Themed Issue Introduction: Promises and Perils of Artificial Intelligence and Advertising

Of all the technological advances in advertising, none are perhaps more exciting than artificial intelligence (AI). However, couched within this excitement is a multitude of questions: What is AI advertising and what are its promises and perils? The promise of AI is about changing the very nature of advertising, affecting all aspects of the advertising process (Qin and Jiang 2019). The perils, according to some critics, pose real threats to advertising and society, leading to reluctance to embrace AI. For consumers, AI threats include loss of privacy and control (Pega n.d.). For marketers, fear of AI stems from a steep learning curve and not knowing the return on investment or if it will help brand image (e.g., Roe 2020). However, as companies continue to generate exponential amounts of data each year, AI is becoming less an option and more a necessity to be on the cutting edge. Consider these statistics:

• Over 75% of consumers already use an AI-powered service or device (Adobe 2018).

• An anticipated 53% growth is expected in AI marketing in 2021 (Gingerich 2020).

• By 2023, global digital advertising is expected to reach $517.51 billion (Enberg 2019) with AI taking up 80% of this (Ad Exchanger 2019).

Indeed, as this themed issue highlights, AI touches on nearly every aspect of society—from retail and entertainment to finance, politics, and health care. This themed issue was in progress before the COVID-19 pandemic started, and the essentiality of AI was clear even then; it is just as necessary during this crisis. For example, in the financial market, a survey of American bankers found that 75% of respondents said AI was “very important to critical” to their ability to recover from the pandemic (Neal 2020). In another survey from IBM, 85% of advanced adopters have already reduced operating costs with AI (IBM, n.d.), suggesting increased need for the very knowledge provided herein as we prepare for the post-pandemic future.

In light of this, themed issue authors undertake a wide range of topics, theories, methods, tools, and applications, and a collection of technologies, answering calls for further exploration of AI advertising (e.g., see Li’s 2019 special section, “Artificial Intelligence and Advertising,” published in the Journal of Advertising, vol. 48, no. 4). Different aspects of the advertising management process are addressed, such as market research, targeting and media selection, ad creation and design, ad placement and execution, and ad performance. For example, authors explore and discuss “promises and perils” of what happens when AI influencers are used in place of human celebrity endorsers (article #1), options for teaching AI to do ad placements automatically in context (article #2), and a framework that shows promise of increasing autonomy and scalability for creative AI advertising (article #3). Authors also examine AI’s ability to detect brand visual–text mismatches as a means of reducing error and increasing profitability in user-generated content in social media (article #4). Even though AI is a behind-the-scenes process, themed issue research illustrates how businesses are making AI tangible and exposing it to their consumers who want enhanced in-store experiences, intertwined with AI-enabled brand communication (article #5). Finally, authors examine the very promises and perils of conducting research in AI advertising, for example, by comparing the ability of AI to perform “content analysis” comparable to that of a human (article #6).

As even this brief summary shows, the themed issue contains many insights that can help to inform advertising theory and industry on efficiency, effectiveness, and safety of how, when, and why AI operates and does not operate in advertising. However, before moving forward, we first need to explore the question of how to define AI advertising.

What Is AI Advertising?

Table 1 shows a list of definitions. While not exhaustive, the list is useful for this discussion. Several broad observations are offered, and some brief comments are made on definitions used in individual articles, indicated by an asterisk in Table 1. A working framework is then proposed. Exact definitions of AI advertising vary, and the differences in definitions arise from the fact that AI covers a lot of territory.

Table 1. List of definitions.

