Advertising Benefits from Ethical Artificial Intelligence Algorithmic Purchase Decision Pathways

Artificial intelligence (AI) has transcended across numerous fields in our lives (Kaplan & Haenlein, 2019, 2020; Russell & Norvig, 2020). Further, AI can dramatically improve individuals and organizations’ efficiencies in almost every endeavor (Frankish & Ramsey, 2014). In marketing, AI and its relevant subsets, such as machine learning, deep learning and neural network, can strongly support resolving issues (Jordan & Mitchell, 2015). Nonetheless, ethical considerations in the design and employment of algorithms are indispensable, since it champions fairness and reduction of harm to individuals (Johnson, 2015). Such work is essential as individuals and organizations attempt to train models on everything from stopping hate speeches online to marketing advertising as well as ensuring more fair and equitable hiring and promotion methods (Kirsten, 2018; Kirsten et al., 2019). To this end, AI and its related technologies are rapidly transforming society and will continue to do so in the coming decades. This social transformation will have profound ethical impact, with these prevailing innovative technologies both enhancing and disrupting individuals and organizations’ actions.

AI is the foundation for representing human intelligence processes through the creation and application of algorithms constructed into a robust computing environment (Russell & Norvig, 2020; Xu et al., 2020). Stated simply, AI is undertaken to make computers reflect human brain neural processing. Accomplishing this end necessitates three vital elements, which are computational systems, data management, and algorithms. AI-based algorithms provide machines the capability to make decisions, recognize patterns and draw useful conclusions, which basically means imitating human intelligence (Rodgers, 2020).

Primarily, the goal of an algorithm is to resolve a specific problem, generally defined by someone as a sequence of steps. To accomplish this, the imitation of human cognition and functions, comprised of learning and problem-solving, is a prerequisite (Russell & Norvig, 2020). In machine learning or deep learning, an algorithm is a set of rules given to an AI program to help it learn on its own. Whereby machine learning is a set of algorithms that enable the software to update and “learn” from prior results without requiring programmer intervention. Furthermore, deep-learning models can be depicted as a related field to machine-learning models that deals with algorithms stimulated by the structure and function of the human brain called artificial neural networks.

Advertising is a marketing area in which AI-based algorithms may assist in formulating enhancement of planning and strategy (Van Esch et al., 2020). Moreover, ethical considerations in AI are critical in terms of fairness in advertising. A central problem in intelligent advertising is that new algorithmic models are needed to better understand ethical considerations and deliver the “appropriate” advertising for customers at the “right” time. While consumers’ desires, needs and wants can be considered as a foundation of intelligent advertising, it is necessary to analyze consumers’ positions (Li, 2019b) or consumers’ journey. This is important in order to understand consumers’ insights in digital advertising (Kietzmann et al., 2018) predicated on AI technologies (Deng, Tan, et al., 2019). Furthermore, advertising views, comments, and unique visits are a set of soft metrics or minor conversions to measure branding. The set of hard metrics and major conversions pertains to application downloads, sign-ups, and purchases in order to measure selling. According to Deng, Tan, et al. (2019) and Malthouse et al. (2019), hard metrics and macro conversions remain comparatively easy to follow in AI-powered analytics unlike soft metrics and minor conversions. Therefore, a huge marketing challenge is to uncover a method to use AI-powered algorithms to analyze hard metrics and macro conversions from consumer’s actions. In other words, this type of procedure may deliver improved intelligent advertising to better understand the purchase decision process.

According to the aforementioned, marketers and advertisers can adopt AI systems (Kietzmann et al., 2018) to be more effective in understanding and reaching consumers at different stages of the consumer journey (Petro, 2018). To appreciate what AI creates for the advertisers and marketers, one must understand ethical considerations and reach the customers at different stages of the consumer journey.

Therefore, this paper imports a Throughput Model (TPM) a decision-making model that encapsulates AI algorithms that can be utilized to generate six dominant ethical behavioral pathways. Therefore, the TPM (Rodgers et al., 2014) is used in marketing to provide an explanation for individuals and organizations’ algorithmic decisions by touching upon several distinct decision-making aspects. Moreover, AI ethical issues such as bias and deception have already started to have an impact on businesses and individual, which the TPM can speak to by means of its algorithms.

