Title : Informed consent in the age of brain data: What people don't know can hurt them
Abstract:
The current neuroscience studies have produced more than ever before, in terms of quantity of brain data, due to the avenues of neuroimaging used such as fMRI, EEG, MEG, and the new brain-computer interfaces. The traditional informed consent models that were developed many decades ago to consider the less complex paradigms of biomedical research do not address the specific vulnerabilities of brain data collection, long-term storage, and extensive sharing models in particular. The presentation discusses in detail the key loopholes that exist in the existing consent protocols resulting in the exposure of research subjects to privacy breaches, possible discrimination, and commercial exploitation against their own obliviousness and substantial consent.
Brain data is unique and the methods of anonymization used in the conventional method cannot be applied to it. Recent studies show effective re-identification of people who previously formed part of allegedly anonymized neuroimaging datasets of unique structural patterning and functional activity patterns. More complex artificial intelligence algorithms can now (extract) brain fingerprints with roughly the same accuracy as a facial recognition system. In addition to the vulnerability of re-identification, brain imaging unintentionally reveals sensitive personal data that the subjects never meant to share with the world, such as genetic inclinations to neurological illnesses, cognitive capacity estimates, mental condition indicators, personality features, and the future course of disease.
The talk discusses the 3 critical areas in which the existing practice of consent is broken: first, the myth about the impossibility of permanently anonymizing brain data in the face of the rapidly expanding computational capacities; second, the problem of unintentional disclosure with individual brain scans easily providing much more information than the small research question that the participants were informed they were answering and paid no compensation to do so; third, the problem of commercial misappropriation, where academic research data is leaked into corporate product development through data sharing requirements without the knowledge or consent of the participants. The proposed practical solutions are updated consent wording that represents the present AI capabilities, tiered consent schemes that provide a real choice of participants, dynamic consent engines that allow controlling participation, and an institutional governance overhaul. These solutions bring the consent practices into conformity with the technological facts without violating the autonomy of the participants and safeguarding the most personal data that humans can share information that has been taken straight out of their brains.