Author(s)	Academic Field	Definition
Baranov et al. (2020)	Psychology	Artificial intelligence (AI) is defined as “the aggregate of the functional capabilities of an electronic computer to solve human problems” (Artificial Intelligence: Regulation Legal Framework, para. 1).
de Miranda (2020)	Human affairs	“AI is a human-designed digital technology that facilitates the identification of meaningful patterns within bodies of data through the use of computers, and the prediction of future patterns, in order to gain insights that facilitate an automated action or a human decision” (p. 598).
Dobrev (2012)	Mathematics and informatics	“AI will be such a program which in an arbitrary world will cope not worse than a human” (p. 2).
Goel and Davies (2020)	Cognitive psychology	“Artificial intelligence is the field of research to understand, design, and build intelligent computational artifacts” (p. 602).
Hancock, Naaman, and Levy (2020)	Communication	“Artificial Intelligence-Mediated Communication (AI-MC) as interpersonal communication in which an intelligent agent operates on behalf of a communicator by modifying, augmenting, or generating messages to accomplish communication goals” (p. 1).
*Kaplan and Haenlein (2019)	Business	AI is “defined as a system’s ability to correctly interpret external data, to learn from such data, and to use those learnings to achieve specific goals and tasks through flexible adaptation” (p. 15).
Kok et al. (2009)	Computer science	AI is computer programs that “imitating intelligent human behavior” (p. 1). More narrowly, AI could be defined as systems that either think or act rationally or like humans.
*Kumar et al. (2019)	Marketing	AI refers to “the broad idea that computers, through the use of software and algorithms, can think and perform tasks like humans” (p. 135).
Lehman-Wilzig (1981)	Criminology	AI is a generation of robots that “is rapidly evolving, a breed that can see, read, talk, learn, and even feel emotions” (p. 443).
*Li (2019)	Advertising	Intelligent advertising refers to “consumer-centered, data-driven, and algorithm-mediated brand communication” (p. 334).
McCarthy and Hayes (1981).	Philosophy in computer science	AI is an entity with “an adequate model of the world (including the intellectual world of mathematics, understanding of its own goals and other mental processes)” (p. 4). The intelligence means its ability to (1) “answer a wide variety of questions on the basis of this model” (p. 4), and (2) “get additional information from the external world when required, and can perform such tasks in the external world as its goals demand and its physical abilities permit” (p. 4).
Meskó and Görög (2020)	Medicine	AI is an interdisciplinary field spanning computer science, psychology, linguistics, and philosophy, among others. It describes “machines that mimic cognitive functions that humans associate with the human mind, such as learning and problem solving” (p. 1).
Panch, Szolovits, and Atun (2018)	Health	Artificial intelligence—Refers to “a broad scientific discipline with its roots in philosophy, mathematics and computer science that aims to understand and develop systems that display properties of intelligence” (p. 2). Machine learning—“A sub discipline of AI, where computer programs (algorithms) learn associations of predictive power from examples in data. Machine learning is most simply the application of statistical models to data using computers” (p. 2). Deep learning—“Deep learning methods allow a machine to be fed with large quantities of raw data and to discover the representations necessary for detection or classification” (p. 2). Supervised learning—“Training computer programs to learn associations between inputs and outputs in data through analysis of outputs of interest defined by a (typically human) supervisor” (p. 2). Unsupervised learning—“Computer programs that learn associations in data without external definition of associations of interest. Unsupervised learning is able to identify previously undiscovered predictors, as opposed to simply relying on known associations” (p. 2). Reinforcement learning—“Computer programs that learn actions based on their ability to maximize a defined reward” (p. 2).