In addition, the TPM addresses online decision-making as a cognitive process, which occurs in the mind of customers prior to a decision choice being made on the Internet. The online decision-making perspective suggests that a decision can be influenced by one of six dominant algorithmic pathways by the context in which it is made. This perspective is focused on learning the factors (i.e., perception and/or information) that bring value to the options (i.e., judgment) before a decision choice is made (see Fig. 1).

TPM is a basic conceptual quantitative model that links the phases of a purchasing process in terms of “perception” (P), “information” (I), “judgment” (J), and “decision choice” (D), where “P” or “I” (or iterations between both) lead to “J,” which then leads to “D” (and/or “P” directly leads to “D”). Hence, understanding the different algorithmic pathways that lead to “D” can help to suitably design a system to uncover consumer decision models. Moreover, the algorithmic pathway that customers link to an option can vary depending on the number of alternatives, the decision maker's mood, her/his former experience with that kind of decision, and so on.

The ethical algorithms can be depicted as follow: (1) P → D represents ethical egoism. (2) P → J → D portrays the deontology viewpoint. (3) I → P → D focusses on the relativist perspective. (4) I → J → D suggests the utilitarian position. (5) P → I → J → D highlights the virtue ethics viewpoint. (6) I → P → J → D represents the ethics of care philosophy (or stakeholders’ position).

These six pathways are viewed as the most dominant and influential for decision-making dominated by moral perspectives. Although, it is important to note that other pathways in the TPM also contribute to the above philosophical positions. Our contention is that the corresponding algorithmic pathway to each noteworthy philosophical view is the most dominant (Rodgers and Gago, 2001). This concept differs from the traditional model.

The traditional model is based on two major assumptions (Tversky & Kahneman, 1981). The first assumption suggests that choices are understood to follow the principle of utility maximization. Utility can be thought of as levels of satisfaction, happiness, or personal benefit. Individuals act to maximize personal subjective benefits by assessing each option aligned with the weighted criteria and choosing the option with the highest total score. The second assumption of the rational model is that individuals' preferences are well-defined and constant over time (Tversky & Kahneman, 1981).

The TPM differs from the traditional economic theory (i.e., rational model) because it is (1) a process model (i.e., opens up the black box), (2) similar to a human neural network providing parallel routes in two stages (i.e., IJ and PJ in the first stage and PD and JD in the second stage) (see Fig. 1), (3) inclusive of a symbolic neural network function (i.e., PI) that imitates a Bayesian model (see Fig. 1), and (4) provides different stages representative of human information processing.

Further, this paper is inspired by the following research questions:

The TPM’s six major ethical algorithmic pathways are discussed later more in detail to indicate their valuable part of any AI-based algorithmic system. The algorithmic pathways help to integrate complex ethical issues into a few central concepts, which can provide widespread commitment to a particular set of values. They can also offer an informal means of holding individuals and organizations accountable to reassure public concerns.

Within this context, the purpose of this article is twofold. First, this paper sheds light on a particular driver of intelligent advertising in the AI age. In particular, the study focuses on the development of intelligence modelling. Moreover, the research investigates the influence of AI and its related subsets of machine learning, deep learning, and neural networks on advertising. Second, this paper proposes to apply ethical considerations for intelligent advertising to the algorithmic purchase decision pathways of the consumers. This procedure may assist marketers and advertisers in understanding and reaching the appropriate ethical considerations for a particular class of customers at the appropriate time. This approach combines the TPM (Rodgers, 2006, 2010) and AI technologies to build the neural network for intelligent advertising based on the consumer’s purchase decision algorithms. Further, this model is imbrued with ethical considerations with the TPM that can assist individuals with a framework enabling them to analyze complicated ethical situations (Rodgers, 2009; Rodgers & McFarlin, 2016).

This paper opens with a brief, general discussion of the relationship between AI and advertising in the AI world and the current issue is expressed. Then, the recent circumstance of intelligent advertising is explained. The next sections present several online algorithmic purchase decision pathways along with ethical considerations for customers and how intelligent advertising can benefit from these algorithms. Finally, the paper concludes by outlining the conclusions, limitations, and future of the research.

Previous research suggests that the impact of AI on advertising has primarily increased efficiency (Li, 2019b; Rodgers et al., 2021). AI has shifted the way advertisers understand and guide consumers (Kietzmann et al., 2018). AI in advertising is considered the main demand in an e-commerce environment and can assist e-commerce platforms and advertisers in developing ethical responsible apparatuses and meet the online market demand (Qin & Jiang, 2019).

