Book Review: AI and Society: Tensions and Opportunities by Christo El Morr Editor

Rather than analyze the technical underpinnings and processes involved in the creation of increasingly capable artificial intelligence (AI) systems, AI and Society: Tensions and Opportunities focuses on social issues of safety, privacy, bias, and ethics associated with AI usage (El Morr 2023). It is noteworthy that of the 40 contributors, most are either from Schools of Health (with a focus on Health Information Science, Policy and/or Management) or Schools of Science (with a focus on Information Technology or Computer Science), with a smaller portion coming from Schools of Law and Philosophy or Ethics programs. The book is 307 pages long comprised of 18 Chapters divided into 4 sections. This review offers a concise overview of each section and aims to familiarize readers with the fundamental research conducted in other fields, which contributes to the broader understanding of the market implications of this emerging technology. The primary aim of this review is to deepen our knowledge of the influence that emerging technology exerts on society and culture, while also exploring the connection between macromarketing and other fields of study.

According to Wilkie (2005), while marketing practice is primarily driven by the marketing concept, it is critical to incorporate society at large into the profession, emphasizing the importance of consumer well-being. Macromarketing, with its focus on the interactions between markets, marketing, and society, is linked to the concept of quality of life (QOL) (Sirgy 2021). The notion of QOL was defined as a marketing practice that aims to improve the well-being of customers while protecting the well-being of the firm's other stakeholders. That is, QOL marketing is founded on an ethical marketing theory with two dimensions: improving consumer well-being (beneficence) and protecting other stakeholders (non-maleficence).

Yet, traditional macromarketing definitions of consumer well-being may be overly concentrated on goods and services, such as materialism (Sirgy 2021). Because effective markets contribute to the well-being of consumers, communities, and stakeholders (DeQuero-Navarro, Stanton, and Klein 2021) this review provides macromarketers additional insights into AI theories and practices related to QOL, ethical challenges, and public policies based on the four sections of the book: 1. AI in health: contextual challenges, 2. AI impact (social and legal aspect), 4. AI in action and ethical challenges, and 4. AI and Humans (Philosophical reflections).

The first section of the book focuses on AI-related challenges to the healthcare industry, such as safety and privacy concerns. It introduces the “mosaic theory” which suggests that the sum of seemingly unrelated data can become greater than the individually meaningless parts (Germani 2023). This theory in practice has the potential to lead to ‘health segregation’ in job applications, where the aggregation of data from credit card purchases, location, social media, and healthcare providers/devices automatically rejects candidates with HIV, cancer, heart disease, mental health issues or who are currently pregnant (Germani 2023). It argues that greater emphasis must be placed on understanding the AI systems and applications, as well as the context in which the technology is employed. AI has been used to improve the safety of healthcare, with enhanced alerts, reminders, and alarms supporting health professionals as they prescribe medications and monitor a patient's health status (Borycki and Kushniru 2023). AI is also applied to electronic health record (EHR) data to facilitate physician decision-making. As the primary healthcare technology, EHRs collect vast quantities of patient data, yet reports referenced in this section found that up to 28% of patient diagnoses were missing (Borycki and Kushniru 2023). Given that hospitals use these systems to facilitate health professional communication, decision-making, and patient care activities, missing, biased, or incorrect data can have dire outcomes. Other applications of AI in drug safety include the use of AI tools to monitor for and prevent safety events such as administering the wrong medication to the wrong patient, providing a patient with the incorrect dose of a medication (leading to a drug overdose), and the reconciliation of medications when a patient transitions from one healthcare setting to another (to ensure the patient is prescribed the appropriate medications) (Borycki and Kushniru 2023). In the United States (US), 96% of hospitals and 86% of office-based physicians used EHRs in 2017, and in Italy, 85% of health services use EHRs (Tikkanen et al. 2020).

