Table 2 Participants’ responses to knowledge sub-domains

Privacy & confidentiality

 Big Data and AI applications in healthcare may predispose patients’ personal details (e.g., health information) to privacy breaches

114 (24.5%)

169 (36.3%)

183 (39.3%)

 Under no circumstances, an in-house breaching of patients’ data might be necessary

207 (44.4%)

68 (14.6%)

191 (41.0%)

 Big Data and AI applications in healthcare may predispose patient’s data to use by unauthorized personnel

107 (23.0%)

140 (30.0%)

219 (47.0%)

 Patients’ data, embedded within Big Data and AI projects, could be used for alternative processes

74 (15.9%)

126 (27.0%)

266 (57.1%)

 Ethical risks associated with Big Data and AI application in healthcare may be present across all steps of data management (e.g., collection, linking, and implementation)

62 (13.3%)

148 (31.8%)

256 (54.9%)

 Linking data from different sources poses significant and novel ethical challenges

92 (19.7%)

152 (32.6%)

222 (47.6%)

Informed consent

 Designing and/or obtaining consent is an ethical limitation of Big Data and AI projects in healthcare

134 (28.8%)

138 (29.6%)

194 (41.6%)

 Data usage permissions granted by informed consent must be determined by legislative authorities

49 (10.5%)

90 (19.3%)

327 (70.2%)

 Obtaining consent for a broad range of future research projects not foreseen at the time of asking doesn’t qualify as “informed” consent

135 (29.0%)

123 (26.4%)

208 (44.6%)

 The informed consent in Big Data and AI projects in healthcare lack transparency due to inherently complex inner-workings of novel AI algorithms

97 (20.8%)

166 (35.6%)

203 (43.6%)

Ownership

 In general, and in Big Data and AI projects in healthcare in particular, data, even at the individual-level, cannot be owned

92 (19.7%)

143 (30.7%)

231 (49.6%)

 Parties conducting Big Data and AI projects in healthcare should be able to exert a quasi-control of patients’ data, as to market or to refrain from alienating intimate data’s core features, to protect data but also to participate in data-driven endeavors, and to use data for one’s own benefit or the benefit of others

46 (9.9%)

145 (31.1%)

275 (59.0%)

 Under certain circumstances, data generated from Big Data and AI projects in healthcare could be utilized for marketization/commodification

231 (49.6%)

107 (23.0%)

128 (27.5%)

Biases & divides

 Big Data and AI application in healthcare could extend economic inequality

117 (25.1%)

190 (40.8%)

159 (34.1%)

 Big Data and AI application in healthcare could promote health discrimination

166 (35.6%)

140 (30.0%)

160 (34.3%)

 Big Data and AI models in healthcare have the inherent risk of augmenting the biases of their developers or the populations on which they were developed

85 (18.2%)

152 (32.6%)

229 (49.1%)

Epistemology

 The data-driven approach of Big Data and AI algorithms in healthcare is equivalent, and at times superior, to theory-based approaches of conventional scientists

112 (24.0%)

179 (38.4%)

175 (37.6%)

 Big Data and AI application in healthcare is prone to the same errors of traditional research, particularly in the acquisition and pre-processing of data (e.g., checking data consistency)

130 (27.9%)

150 (32.2%)

186 (39.9%)

 Due to our lack of understanding, analytical interpretations of Big Data and AI algorithms in healthcare are essentially “blind” (i.e., lack context for clinical integration)

76 (16.3%)

193 (41.4%)

197 (42.3%)

Accountability

 It is the responsibility of individual researchers to ensure that big data in healthcare is used ethically

59 (12.7%)

92 (19.7%)

315 (67.6%)

 It is the responsibility of institutions to ensure that big data in healthcare is used ethically

32 (6.9%)

75 (16.1%)

359 (77.0%)

 It is the responsibility of legislative and regulatory bodies to ensure that big data in healthcare is used ethically

37 (7.9%)

93 (20.0%)

336 (72.1%)

 Big data and AI application in healthcare might have an impact on the environment

140 (30.0%)

157 (33.7%)

169 (36.3%)