Both practitioners and academics regard AI as an important influence on not only advertising dimensions, such as advertising process, advertising operation, advertising design (Qin & Jiang, 2019), and advertisement production and execution (Lee & Cho, 2019), but also each phase of digital marketing, such as programmatic advertising (Bakpayev et al., 2020; Chen et al., 2019) in the technology world, which has exploded and changed dramatically in recent time. Further, as depicted in the TPM pathways, AI-based algorithms can be adapted to benefit advertisers by transforming big data (including structured and unstructured data) to understand and reach consumers via consumer journey (Petro, 2018).

With the help of AI-based algorithms and machine learning, consumer data can be combined and mined to understand the current customer’s insights. That is, data can be collected from various sources by marketers and advertisers in order to connect actively back to consumers. The growth of intelligent advertising (such as interactive advertising and programmatic advertising) can be viewed as outwardly driven by big data, cloud computing and algorithms (Van Esch et al., 2020). Thus, it is important to implement the TPM algorithmic pathways to assist in better understanding how intelligent advertising can deliver personalization and real-time for customers based on their activities or behaviors.

Cho and Lee (2018) and Lee and Cho (2019) advocated for adaptation to a changing environment by implementing digital advertising along with the combination of advertising concepts such as interactive advertising, Internet advertising, online advertising and/or smart advertising. Furthermore, Li (2019b) confirmed that there are three phases of digital marketing, including interactive advertising, the first stage; programmatic advertising, the second stage; and intelligent advertising, the third stage. Similarly, the TPM follows a stage-by-stage process involving innovative attributes (i.e., perception and information concepts), as well as an options stage of digital advertising (i.e., judgment) before a decision choice. For instance, the TPM incorporated stages involve programmatic advertising that uses algorithms to depict digital advertising (IAB, 2014; Li, 2019a).

With the emergence of the Internet, the nature of advertising changes (Lombard & Snyder-Duch, 2001). The marketer is progressively more relying on interactive technologies, which refer to “methods, tools or devices that allow various entities (individuals, machines, or organizations) to engage in mediated communication to facilitate the planning and consummation of exchanges between them” (Varadarajan et al., 2010, p. 97). Furthermore, this process helps reach customers who adopt the Internet as an information source. This type of marketing communication can be ramped up by the implementation of algorithms and interactivity of AI concepts as displayed by the TPM. According to Johnson (2000), interactivity can make advertising more effective. Based on previous researchers, a model employing an algorithmic process can significantly enhance individuals and organizations’ decision making abilities (Heeter, 2000; Li, 2019b; Liu & Shrum, 2002; McMillan & Hwang, 2002; Pavlou & Stewart, 2015).

Adopting AI is the new wave in many subsets of the advertising field. With the power of AI, advertising process (Qin & Jiang, 2019; Van Esch et al., 2020), automatic advertising personalization (Deng, Tan, et al., 2019; Malthouse et al., 2019), and the main foundation of intelligent advertising, such as consumer’s insights (Li, 2019b) or consumer decision journey (Kietzmann et al., 2018), are supported to develop rapidly and effectively.

The following literature reviews the online decision process and the advantages and disadvantages of current online decision-making models.

Based on modelling the purchase decision pathways, the authors suggest several novel ways for marketers and advertisers to apply in intelligent advertising supported by machine learning. To date, a paucity of research has systematically assessed the ethical issues encountered or considered by those using the Internet purchasing habits. This is where TPM, as highlighted by six dominant algorithmic pathways, can assist customers and marketers.

In addition, the ability of consumers to make better quality decisions in online stores is related to their ability to take advantage of the characteristics of online settings that enhance decision quality, while avoiding those which impair it. As discussed previously, the TPM ethical algorithmic pathways provide characteristics that include (1) time constraints, (2) cognitive costs, (3) perceived risks, and (4) product knowledge.

First, the algorithmic pathways (e.g., quick buying pathway) considers time constraints, which accounts for the physical effort required to conduct a search (Johnson et al., 2003). Moreover, the typical online consumer is “time starved” and shops online to save time (Bellman et al., 1999). Online consumers also exhibit search and evaluation patterns that are consistent with time constraints (Sismeiro & Bucklin, 2004).

