Artificial intelligence (AI) advancements have begun to transform social interaction and business operations. These advancements have impacted diverse business functions, including financial fraud detection, risk profiling in healthcare, decision-making, and advertising. Organizations have increasingly invested financial and human resources in AI-related initiatives.  

While digital innovations have delivered numerous benefits to society, they also result in unintentional, adverse effects—especially when these innovations result in inequity that separates those with access to the technology. 

Rogers (1962) early conceptualized the digital divide by recognizing a digital gap between users and potential users. However, Rogers’s theory highlights only the impact of user requirements on ICT access and use. It does not include other factors such as individual attributes (e.g., demographics), technical conditions, and social environments. 

Researchers and practitioners have designed AI-enabled artefacts to facilitate various business applications and operations, such as search advertising, copycat detection, risk profiling in chronic care, dynamic decision-making, financial fraud detection, customer social network analysis, knowledge management, user emotion recognition, and user performance prediction. 

Modeling the AI Divide 

AI enables a program or a machine to complete tasks that a human normally performs, such as planning, reasoning, problem-solving, and even acting. Russell and Norvig (2016) identified four types of AI: thinking humanly, thinking rationally, acting humanly, and acting rationally. Cognitive scientists have used psychology theories to imbue AI with “humanness”, while computer scientists and mathematicians have emphasized AI’s logical and unemotional “rationality.” Rationality, the capability to produce ideal solutions, and humanness, the extent to which technology mimics humans, represent two sides of the same coin.  

AI techniques are vast. Some frequently used techniques include natural language processing, knowledge representation, automated reasoning, machine learning, computer vision, and robotics. We can categorize AI innovations as visible and invisible. When users can readily recognize AI’s presence, we refer to that innovation as visible AI. Visible AI innovations have discernible outcomes and have a “user-invisible” side.  

Components to look into 

According to “A Comprehensive Framework for AI Divide Research”, there are three levels of AI divide Access, Capability and Outcome. Social factors such as Demographic and economic factors such as education, ethnicity, race, gender, social class and income contribute to the AI divide.  

Socio-technical factors such as Skills and digital literacy, which include computer experience, general interest use, online purchase, and online information search, are significant components of the AI divide. Furthermore, technical factors such as the infrastructure add to the influencing components of the AI divide. 

AI applications permeate global, organizational, and individual interactions. However, actors will not likely share AI innovation’s benefits in an impartial manner. Developing economies with insufficient digital infrastructure and limited capacity for innovation may not realize the same benefits and convenience from AI innovation as developed countries. Similarly, companies that fully adopt AI technologies may gain an advantage over companies that do not. For instance, an organization may deploy AI to replace human labour and lower operational costs. At the individual level, we will see variances in AI access, comfort, and outcomes. 

• Demographic and socioeconomic factors: Factors such as income, gender, education, ethnicity, and age differentiate who can access and use ICTs. These factors may also differentiate who can access and use AI. Future research needs to examine the impact of demographic and socioeconomic factors on the AI divide. 
• Infrastructure: System developers and designers build “intelligence” into information systems. For example, Google Photos can use facial recognition to allow users to search their photos by people, things, and places. AI-enabled facial recognition helps Google Photo identify a person from a digital image or a video source by analyzing features and building a machine-learning model.  
• AI-specific factors: AI-specific factors will impact the AI divide—algorithm and data. For example, suppose one trained a machine learning algorithm for clinical decision support with data in one country. In that case, the system may not perform with the same level of accuracy when applied to citizens in a different country. 
• Skills and Digital Literacy: Computer-based skills and digital literacy, which contribute to the existent digital divide, will also impact the AI divide. Van Deursen and Van Dijk argued that as the digital divide evolves, differences in skills using technology may create inequity in technology use.  
• Beliefs: Individual concerns about AI-related risks may reduce users’ willingness to engage with AI tools or systems. For example, facial recognition can identify human faces in photos. The social-technical framework we propose identifies how actors can generate, mitigate, or bridge the AI divide as AI permeates society. 

The digital divide has traditionally focused on human access, skills, and capacity. However, invisible AI complicates the interaction between humans and AI-enabled systems. AI innovations interact with users via front-end interfaces, and the training data actors select and manage.  

Exploring the Intersection of the Digital Divide and Artifificial Intelligence: A Hermeneutic Literature Review