The use of machine learning presents a number of challenges, especially in terms of health safety for both physicians and consumers during the service encounter. Physicians are concerned about the impact of AI on their decision-making (e.g., presenting erroneous information or lowering the quality of their decisions due to AI bias) (Price, Gerke, and Cohen 2019). Patients are concerned that AI technology would lead to fewer treatment options, and inaccurate diagnoses (Richardson et al. 2021). If the AI misdiagnosed a patient and subjected them to risky procedures and emotional stress, who would be held accountable: the physician acting as a human supervisor or the AI itself? Would the patient be granted access to the AI algorithms on which their life depends, or would they be considered company secrets?

Further, a well-known fact among academics is that data is only as good as the technique used to gather it. Secondary data analysis is plagued with difficulties, ranging from inconsistent labeling and classification of data to poor collecting quality in healthcare. Increased data hygiene can be attained by establishing a standard data gathering system, such as EHR software, but this is out of reach for most low-income countries and institutions (Kumar and Mostafa 2020).

The second section of the book focuses on AI and its social/legal impact. Comprised of 8 chapters, it accounts for nearly 50% of the book's pages. The section begins with AI-generated social media recommendation systems (Vybihal and Desblancs 2023) and covers interesting topics related to sports betting and referee integrity (Dinca-Panaitescu and Dinca-Panaitescu 2023) and Citizen Digital Twins (CDT) (Kopponen et al. 2023). As digital and physical worlds become increasingly entangled, CDT may be of particular interest to macromarketers because of its focus on Human-Centric AI Transformation (HCAIT), which creates the opportunity for citizens to exert greater influence on the development of digital services meant to improve well-being and increase decision-making capacity focused on desired end-states, rather than external societal and commercial forces (Kopponen et al. 2023).

Another interesting topic addressed in this section is how AI interferes with the bargaining system and contracts related to marketing systems (Turnbull 2023). That said, the essence of a contract is the exercise of two parties’ will, which results in a bargain. This traditional concept of bargained agreement is a “creature of the nineteenth century” (Horwitz 1974), and is related to consensus theory, which holds that a contract is an artifact of consensus ad idem, or a meeting of the minds. By the end of the twentieth century, standard form contracts become ubiquitous, accounting for more than 99% of all contracts used in both business and consumer transactions (Burke 2000). They became even more pronounced as they transitioned from paper to electronic form. This meant that any procedures developed using technology were acknowledged as functionally equivalent to their paper-based counterparts.

The increasing sophistication of AI compounds the controversy surrounding standard form contracts. Ultimately, the increased use of AI in the contracting process indicates a shift away from conventional contracting. While some scholars have referred to digitally mediated contracts as “pseudo-contracts” (Kar and Radin 2019), Scholz 2017 provides specificity with the term “algorithmic contract”. She explains that algorithmic contracts and standard form contracts share certain characteristics, including the use of regularized language. The primary distinction between a standard form contract and an algorithmic contract is that an algorithm takes the place of a human in the decision-making process (Scholz 2017). Simply put, algorithmic contracts are contracts that are legally enforceable and rely on algorithmic decision-making.

Algorithmic management supports the gig economy (Duggan et al. 2020). Ride-hailing marketplaces like Uber and Lyft match millions of riders and drivers, and a key component of these marketplaces is a surge (dynamic) pricing mechanism (Garg and Nazerzadeh 2022). For example, Uber gathers enormous quantities of data from its drivers and passengers and is well-known for its surge pricing. This information is saved and utilized to forecast supply and demand, which sets fare prices. Adjusting for supply and demand makes economic sense in principle, but the outcomes are unclear when an algorithm is in charge of surge pricing or other contract details. When a passenger looks for a vehicle, the contractual relationship is established and governed by an algorithm. Similarly, the driver's contract with the passenger also relies on the algorithm and sometimes the driver is not even aware of the terms of the contract they are responsible for fulfilling. For example, until changes were implemented in 2022, drivers did not know the final destination until the passenger was in their vehicle and the app acknowledged that the trip had started. At the same time, they introduced ‘upfront fares’ for drivers in select US markets (Clark 2022). However, there are still domestic and global markets in which Uber drivers are not privy to the algorithm-determined fare. This means they are expected to accept contract terms and agree to work without prior knowledge of how much they will be compensated.