Second, the algorithmic ethical pathways involve cognitive costs, which are lower in digital environments since cognitive effort can be shifted to the recommendation agents that are typically obtainable in these environments (Johnson et al., 2003).

Third, based upon a particular algorithmic ethical pathway engaging perceived risks, consumers require stronger signals (e.g., brand names, retailer reputation) to lessen risk (Biswas & Biswas, 2004). Further, risk considerations may be offset by the convenience of a particular algorithmic pathway when purchasing online (Bhatnagar et al., 2000).

Fourth, the algorithmic ethical pathways employed can enhance product knowledge for consumers. In addition, the Internet coupled with AI algorithms can proficiently research products that are an important source of information (Ratchford et al., 2003).

This research provides insight into drivers of intelligent advertising, such as intelligence modelling, in the AI era. First, the research explored the stimulus of AI and its related subsets of machine learning, deep learning, and neural networks on advertising. Second, this paper provided a basis to apply ethical considerations for intelligent advertising in customers’ algorithmic purchase decision pathways. This technique may assist marketers and advertisers in comprehending and reaching suitable ethical concerns for customers. This process is useful since biased human judgments can influence AI-based systems in two distinct manners. The first is bias in the data that systems are trained from for implementation. A second source of bias transpires in the way algorithms are developed for application. This paper suggests six ethical algorithmic pathways that may address the latter problem. Moreover, algorithms are responsible for assisting customers’ online decisions that impact their decision-making more than ever before. Nonetheless, it is becoming progressively apparent that this is not always being done fairly or transparently. Further, not implementing the correct algorithm may trigger harm to customers and influence behaviors in morally dubious ways.

Ethical algorithmic pathways implanted in the TPM refers to the process of evaluating and choosing among purchasing alternatives in a manner consistent with ethical principles. In making ethical decisions, it is necessary to understand different algorithmic ethical pathways to eliminate unethical options and support the best ethical alternative. Connecting ethical theories to algorithmic pathways allows us to better understand the cognitive and information components that are undergirding decision-making used in AI systems.

In addition, based on the TPM, this article examines a more inclusive comprehension and springboard of the numerous ethical algorithmic purchase decision routes of consumers in the online environment. Further, this paper serves to fill an existing research gap in the literature by illustrating how purchase decision-making ethical algorithms based on the TPM are more transparent and unbiased. Moreover, by emphasizing the different perspectives of purchase decision-making ethical algorithms, this paper advances algorithms for intelligent advertising tasks, messages, and processes.

From a theoretical perspective, this study contributes to the literature in the following way. This paper commences by stressing that the TPM can designate the buyer’s characteristics by showing six dominant ethical algorithmic purchase decision pathways for online consumers. Supplementary, this methodology elucidates each purchase decision pathway in detail. Depending on the type of products/services, this procedure has three levels (i.e., one, two and three-step decision-making algorithmic approaches) for the consumer to decide. Further, this paper provides distinctive insights and a wide-ranging reflection on the utilization of the TPM in consumer purchase decisions. In addition, this decision-making model makes use of neural networks to better explain marketing post-purchase decisions by consumers. Second, this paper explores the significance of these purchase decision algorithms for intelligent advertising. By breaking down the four concepts of the TPM and applying them in the consumer decision-making process in four layers like a neural network, these ethical algorithmic purchasing decision models might support intelligent advertising, which can be used in machine learning and deep learning to predict and approach the right advertising tasks and message for potential shoppers.

The TPM advances the marketing discipline as well as other disciplines since it is a process model that (1) allows ethical considerations and transparency by opening the black box, and (2) imitates human brain neural network function by providing parallel routes in two stages (i.e., IJ and PJ in the first stage and PD and JD in the second stage), (3) includes a figurative neural network function (i.e., PI) that trains perception for selection of certain information for later processing and mimics information updating from outside sources, and (4) provides, before a decision is made, different cognitive processing stages, which can be broken into several ethical pathways, namely: ethical egoism, deontology (i.e., rule-based), relativist, utilitarianism, virtue ethics, and ethics of care (i.e., stakeholders’ perspective).

In summary, the ethical algorithmic pathways’ impact on people, organizations, and society shapes practically every question of business and public policy. Nonetheless, core issues, such as bias, transparency, ownership, and consent, can be subject to different meanings in different contexts. Therefore, the six algorithmic ethical pathways may alleviate these problems.