Leiden Declaration on FAIR Digital Objects

Leiden Declaration on FAIR Digital Objects

"Acknowledging that, the innovations brought by the current Internet have great benefits for society, but also come with inherent challenges such as the nearly uncontrolled spread of misinformation and violations of privacy, we affirm that:"

• There is an imperative to harness the benefits and mitigate the unwanted side effects of the Internet and the Web.
• Once machines enter the equation as ‘users’ of the internet and increasingly independently from human intervention, the benefits as well as the challenges will be further aggravated.
• The development of combined use of data and services by humans and machines (‘Hybrid Intelligence’) and by increasingly powerful machine learning algorithms operating independently (‘Artificial Intelligence’) requires urgent action and decision making regarding the proper and controllable reuse of data, and information, with built-in ‘privacy by design’.
• Transparency, scrutiny and trust in data and the associated processing (more and more by machines) is essential if we are to avoid a future web that will be increasingly polluted with unreliable information and dominated by the use of algorithms without clear, explicit and comprehensible origins and operations. 

• Current society is increasingly data-driven and faces unprecedented global challenges, reflected for instance in the SDGs [https://www.un.org/sustainabledevelopment/]
• In order to benefit society and our planet in general, access to the underlying data and infrastructure should be global in nature, as open as possible and as equitable as possible.
• This implies optimal interoperability of data and the associated infrastructure across domain, geographical and cultural borders.

In response to these challenges, we contend that:

• The future of research and innovation will be dependent on effective machine-assisted methods.  
• For machines to assist people in the research and innovation process we need iterative human-machine interaction, here referred to as ‘Hybrid Intelligence’.
• For machines to optimally serve this process, data and the accompanying software and infrastructure should enable machines to independently Find, Access, (Inter-)operate on, and Reuse (and repurpose) data and the accompanying workflows and services. 
• Data formats should enable transfer over the internet as FAIR, machine-interpretable and self-explanatory units of information/knowledge (the FAIR digital object concept).  
• Data integrity should be protected to avoid data corruption and context/provenance degradation.
• For humans to retain trust and endorse the benefits of machine assistance, all machine actions should be properly logged, and explainable at request, allowing full transparency and traceability (provenance). 
• Adequate infrastructures are essential in reaching the full potential of available data and services. 
• FAIR digital objects on the Internet, including the Web, should ideally be actionable by machines alleviating unnecessary burden of human intervention.
• The use of the desired commonly usable infrastructure, provisionally called the web of FAIR data and services, therefore needs to be built around machine actionable units of information. 
• Data and associated services (protocols, software, algorithms) should be as open as possible, and only as restricted as necessary.
• FAIR digital objects should not only be unambiguously identified, but also convey to the consumer their accessibility conditions.
• FAIR digital objects can contribute to scalable, equitable and transparent access to information and knowledge. 
• Standards facilitating global implementation of FAIR digital objects should be developed as open and publicly accessible resources.

Therefore, we commit to:

• Support the FAIR guiding principles to be applied (ultimately) to each digital object in a web of FAIR data and services;  
• Support open standards and protocols;
• Support data and services to be as open as possible, and only as restricted as necessary;
• Support distributed solutions where useful to achieve robustness and scalability, but recognise the need for centralised approaches where necessary;
• Support the restriction of standards and protocols to the absolute minimum;
• Support freedom to operate wherever possible;
• Help to avoid monopolies and provider lock-in wherever possible.

We acknowledge the FAIR digital object concept is a set of behaviours and properties as described above and should serve as the basis for the FAIR publication of digital entities, encouraging the community to design and implement protocols and specifications that enable these FAIR features, irrespective of the technology used.