Two central notions in bioethics are “beneficence,” the provision of benefits to the patient, and the absence of “maleficence,” the “do no harm” principle. Certainly in constructing a framework for data ethics, these must also be central goals.
There doesn’t appear to be a perfect mapping of these ideas to statistical practice aimed at describing the attributes of large populations. IRB’s have largely adopted the practice that informed consent requires the participant’s understanding of both costs and benefits of their decision. For statistical information that benefits serve the common good not the individual, see last week’s blog, A Little More on Data Ethics.
Bioethics focuses on interventions (or lack thereof) on individuals. For example, medical procedures should be designed to benefit the individual.
Statistics for the common good seek beneficence for the whole population. Measuring the unemployment rate through self-reports of a sample of persons seeks to inform the citizenry of an indicator of the society’s well-being. Participants in employment surveys are promised that no action on themselves will be taken based on their responses. Statistical uses of data by definition are uninterested in the individual.
Further, not all statistical aggregations of data avoid maleficence for all persons. If the unemployment rate leads policy makers to introduce increased taxes to support the unemployed, then the relative socioeconomic status of employed persons is diminished. They, as a group, are harmed. In general, when data produce statistical information that leads to a reallocation of resources in a society, group harm results.
Thus, when individual data are used for statistical purposes, the interplay of individual versus group beneficence and maleficence is different than for the biomedical case.
When we move our attention to unobtrusive use of personal data, another concept seems useful for data ethics – instrumentalization. This is the use of other humans merely as a means to some other goal. One might consider this notion as relevant to informed consent. If researchers do not reveal their goals to the participant, the individual is not fully able to weigh costs and benefits of participation. Consent is then not “informed.” For example, users are routinely informed that their mobile phone positional data are used to provide traffic estimates. I have the right to refuse participation if I feel that I am merely being used as an instrument for a purpose that does not benefit me. Informed consent, in that sense, is the protection against instrumentalization in the researcher-participant relationship.
As we think about how to build a sustainable environment for common good uses of large digital data resources, we must address our ethical obligations to those providing the data. I suspect we’ll be talking about beneficence, maleficence, and instrumentalization in the future.