*Qin and Jiang (2019)	Advertising	AI is defined as “a set of disruptive technologies which simulate human intelligence and realize machine intelligence” (p. 339).
Ransbotham et al. (2017)	Business and management	“AI is the theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages” (Oxford Dictionary p. 2).
Russell and Norvig (2010)	Computer science	AI is “the study and construction of rational agents” that act based on given inputs to achieve the best expected outcomes or goals (p. 7).
Ryan (2020)	Science and engineering ethics	AI refers to “technologies that demonstrate levels of independent intelligence from humans” (p. 2751). Narrow AI, as defined in Macnish et al. (2019), refers to “a system that is designed and used for specific or limited tasks” (cited from p. 2751). General AI, as defined in U.K. House of Lords (2018), refers to AI that “has similar cognitive abilities as humans and can convincingly demonstrate these capacities” (p. 2751).
Schmidt (2020)	Computer science	“An Interactive Human Centered Artificial Intelligence is an Artificial Intelligence that enables interactive exploration and manipulation in real time and is designed with a clear purpose for human benefit while being transparent about who has control over data and algorithms” (p. 1).
*Shankar (2018)	Retailing	AI refers to “programs, algorithms, systems or machines that demonstrate intelligence. More generally, it is used to denote a set of tools that can enhance the intelligence of a product, service, or solution” (p. vi).
Shymko (2018).	Psycholinguistics	AI is defined as “the study of agents that receive percepts from the environment and perform actions” (p. 330).
Stephan and Klima (2020)	Philosophy	AI is the modern term for Adler’s (1967) “robot” that refers to a system that can learn from its environment, can take independent autonomous actions without direct instructions, and can engage in ordinary-language conversation with humans.
Sun, Anbarasan, and Praveen Kumar (2020)	Technology and education	“Artificial intelligence (AI) is the combination of intelligence, that is, machines capable of demonstrating human intelligence and of making decisions with human skills. The main objective is to build highly advanced machines that can make smart decisions” (p. 2).
*Toivonen and Gross (2015)	Computational creativity	Computational creativity is a sibling of AI: “Artificial intelligence studies how to perform tasks which would be deemed intelligent if performed by a human; While computational creativity studies performances which would be deemed creative if performed by a human” (p. 1).
Wahl et al. (2018)	Health	“AI is the branch of computer science that deals with the simulation of intelligent behavior in computers” (p. 1).
*Weber and Schütte (2019)	Big Data and cognitive computing	AI is the science that “tries to figure out how to get machines to use language, to form abstractions and concepts, to solve types of problems that are currently reserved for humans, and to improve themselves” (p. 3).
*Xu et al. (2020)	Customer service	AI is defined as “a technology-enabled system for evaluating real-time service scenarios using data collected from digital and/or physical sources in order to pro- vide personalized recommendations, alternatives, and solutions to customers’ enquiries or problems, even very complex ones” (p. 189).
Yanke (2020)	Law and sociology	Black’s Law Dictionary (2009) defines an artificial person as an entity “given certain legal rights and duties of a human being; a being, real or imaginary, who for the purpose of legal reasoning is treated more or less as a human being” (cited from section 3, para. 2).

