Representational Veracity Domains in Stakeholder Reasoning about iPSC Derivation from Biobanked Samples

We recently derived representational veracity (RV) as an upstream condition for the governance of data science health research: that banked data, the models derived from them and the population descriptors that travel with them continue to represent their sources truthfully at the point of use. The derivation was normative and specified for data; we asked whether the RV domains correspond to anything stakeholders in a real reuse decision recognize, and whether they apply to biological samples.

We conducted a two-stage secondary analysis of the KACIY study, which examined reuse of banked HapMap YRI samples for induced pluripotent stem cell (iPSC) research, using interviews with six community advisory board (CAB) members and five North America-based bioethicists, and two focus groups with twenty community members. We adopted two thematic architectures generated independently before RV existed, then asked how far each corresponds with the four RV domains, returning to the transcripts where needed. An external auditor reviewed the analysis, reading the transcripts before seeing the framework or our conclusions, under responses we specified in advance.

Normative authorization and relational community were organizing themes in both architectures and all three groups. Material provenance had a counterpart in the community data only as the legibility of reports returned to the community. Informational descriptors had no counterpart in any community-facing document. No descriptor question had been put to any participant; the fullest descriptor testimony came unprompted from the repository steward. The auditor found descriptor references by five of five bioethicists under a broad definition, three of five under a narrow one, and none by community members.

The RV domains are present, unevenly, in stakeholders' reasoning about this reuse decision. The domains concerning persons and a collective were raised by the community that authorized the use; the domains concerning the artifact as it travels largely were not. The audit supported two propositions: that the domains are not equally answerable from every position in the chain of custody and use, and that an RV review should draw on several groups of informants. The reason for the difference was not established, and utility remains to be shown prospectively.