Data divide as the fourth level: An extended digital divide model

The rapid advancement of AI technology and its applications in healthcare have triggered a heated debate on whether the accelerated e-health applications would narrow existing health disparity. The conventional three-level digital divide framework becomes inadequate to address this debate because it has not considered digital data representation, an emerging dividing force. This study first proposes an extended digital divide model and we argue that the data divide has become the fourth level of divide shaping the health disparity. Based on the survey data of 1006 respondents collected between 20 and 27 October 2023 in Hong Kong, we investigate the magnitude of four levels of digital divide across various socioeconomic groups in the context of e-health applications.
Regarding the digital divide as “differences between individuals, households, companies, or regions, related to the access and usage of ICT” (OECD, 2001), the early studies emphasize the physical ICT connectivity, the first level of the digital divide, and treats the digital divide as a binary divide between “haves” and “have-nots” (van Dijk, 2006). Recent research has gone “beyond access” by paying more attention to the second level of the digital divide: the skill and knowledge gap among ICT users. Researchers strive to understand the reasons for the differences in the spectrum of “knows vs. know-nots” and “uses vs. use-nots,” as well as the social, cultural, and psychological background of the usage gap (Büchi, Festic, & Latzer, 2018; DiMaggio & Hargittai, 2001; Hargittai, 2010). Current studies on the third level divide focus on the disparity in the ability to translate digital resources to achieve specific offline outcomes (Ragendda, 2017; Stern et al., 2009; van Deursen & Van Dijk, 2014; van Deursen & Helsper, 2015). However, many of today's AI-enabled e-health applications are built using training data from established digital systems. Those having no digital footprints would have been excluded from the design of these applications. Thus, we propose an extended digital divide model to address the data divide between digitally active and idle as the fourth level divide.
Among all 1006 respondents, 53.7% were under 45 years old, and 23.16% were above 65. 49.85% of respondents were female, and 18.63% obtained a graduate or above degree. As for the domestic family income, 10.14% of respondents reported less than 20000 HKD per month, while 16.7% reported over 80000 HKD per month. The preliminary results based on this sample present three consistent patterns. 1) Across all SES groups, the know-how divide (the second level) is consistently the most salient, followed by the data divide (the fourth), the transformative capacity (the third level), and then the access divide (the first level). 2) It is unsurprising that these four levels are interdependent, with mild correlations ranging from 0.41 to 0.60. 3) Compared with other Page 53 levels of the divide, the data divide is most salient across generations and education levels. In contrast, the access/usage divide is most significant between the rich and the poor, suggesting complex cross-cutting patterns of the digital divide.