* Indicates definitions cited in themed issue articles.

First, definitions of “artificial intelligence and advertising” vary for the six articles comprising the themed issue. Despite these variations, all definitions share a common conception that AI is data driven for the purpose of making intelligent decisions (see article #5 for a similar argument). The term artificial intelligence broadly refers to a range of machine functions that learn with the help of humans or completely on their own (Kaput 2020). The term advertising, as used by our authors, refers to (a) brand communication (b) with an intent to persuade (see Dahlen and Rosengren 2016; Thorson and Rodgers 2019). As such, AI advertising can be viewed as being of a specific type (e.g., AI endorsers in article #1) or as part of a larger process (e.g., creativity in article #3). It can be experienced within a specific context (e.g., social media in article #4) or as part of a larger branded experience (e.g., in-store shopping in article #5). It can also be supervised or unsupervised, can imitate humans or other machines, and can take what is learned and reconfigure itself, as in the case of AI influencers who can make themselves look and act like humans and even stage fights with other machines (article #1). This last example helps illustrate that an “AI advertisement” is not always of the human type but can also be of the machine type or a blend of human and machine. Truly, AI advertising has become complex (see Li 2019) and has come to mean many things, as represented by the research in this themed issue.

To sum it up, for purposes of the themed issue, AI advertising is defined as brand communication that uses a range of machine functions that learn to carry out tasks with intent to persuade with input by humans, machines, or both. The perspective guiding this themed issue is that AI advertising can and should be viewed as a distinctive subdiscipline of advertising that is situated at the intersection of cognitive science, computer science, and advertising. This intersectionality illustrates a very broad scope and helps define AI advertising as a distinct albeit related concept from other concepts in advertising, such as computational advertising (e.g., see Huh and Malthouse’s 2020 special section, “Advances in Computational Advertising,” in the Journal of Advertising, vol. 49, no. 4).

Second, it is important to understand the domain of inquiry, as this determines the types and scope of questions being addressed (Duff, Faber, and Nan 2019). Returning to Table 1, AI (which is basically a computer science domain) is applied to form a distinctive subdiscipline within many academic fields or domains: business, communication, psychology, linguistics, philosophy, mathematics, medicine, law, and sociology. Indeed, the wide variety of authors and program affiliations published in this themed issue speaks to the various perspectives that can be applied to the study of AI advertising. You will note that most of their definitions fall within domains of business, marketing, and advertising (see Table 1). From marketing, authors cite Kumar et al. (2019); from business, Kaplan and Haenlein’s (2019) definition is used; and from advertising, authors cite Qin and Jiang’s (2019) definition of AI and Li’s (2019) definition of intelligent advertising (see Table 1 for definitions). This suggests that the intellectual space currently occupied by themed issue research mainly comes from advertising and domains that are historically associated with advertising, namely, business and marketing. Definitions from other domains are also used, suggesting there is room for a diversity of theoretical approaches and practices for the study of AI advertising. For example, authors cite Weber and Schütte’s (2019) definition from big data and cognitive computing and Xue et al.’s (2020) definition from customer service (see Table 1). The themed issue has also prompted new definitions specific to the study of AI advertising. An example is Thomas and Fowler’s (article #1) definition of AI influencer: “a digitally created artificial human who is associated with Internet fame and uses software and algorithms to perform tasks like humans.”

From this discussion, it can be concluded that the definitions and questions addressed by this themed issue fall squarely within domains that can be legitimately addressed by AI advertising. There are also domains listed in Table 1 whose definitions were not applied to our themed issue but which may be useful for the future study of AI advertising. For example, an ethical perspective is offered by science and engineering and a legal perspective is offered by law and sociology. In light of various definitions, a framework is proposed.

Toward a Framework

Figure 1 shows a classification schema that has potential to house research on AI advertising. The intent is to find common ground among themed issue articles. Three dimensions are identified: artificial intelligence types, artificial intelligence functions, and learning types. Dimensions are briefly explained, and some examples are provided. The proposed framework is meant to be illustrative rather than all-encompassing.

Figure 1. Classification Schema.

First, research can be categorized into three main types of AI: narrow or weak, general or strong, and super (see Ryan 2020 in Table 1), indicated by small, medium, and large circles (see Figure 1). The circle’s size indicates the strength of the AI intelligence. Narrow AI (NAI) or “weak AI” involves a basic level of intelligence that is likened to that of an infant, as it can perform only a specific task or one task at a time (much in the way that a baby will crawl before it can walk). General AI (GAI) refers to machine intelligence at the level of an adult, meaning the machine can be trained to handle more cognitively demanding and complex tasks (e.g., perception, learning, problem solving). GAI can also have human sense abilities, such as seeing, hearing, and feeling, making it a “strong” type of AI. Super AI (SAI) refers to technologies that supersede human intelligence and act independently from humans. SAI, the strongest of all intelligence types, is not yet possible but is of the futuristic type that Stephen Hawking believed could pose a threat to humankind (see article #5). Although the debate continues over AI advertising’s current level of intelligence—whether NAI or GAI—arguably, the research in this themed issue can be classified as GAI. Two examples are the fully autonomous AI influencer examined by Thomas and Fowler (article #1) and the relative autonomous AI-enabled creative advertising system proposed by Vakratsas and Wang (article #3).