In the business-to-business context, Scholz 2021 distinguishes two types of algorithmic contracts. First, either before or after a contract is formed, gap-filler algorithmic contracts rely on the algorithm to determine and fill gaps with standard legal terms. In terms of both price and conditions, algorithmic gap-filler contracts pose determinacy difficulties in the marketing exchange process. For example, Amazon employs standard form terms and conditions, but an algorithm determines the exact price of an item for each entity (Scholz 2021).

With the negotiator algorithmic contracts, an algorithm chooses which terms to offer or accept. In other words, algorithms serve as negotiators prior to the formulation of the contract. Negotiator algorithmic contracts have been utilized in high-frequency trading of financial products, where investment banks use algorithms to develop real-time purchasing and selling strategies (Scholz 2021). Consumers, and suppliers in some situations, have a limited comprehension of the nature of algorithmic contractual terms and virtually no control over them.

The notion of agreeing to contractual terms determined by an algorithm causes concerns. On the one hand, “clicking to agree” has become part of the daily routine (e.g., Uber, Lift). Consenting to terms and conditions to trade goods and services has become social norm. Thus, we propose some concerns and questions that may be of interest to macromarketers.

The last two sections of the book address ethical challenges and philosophical reflections. Socioeconomic and systemic biases may be exacerbated by AI due to input, algorithmic, and cognitive bias (Kundi et al. 2023). Social bias in AI can be caused by biased algorithms due to biased data caused by human factors, missing data, and underrepresentation of certain social groups. An AI algorithm that learns from older EHR data may not recommend that older women be tested for cardiac ischemia, delaying potentially life-saving treatment and reinforcing implicit social biases. Social biases within healthcare systems lead to inequity in healthcare delivery, which results in unfavorable outcomes for underrepresented groups. Black women with breast cancer were less likely to be tested for high-risk mutations than White women because the data used to train the AI were influenced by the lack of Black women's representation. Similarly, because of historical data showing a correlation between ethnic minorities and patient no-shows, predictive booking algorithms for medical appointment scheduling can lead to biased results that limit options for certain groups (Kundi et al. 2023). AI has the potential to reinforce and perpetuate social biases that effectively limit care options, thereby worsening patient outcomes and health disparities. If the product of efficient market systems is an increased assortment of offerings that increase QOL (Layton 2019), these examples demonstrate that the current market system has yet to reach optimal efficiencies. This gap between current and ideal system operation is opportune for macromarketing research.

While automation with AI is anticipated to relieve humans of tedious and unnecessary Internet of Things (IoT) tasks, working individuals are concerned that AI will eliminate their jobs in the future. It is predicted that by 2030, intelligent robots and machines could replace as much as 30 percent of global labor force hours (Ellingrud et al. 2023). Depending on various implementation scenarios, automation will displace between 400 and 800 million jobs, and approximately 375 million working individuals will be required to completely switch job categories.

It has been said that AI will be as transformative as the Industrial Revolution (Devlin 2023) and is already impacting numerous facets of our lives. While this book would not be appropriate for someone wanting to understand the technical mechanics of AI, it is thought-provoking and thorough in its focus on societal implications, especially as it relates to healthcare, ethics, and bias. The book was not written by or for those tending to the business of AI. Nor does it rely on the sensational blurring of science fiction and science to make its arguments. It presents sound, supported findings for consideration of the eminent societal issues arising from AI usage without dramatic speculation or exaggerated statements that read like clickbait. We need to heed the call and use our skills to tackle the consequences to consumers’ QOL if the issues presented here continue to go unexamined or unregulated by those with commercial influence and interests in AI deployment (Iordanou and Antoniou 2023). It is our responsibility as macromarketers to research and report the disruptive influence, both positive and negative, that AI will have on market systems, and offer solutions that benefit all stakeholders. This book provides a variety of theories and frameworks from other disciplines that can and should be incorporated into our examination of AI's macro-level impact.