3.4 The FAIR Principles are neither rules nor standards

The FAIR Principles should not be seen as rules or standards to be implemented in a strictly formalised way. Rather, they are intended as guidelines that should be followed when handling research data at every stage of the data life cycle. While the FAIR guiding principles are naturally interconnected, they are technically independent of one another and can be implemented incrementally in any combination. This means that the entry-level for data producers can be relatively low, which is often the case when dealing with data that was created a long time ago. As part of data quality management, the goal should be to gradually increase the level of compliance with each of the FAIR criteria.

Adopting the FAIR Principles will often involve a gradual adjustment of workflows, but it can also occur during a more significant shift, such as when one type of infrastructure is replaced by another.

The purpose of this document is therefore not to propose a specific technological implementation of the FAIR Principles, but rather to describe the characteristics that data resources, tools, and infrastructures should have in order to be considered FAIR. The FAIRness of data can be achieved with a wide range of technologies and procedures.

The Data FAIRport Initiative has developed a tiered model for assessing the FAIRness of digital objects, which can serve as a guideline for the gradual and systematic enhancement of the FAIRness of data collections. The FAIR requirements can be evaluated and implemented differently for each of the three components of the digital object. For example, an image file may be subject to different access restrictions than its associated metadata.

Digital objects can contain different FAIR levels:

1. Metadata is assigned to the bit sequence, but the digital object does not have a persistent identifier. The metadata is not, or only partially, machine-readable, which also limits the machine processability of the bit sequence (gray).
2. Each digital object has a PID. However, the metadata has not yet been optimised for machine processing, so the digital object is primarily usable by humans. It cannot yet be considered FAIR for reuse.
3. Each digital object has comprehensive, machine-readable FAIR metadata (green). However, important information about the bit sequence is still missing, such as licensing, data provenance, or the structure of complex digital objects.
4. The bit sequences in the digital objects are also technically FAIR, but not freely accessible and not fully reusable, due to legal restrictions or because they are proprietary data (red). Since the information about these restrictions is machine-readable, this digital object is considered FAIR.
5. Both the metadata and the bit sequences themselves are fully FAIR. They are under a clearly defined open license, allowing for comprehensive reuse of this FAIR digital object.
6. In addition, the digital objects are functionally linked with data sets that are related to them and can be integrated into analysis using semantic technologies (green background). Data is available as RDF statements, granular assertions about the resource in the form of triples (subject-predicate-object). It is also ensured that data provenance information is accessible for machine-based analysis methods.

If the digital object is provided according to the FAIR Principles, it ensures, as a FAIR digital object (FDO), the reliable interpretation and processing of the data represented by the object for both humans and computers. FDOs are also the foundation for a future FAIR Digital Object Framework (FDOF), which will significantly improve the machine-processability of FAIR data. This framework aims to define predictable identifier resolution behaviour, a mechanism for retrieving an object's metadata, and a system for object typification.