Second, research may also be categorized in terms of AI building blocks or functions, for example, natural language processing, image and speech recognition, and problem solving (see article #1 for definitions). Themed issue authors identify and/or use all of the building blocks outlined in Figure 1. For example, Ha et al. (article #4) use computer vision, which relies on image recognition, and Hayes et al. (article #6) use natural language processing, which relies on text (and speech) recognition. Why does this matter? AI building blocks form the basis of theory and shed light on the kinds of data that can be analyzed and types of problems that can be addressed.

Finally, definitions may be classified by learning type. Figure 1 shows five types of learning: machine, deep, supervised, unsupervised, and reinforcement (for definitions, see Panch, Szolovits, and Atun 2018 in Table 1). “Learning” broadly refers to a machine’s cognitive abilities. As stated, cognitive levels range from lower (infant) to higher (adult), and learning can come from input by humans, machines, or both. Research by Watts and Adriano (article #2) is an example of supervised machine learning, where algorithms are trained to learn associations based on examples that humans provide. Supervised deep learning can be illustrated by the computer vision method employed by Ha et al. (article #4), which learned to detect a mismatch in user-generated posts in Instagram based on text, image, and metadata cues. Unsupervised learning refers to the ability to identify predictors not previously known and is perhaps illustrated by the autonomous AI endorsers examined by Thomas and Fowler (article #1). Although the final type, reinforcement learning, was not examined by any of the studies in this themed issue, its utility may be better suited for future studies where the machine’s ability is maximized by a set reward, as in the case of gaming (see Panch, Szolovits, and Atun 2018 in Table 1).

As even this abridged exercise shows, the many theories of AI advertising vary considerably but can be categorized along some common dimensions. Figure 1 identifies three dimensions—AI type, AI function, and learning type—as outlined. This, of course, is not all the possible dimensions, and dimensions are not perfect in the sense that they overlap (e.g., machine learning is considered both an AI function, as defined by Kietzman, Paschen, and Treen 2018, and a learning type, as defined by Panch, Szolovits, and Atun 2018). As technology is always evolving, we assume that frameworks on AI advertising will keep evolving too. What this means is that AI research specifically related to advertising has the challenge of building and testing theories that can accommodate and contemplate all of that complexity, as reflected by the research in this themed issue. Because of the expertise needed, it is highlighted that themed issue authors come from diverse disciplines; three of the articles (i.e., articles #2, #4, and #6) created their own original AI algorithms and/or lexicon to make the research possible. Such endeavors suggest promising avenues for future interdisciplinary teams of researchers working to solve AI advertising problems that are firmly rooted in the field of advertising.

Promises and Perils

As the name implies, the primary goal of the themed issue was to stimulate interdisciplinary thinking on promises and perils of AI advertising. Should we welcome AI for all its promises? Or are the perceived perils too great? Here is what the authors of the six accepted article (out of 31 total submissions) had to say about it.

Article #1. The first article by Thomas and Fowler makes a novel discovery regarding advertising endorsements by using AI influencers as alternatives to traditional human celebrity endorsers. Titled “Close Encounters of the AI Kind: Use of AI Influencers As Brand Endorsers,” the article shows through two studies that AI influencers perform just as well as celebrity endorsers to produce positive brand benefits. Where it can backfire is when AI influencers commit transgressions, for example, saying negative things about a competitor brand, in the same way that celebrity endorsers commit transgressions. In the case of AI influencers, consumers are more likely to think a transgression applies to all AI influencers, but perceive celebrity endorser behavior as more independent and less interchangeable. Taking it a step further, the research finds that one way to reduce negative brand perceptions resulting from transgressions is to replace the AI influencer with a human celebrity endorser, suggesting humans might have to clean up the messes left behind by machines.

Promises: AI influencers are just as effective as human celebrity endorsers in evoking favorable brand responses (i.e., attitudes, purchase intentions) and, therefore, could be a viable alternative to human endorsers.

