The Adoption of Artificial Intelligence in Health Care and Social Services in Australia: Findings From a Methodologically Innovative National Survey of Values and Attitudes (the AVA-AI Study)

Abstract

Background Artificial intelligence (AI) for use in health care and social services is rapidly developing, but this has significant ethical, legal, and social implications. Theoretical and conceptual research in AI ethics needs to be complemented with empirical research to understand the values and judgments of members of the public, who will be the ultimate recipients of AI-enabled services. Objective The aim of the Australian Values and Attitudes on AI (AVA-AI) study was to assess and compare Australians’ general and particular judgments regarding the use of AI, compare Australians’ judgments regarding different health care and social service applications of AI, and determine the attributes of health care and social service AI systems that Australians consider most important. Methods We conducted a survey of the Australian population using an innovative sampling and weighting methodology involving 2 sample components: one from an omnibus survey using a sample selected using scientific probability sampling methods and one from a nonprobability-sampled web-based panel. The web-based panel sample was calibrated to the omnibus survey sample using behavioral, lifestyle, and sociodemographic variables. Univariate and bivariate analyses were performed. Results We included weighted responses from 1950 Australians in the web-based panel along with a further 2498 responses from the omnibus survey for a subset of questions. Both weighted samples were sociodemographically well spread. An estimated 60% of Australians support the development of AI in general but, in specific health care scenarios, this diminishes to between 27% and 43% and, for social service scenarios, between 31% and 39%. Although all ethical and social dimensions of AI presented were rated as important, accuracy was consistently the most important and reducing costs the least important. Speed was also consistently lower in importance. In total, 4 in 5 Australians valued continued human contact and discretion in service provision more than any speed, accuracy, or convenience that AI systems might provide. Conclusions The ethical and social dimensions of AI systems matter to Australians. Most think AI systems should augment rather than replace humans in the provision of both health care and social services. Although expressing broad support for AI, people made finely tuned judgments about the acceptability of particular AI applications with different potential benefits and downsides. Further qualitative research is needed to understand the reasons underpinning these judgments. The participation of ethicists, social scientists, and the public can help guide AI development and implementation, particularly in sensitive and value-laden domains such as health care and social services.

Practical, epistemic and normative implications of algorithmic bias in healthcare artificial intelligence: a qualitative study of multidisciplinary expert perspectives

“…We found that some stakeholders had the view that bias does not exist in healthcare AI, based on the epistemic belief that these knowledge systems are not biased. People taking this view are incorrect as a matter of fact: algorithmic bias has been demonstrated across domains including healthcare, 33 social welfare, 34 legal systems 35 and finance. 36 In healthcare, studies continue to demonstrate the persistence of racial disparities in diagnostic imaging that predates AI applications.…”

Measures of socioeconomic advantage are not independent predictors of support for healthcare AI: subgroup analysis of a national Australian survey

“…We used the shortened version of the questionnaire as a reference survey to produce weights for the non-probability sample that account for characteristics that influence people's propensity to participate in the online panel. A more detailed description of the data collection and weighting methodology is provided in Isbanner et al [14]. For this analysis, we report on results from the weighted non-probability sample using data obtained from the full questionnaire.…”

Centring equity in data‐driven public health: a call for guiding principles to support the equitable design and outcomes of Australia's data integration systems

“…Others are analysing linked data to improve evidence‐based policy around the social and health outcomes of early childhood experience, which carries long term health implications 6 . Other applications include the monitoring and improvement of prescription practices, 7 and the development of artificial intelligence (AI)‐supported medical technologies 8 . In addition to universities and learned academies, the benefits of data integration are widely recognised across Australian Government agencies, so that considerable investment and reform have been underway to strengthen data integration capacities to support research and innovation, create efficiencies and private–public partnerships, and accelerate data‐driven decision making within policy and routine modes of government 9…”

“…In countries that are further advanced in data integration, such as the United Kingdom, private sector access to health data has provoked public controversy, indicating that Australia should be proactive in developing strategies to regulate and share data from and with the private sector in ways that allow innovation while also aligning with broadly shared visions of the public good 12,13 . The imperative to keep human decision making visible, and data systems transparent and accountable, is a major concern globally 8,10,13‐15 . Many agree that public confidence in the secondary uses of health data requires a social license 12,13,15 .…”

“…Conversations about the overarching purpose of data systems need to take place in a public forum 14,15,17‐19 . In addition to epidemiologists, data analysts, consumer groups and legal experts, a burgeoning number of researchers in Australia are examining the social implications of AI, automated decision making, big data, and the regulation of technology, work which can support the equitable design of our data systems 8,11,12 . The Indigenous Data Sovereignty movement, disability rights activists, and other data‐related movements have vital expertise to bring to this conversation and to the design of equitable data systems 11,19…”