merit. Arguments against a public good approach could be made. It is difficult to deny, however, that science flourishes with the efficient exchange of information and that scientists flourish in environments that facilitate rigorous work in pursuit of intellectually satisfying ends. It is a reasonable view that these criteria will be more fully met from within a public good model. Incidentally, the advantages of a thoroughgoing collective approach to science were identified early in the scientific enterprise. Francis Bacon's ‘House of Salomon’, in which science was conceived as a team exercise involving the coordination of specialisms, was an inspiration to the foundation of the Royal Society.

5. Differentiating the science economy

For there to be an efficient exchange of value, the differentiation of an economy needs to be commensurate with the activity which it supports. In the science economy an efficient value exchange mechanism is one which allows investigators, research organizations, infrastructures and funders to receive recognition for their contribution proportionately and rapidly.

An example of efficient value exchange is the acknowledgement of grant funding. Typically, this occurs at the end of an article in the form of a funder's name and grant reference number. This concise and precise formula satisfies the needs of funders for their contribution to be recognized, the needs of researchers to demonstrate funding success and the needs of publishers to use column inches carefully. That the funding acknowledgement has become ‘standardized’ indicates that it works, i.e. value has been efficiently transferred for all stakeholders. There is no reason why this system should not continue to provide efficient value exchange in a platform context. This example illustrates the benefit of clearly defined property rights. That the funding agency provided the resources required to conduct the study is not disputed. That this benefit should be recognized is also agreed. The critical point is that the value of the project to the funders and to the investigators is differentiated and may be rewarded separately and with no additional transaction cost, i.e. one stakeholder may be rewarded without any cost to another stakeholder. Of course, multiple benefits between stakeholders are welcome.

An example of value exchange in transition is authorship [4]. In biomedical science, the conventions surrounding authorship attribute high value to the recognition of principal investigatorship (typically the last author) and senior authorship (typically the first author). This practice of exclusive authorship has evolved in the context of small research teams and, within that context, may be presumed to exchange value reasonably efficiently. It becomes less elegant, however, in addressing the needs of consortium-based research where many research teams have contributed to the dataset under analysis. In these circumstances, collective authorship is frequently adopted, with authors listed alphabetically or haphazardly, and a writing team identified. Collective authorship is a blunt instrument offering limited opportunity for value exchange. For example, it is uninformative to research organizations for identifying specialist skills for recruitment and career progression purposes, and of little value to individuals managing a specialist career track. This is an example of under-differentiation, i.e. the economy does not have the means for fully extracting that benefit available from the activity. Arguably, collective authorship recognizes the value of the underlying collective activity, a kind of joint-value, but this may be considered a poor substitute for extracting multiple and diverse benefits. This problem has been recognized and more nuanced approaches are evolving. One approach is badged authorship, which identifies specific contributions to the published work, provides a convenient means for discovering who did what and is not constrained by the number of authors. The challenge with badged authorship is establishing a meaningful taxonomy of ‘badges’ that is acceptable to both scientists and journals, i.e. can be used as a community standard. A proposal that is being promoted by a slowly increasing number of journals comes from the digital taxonomies project [5]. Fourteen categories of contribution (badges) are proposed covering a wide range of activity from study conception, through data curation to formal analysis and manuscript preparation.

An example of value exchange failure is infrastructure acknowledgement. Infrastructure is haphazardly and idiosyncratically acknowledged, and the acknowledgement is rarely scientifically informative, diminishing its ability to exchange value. For example, a data collection, from which a specific dataset has been drawn, may be alluded to, but metadata allowing the specific dataset or sample set used for the analysis and its provenance to be unambiguously identified by other researchers is rarely available, let alone explicit. This is largely a legacy issue from data collections being led by a principal investigator, and the principal investigator's acknowledgement of the underlying infrastructure. Moving from this ‘cottage-industry’ model to platform science requires a differentiation of property rights commensurate with the complexity of creating and sustaining the wider infrastructure. For an infrastructure acknowledgement convention to be widely adopted it needs to give value to investigators (users), funders (payers) and the support scientists (builders and maintainers), in ways that are scientifically informative, career relevant and convenient. A potential and relatively simple solution would be providing a standardized digital object identifier (doi) for specific data instances as a mandatory component in the acknowledgements section. Just as the doi has been used to disambiguate scientific publications, it could be used to disambiguate datasets, giving information on origin, curation, pre-processing and instancing. For investigators this allows rapid and precise identification of datasets used for specific analysis, enabling replication and further analysis. For funders (and research institutions), as the doi can be linked to specific grants, projects and individuals it can be used as a tool for automated impact evaluation and audit as well as recruitment and career progression decision-making. For data collectors and support scientists, an infrastructure doi provides visibility, enabling value to be assessed and to incentivize the development and adoption of best practice, e.g. the upgrading of the management of data collections to a common standard and more intuitive database navigation. For science, in general, these outcomes for each stakeholder group are desirable as they improve the transparency of increasingly complex data management analysis and improve the user experience, each of these applying downward pressure to transaction costs.

6. Future directions

Science platforms, though not replacing investigator-led research groups, increase capacity and opportunity by accelerating knowledge diversification, knowledge sharing and interdependence. However, it will require a collective effort by the entire science community for this potential to be realized. Value structures of resource owners need to be revisited to incentivize the reduction of transaction costs to a level at which they are absorbable prior to the point of use. As this is challenging for many reasons, a process of incremental and opportunistic change might be anticipated. Nevertheless, it is not difficult to suggest a direction of travel:

1. Fundamentally, the value of the science community to wider society lies in knowledge generation. To facilitate this and avoid confusion with secondary, but often rivalrous, principles such as maximizing research income and maximizing organizational prestige, knowledge itself (including publicly funded datasets) could be recognized as a public good. This would stimulate the process of developing best practice for data, sample and technology sharing around cost neutral or cost minimal solutions.

2. Institutions are required to give recognition through appropriate and clearly defined property rights to infrastructure contributors (data collectors, infrastructure builders and maintainers). To deliver a growing and sustainable science economy these institutions must be capable of exchanging sufficient value to incentivize further data collection and infrastructure build, for example by enabling incremental value exchange for greater data and infrastructure use. The adoption of badged authorship and infrastructure doi acknowledgements are examples of institutions designed to reward infrastructure contributors.

3. Efficiency metrics might be used to reward the adoption of low transaction cost practice. For example, research groups of all sizes might be incentivized to outsource routine informatics and biosample storage to specialist and more cost-efficient facilities through an efficiency premium attached to research funding.

In conclusion, the main asset of science is the scientist. To create an environment which attracts, stimulates and rewards our best intellects is a reasonable heuristic for successful knowledge generation. However, the fundamental challenge that will underpin the development of this environment, the fruitfulness of its inhabiting community and the benefit obtained from science platforms will be the desire of the community itself to create a fair economy—one that facilitates the exchange of value across the community to the benefit of all. Curiously, this is not a structural, technical or even governance issue, it is moral.

Competing interests

I am Director of the MRC Dementias Platform UK.

Funding

The MRC Dementias Platform UK is supported by MRC grant MR/L023784/2.