Perils: As with celebrity endorsers, any transgressions committed by an AI influencer could negatively influence consumers’ brand perceptions and may even prompt consumers’ disavowal of all AI endorsers, suggesting caution is needed.

Article #2. The second article addresses the important issue of how to improve advertising placement effects using machine learning algorithms. In “Uncovering the Sources of Machine-Learning Mistakes in Advertising: Contextual Bias in the Evaluation of Semantic Relatedness,” authors Watts and Adriano argue that machines make costly mistakes with brand placement that could be avoided if they were taught about semantic relatedness, which they define as “the conceptual distance between words in the human mind.” To address shortcomings, a context-aware database is created, and results of the database are compared to current best practices. Improvements in brand placements are seen after teaching machines how to be more context-sensitive when certain words become associated, bringing machines closer to understanding contextual nuance much in the same way that humans understand it.

Promises: AI is useful for training machines to overcome ad placement mistakes by becoming more context-aware, thereby improving advertising efficiency and effectiveness.

Perils: Contextual nuance training is only as good as the assumptions of people doing the training; a broader vocabulary and deeper understanding of contextual nuance are needed.

Article #3. Creativity in advertising is addressed in this next article, authored by Vakratsas and Wang, who propose a creative advertising system (CAS). Titled “Artificial Intelligence in Advertising Creativity,” the CAS is designed to generate and test advertising creative ideas with an AI advertising system, based on both data and human input that builds up an “advertising library.” To do this, the authors provide a broad definition of creativity: “as a search process, the outcomes of which should be evaluated based on a set of rules.” Four interrelated steps are outlined: (1) knowledge development and tracking needed to build up the advertising library; (2) knowledge classification into styles; (3) creative generation using the set of traversal rules; and (4) creative testing using a combination of evaluation rules. According to the authors, the proposed road map is flexible enough to be used with existing advertising concepts and can help explain why there might be inconsistencies for some executional creative elements. Rather than “set it and forget it,” the framework provides a way of knowing when creative work gets divided between machines and humans based on the task and strength of the work.

Promises: The proposed CAS framework provides a road map to potentially rethink and reshape advertising creativity with the development of an AI advertising system.

Perils: It is difficult to assess the value of novel transformational creative ideas entirely based on computational advertising systems considering the difficulties of observing perceived creativity within the consumer behavior matrix.

Article #4. In their article “Automatically Detecting Image–Text Mismatch on Instagram with Deep Learning,” Ha et al. develop and test an innovative machine learning method using computer vision to detect the image–text mismatch in consumer-generated branded content in Instagram. Being able to identify mismatched advertising–editorial content is key to enhancing consumer search and brand performance. Their proposed multimodal approach combines features extracted from image, text, and metadata and is, therefore, useful for garnering consumer insights using multiple sensory modes. After analyzing 452,616 Instagram posts on fashion brands, the authors find the method can efficiently and accurately detect advertising visual–text mismatches and may even outperform a general (nonbrand) model when properly trained.

Promises: Computer vision as a proposed method can help open new research streams in advertising, especially where the concept of contextual congruence can be further developed.

Perils: Substantial training is involved in teaching machines to recognize visual mismatches, and a broader spectrum of visual inputs is needed.

Article #5. The next article, “Stimulating or Intimidating: The Effect of AI-Enabled In-Store Communication on Consumer Patronage Likelihood,” by van Esch, Cui, and Jain, examines “just-walk-out” retail technology illustrated by stores like Amazon Go that let you walk in, grab what you want, and leave without scanning anything, as the AI automatically knows what you bought and charges your card. The authors hypothesize circumstances in which AI-enabled checkout services may stimulate or intimidate consumer patronage. The study, thereby, expands the technology-based brand communication research to offline settings by focusing on AI advertising in the in-store shopping environment. Although AI technology can increase consumers’ store atmosphere evaluations and purchase intentions through in-store communication sensory stimulation, the research finds that any positive effect could be undermined once consumers perceive threats from AI. Even though the research was conducted before the COVID-19 pandemic, the authors note the applicability and relevance of the results for the current and postpandemic world, considering the value of “human contactless” purchase experiences enabled by AI.

