Beyond F-UJI: reuse, sensitivity, hygiene, and FAIR-TLC
In rfair: Assess the FAIRness of Research Data Objects and Software

knitr::opts_chunk$set(collapse = TRUE, comment = "#>")
library(rfair)

Automated FAIR tools have well-documented blind spots. In peer review of a COVID-19 FAIR-assessment study, the reviewer (Melissa Haendel) noted that such tools reward the presence of a license, an identifier, or a metadata field without checking whether the data is actually reusable, legitimately restricted, or properly identified. rfair adds checks for exactly these.

A license can be present yet not open for reuse

Detecting that a license exists says nothing about whether you may reuse the data. license_reuse() classifies the actual permissions, and maps each license to the six-category taxonomy of the (Re)usable Data Project (Carbon et al. 2019).

license_reuse("https://creativecommons.org/licenses/by/4.0/")[c("category", "rdp_category", "facilitates_reuse")]
license_reuse("https://creativecommons.org/licenses/by-nc-nd/4.0/")[c("category", "rdp_category", "facilitates_reuse")]

Only permissive licenses facilitate reuse without negotiation; CC-BY-NC-ND is present and standard, yet restrictive.

Controlled-access and sensitive data is not a FAIR failure

Data behind a data-use agreement (e.g. human/clinical data) is legitimately restricted; it should be judged on metadata richness, not open download. classify_access() flags this, drawing on the (Re)usable Data Project curations.

classify_access(access_level = "closedAccess",
                urls = "https://www.ncbi.nlm.nih.gov/gap/?term=phs000424")[c("access", "controlled_access", "sensitive")]

Identifier hygiene

Layered identifiers (an identifier minted on top of another) and non-persistent identifiers reduce interoperability.

identifier_hygiene("RRID:MGI:5577054")$issues
identifier_hygiene("https://doi.org/10.5281/zenodo.8347772")$hygiene_ok

FAIR-TLC: Traceable, Licensed, Connected

The reviewer's own framework extends FAIR with three principles (Haendel et al., FAIR+): data should be Traceable (provenance, attribution), Licensed (clearly and reusably), and Connected (qualified links to related entities). fair_tlc() computes these from an assessment.

a <- assess_fair("https://doi.org/10.5281/zenodo.8347772")
fair_tlc(a)
#>   dimension                             indicator   met
#> 1 Traceable                         T1 Provenance  TRUE
#> 2 Traceable                        T2 Attribution  TRUE
#> 3  Licensed L1 Documented & minimally restrictive  TRUE
#> 4  Licensed           L2 Flowthrough transparency  TRUE
#> 5 Connected                      C1 Connectedness  TRUE