Getting started with rfair

rfair assesses how well a research data object satisfies the FAIR principles (Findable, Accessible, Interoperable, Reusable), entirely in R. It is a native port of the F-UJI metrics, so it needs no external assessment server.

library(rfair)

Assessing an object

Pass any DOI, persistent identifier, or URL to assess_fair(). It resolves the identifier, harvests metadata, and scores it against the FAIRsFAIR metrics.

a <- assess_fair("https://doi.org/10.5281/zenodo.8347772")
a

The returned fair_assessment object prints an F/A/I/R summary. The numbers come from up to 17 metrics; each is one row of:

as.data.frame(a)

summary(a) gives the per-principle score table, and the maturity column reports a 0–3 CMMI level (incomplete → advanced).

summary(a)

Interpreting beyond the score

Automated FAIR scores have well-known blind spots. rfair surfaces three:

# A license being *present* does not mean the data is open for reuse.
a$reuse                    # per-license: open / restrictive, commercial, derivatives

# Restricted access can be legitimate (e.g. sensitive human data) and should not
# be read as "not FAIR".
a$access                   # access level, controlled_access, sensitive

# Identifiers should follow best practices.
a$identifier_hygiene       # layered / non-persistent identifier warnings

You can call these directly too:

license_reuse("https://creativecommons.org/licenses/by-nc-nd/4.0/")
identifier_hygiene("RRID:MGI:5577054")
fair_principles()          # canonical FAIR principle definitions

Exporting results

as_fuji_json(a)            # F-UJI-compatible FAIRResults JSON
as_rdf(a)                  # W3C DQV + schema.org Rating (JSON-LD)

Interactive use

launch_rfair()             # Shiny app

A no-install browser version is at https://choxos.github.io/rfair/app/; because browsers cannot fetch landing pages cross-origin, it scores from registry metadata (DataCite/Crossref) only, so some metrics are lower than the R engine.