Promises: AI advertising in the in-store shopping environment can stimulate consumer patronage likelihood through in-store brand communication sensory experiences.

Perils: The positive effect yielded by in-store communication sensory experiences could be undermined if consumers perceive AI as a threat to their privacy.

Article #6. In the final article, Hayes et al. compare results of a human content analysis to a machine-based analysis using Linguistic Inquiry and Word Count (LIWC), a text analysis program. Their article, “Can Social Media Listening Platforms’ Artificial Intelligence Be Trusted? Examining the Accuracy of Crimson Hexagon’s (Now Brandwatch Consumer Research’s) AI-Driven Analyses,” analyzes a random sample of 10,000 electronic word-of-mouth (eWOM) comments from the Nike “Dream Crazy” advertisement. Results suggest that humans may perform better than machines at coding some brand-specific content, such as brand identification and brand sentiment. Rather than throw out the machine, the authors suggest steps researchers should take to reduce detection errors, such as (1) examine the algorithm documentation, (2) train data sets, and (3) assess AI-generated data prior to their use in research models and decision making.

Promises: Social media listening platforms (SMLPs) can increase data collection speeds and enhance research robustness and efficiency for both industry and academic researchers.

Perils: Machines might not be as nuanced as humans in detecting certain brand-specific classifications (e.g., brand identification). However, this can be both a peril and a promise, because the currently available tools are quite limited and problematic, but there is ongoing research and potential for developing more advanced deep learning AI tools for enhanced and more accurate detection and classification of brand-specific text that does not contain specific brand names. This means that researchers will need to prioritize the most important data to be obtained and the right tool or method for the job.

Closing Remarks

This themed issue has brought together high-quality research that explores promises and perils of AI advertising. Combined, these six articles represent what we believe is a unique contribution to advertising theory development and practice. A key point to remember is that theoretical-based considerations are necessary for guiding the improvement of algorithm performance in AI advertising. In addition to the efficiency and privacy trade-off issue mentioned (article #5), another important problem connected to all these themed issue articles is an emerging issue of algorithm bias and algorithmic transparency and fairness. Algorithmic transparency is an emerging interdisciplinary subfield in computer science that focuses, in part, on understanding why algorithms make the decisions that they do (see related discussion on this issue in Helberger et al. 2020). A lack of awareness and understanding of algorithm bias among the general public is a significant emerging issue that requires serious intervention to improve algorithm literacy; we also notice that a lot of social science scholars, including advertising researchers, also seem to assume a value-neutral and unbiased nature of AI algorithms, which is quite dangerous in many ways. Thus, in addition to offering suggestions for improving the technology needed to examine theories of AI advertising, themed issue studies offer explanations and guidance for understanding how or why AI may be helping and/or standing in the way of future progress in advertising. Our intention is to stimulate interdisciplinary research which will have a substantial impact on thinking and which leads to additional research on AI advertising in the years ahead.

Acknowledgments

The editor wishes to thank all of the Journal of Advertising’s loyal contributors and reviewers who participated in and made this themed issue possible; Drs. Hairong Li and Shyam Sundar for commenting on an early version of the call for papers; Dr. Jisu Huh for her valuable feedback on this introduction; Weilu Zhang for her assistance, especially with Table 1; and Evgeniia Belobrovkina for her tireless commitment to keeping the front door of the Journal of Advertising running to perfection in her role as editorial assistant.

Additional information

Notes on contributors

Shelly Rodgers

Shelly Rodgers (PhD, University of Missouri) is a professor of strategic communication and the Molly Phelps Bean Faculty Fellow, School of Journalism, University of Missouri, Columbia, Missouri, USA.