AEJMC Advertising Division 2023 Teaching Pre-Conference Review: Innovating Data Storytelling and Visualization With Artificial Intelligence and Chat Generative Pre-trained Transformer

Abstract

The 26th annual Teaching Pre-Conference organized by the Advertising Division of the Association for Education in Journalism and Mass Communication focused on the topic of innovating data storytelling and visualization with AI and ChatGPT. Five prominent speakers from leading media companies and universities shared insights with advertising educators, covering the application, impact, and challenges of generative artificial intelligence (AI) in the advertising industry. The five panels also delved into effective ways of integrating generative AI tools into the classroom. Three key trends that arise from the panel presentations are discussed below. Relevant advertising AI tools and class activities are also shared in the report.

The Advertising Division (ADVD) held its 26th annual Teaching Pre-Conference Workshop on Sunday, August 6, 2023, at the Association for Education in Journalism and Mass Communication’s (AEJMC) annual convention in Washington, D.C. The pervasiveness and complexity of generative artificial intelligence (AI) are undeniably relevant to the advertising community; professionals and educators alike. Given the rapid growth and evolving nature of AI, it was an important topic for the Advertising Division Teaching Pre-Conference and was sold out due to high interest.

The planning committee invited a combination of professionals and professors to speak about how they use AI at work and in the classroom. The pre-conference provided valuable time and space for attendees to understand, explore, and ask questions about various AI tools and their impact on the industry and advertising education. The speakers and attendees were invited to share AI resources which have been compiled and are available here and in the AEJMC Advertising Online Community. What follows is a synopsis of the pre-conference and the points our esteemed speakers made.

Morgan Bramlet- Principal and Creative Director, Blue Fusion and Director, Brand and Creative, American Nurses Association/ANA Enterprise.

Our first speaker, Morgan Bramlet, is an award-winning creative with over 20 years in agency and client-side advertising. His wife, Eileen Clark Bramlet, is the Senior Vice President of Communications and Events at the Copyright Alliance. She and her colleague, Courtney Lang, infused some copyright insight into the conversation. Morgan had a two-hour time slot packed with information.

As a curious creative, Morgan began exploring AI applications in their early stages to see what they could do. Earlier versions of some applications could yield “decent” renderings but there were glaring inaccuracies like extra arms and fingers and other strange phenomena. AI tools change rapidly, and Morgan continues to play around, finding that speed, efficiency, and quality of output increase with each update. Costs can be reduced; time can be saved; an appealing proposition to agency owners. However, concern mounts for the future of creatives, artists, actors, and writers as generative AI takes over (DeCremer et al., 2023).

Interesting examples of how a variety of generative AI tools can create content were highlighted. Morgan created a lifelike model with Midjourney that explained what his presentation would include, personalizing it for our specific pre-conference audience. The script was created in conjunction with ChatGPT and used a synthesized voice for the avatar created in D-ID, which is an AI video creation platform.

The effect was highly engaging. You can see the AI engineered model, created via multiple generative artificial intelligence tools, in the first 3 minutes of the AI Mashup video here.

AI isn’t new, it goes back 100 years or more, explained Morgan. AI was an idea first; a concept, a fad, a toy. Then, an AI became an assistant (think chatbots) and finally a tool to generate content in many forms. He ran through a brief history, shown below:

1927 = Metropolis(film) and the “idea of intelligent machines”.

1950 = Alan Turing “Artificial intelligence pioneer,” inventor of the Turing Test (Copeland, 2023).

1966 = Eliza the first chatbot at M.I.T. combining human language and machines (Wikipedia).

1998 = Furbie a toy that developed human language skills over time (Wikipedia).

2011 = Siri natural language, machine learning personal assistant developed at Stanford (Sanchez, 2023).

2014 = Alexa artificially intelligent personal assistant independent of an iPhone (Stone, 2021).

Tools continue to expand, generating text, images, video, information, and analytics in response to prompts due to generative AI technology. Put in a prompt, and generate content—you know how it works. AI tools are trained on data sets with algorithms to teach the machine how to respond and produce content. As we have seen, scraping the internet full of individually created and copyrighted material can be problematic (Allyn, 2023).

Morgan then reviewed some of the more recent AI major players, which you can access on our resource sheet (to be certain, more AI developments have come on the scene since we created this list). For advertising and marketing professionals, Morgan notes that AI can be beneficial for creativity, content creation, analytics, data visualization, strategy, and other future uses. He demonstrated how he generated iterations of copy and visuals in an ad for a made-up brand using ChatGPT-4 and Midjourney. Morgan showed interesting visual content using prompts in Midjourney and discussed different ways the tools can be helpful in the industry and for an assignment in the classroom. He also explained that the images were generated from billions of bits of data, making attribution a challenge.

The issue of AI and copyright was discussed, and fortunately, Eileen Clark Bramlet, Senior Vice President of Communications and Events at the Copyright Alliance, was in attendance and weighed in. At the time of the pre-conference, there were a lot of gray areas regarding generative AI and copyright. Eileen mentioned the Copyright Alliance had been working with Congress and giving presentations on “the hill” as our representatives work to develop more laws and guidelines in this area.

For now, AI content can be used for inspiration or as a first draft to put the creator’s own spin on; but unless one makes the AI-generated content uniquely their own, it can’t be copyrighted. A creator cannot publish AI work as is without attribution and there is no way to know who contributed to the bits that constitute the output.

As educators, it is important to let students know they cannot use AI content straight up and call it their own legally unless they do something to change it significantly and it cannot be recognized as the original creator’s work. Otherwise, someone will recognize their work in AI-generated content, which can mean legal trouble. Ultimately, students need a solid foundation in their discipline to develop fundamental skill sets without relying solely on AI. AI should be used more as inspiration than full-on creation. There are a lot of legal gray areas and it is important to own up to your process and be transparent on how content has been created. In other words, show your work.

Currently, the U.S. Copyright Office won’t copyright anything that isn’t “human touched,” yet there is no formula on how much of the content must be human-generated. This is an ongoing debate and there are no definitive answers at this time. The best thing to do is disclose your process when using generative AI tools for content creation. As we know, many lawsuits are working themselves through courts. For example, the New York Times sued ChatGPT and others for scraping their copyrighted work to train the AI platform (Allyn, 2023).

Takeaways: AI isn’t new but is proliferating in powerful ways with multiple uses across a variety of platforms. There are pros and cons to AI including ethical and legal concerns. AI results can be inaccurate, outdated, and nonsensical. Deepfakes “a fake, digitally manipulated video or audio file produced by using deep learning, an advanced type of machine learning, and typically featuring a person’s likeness and voice in a situation that did not actually occur” are misleading at best and dangerous when created for ill intent (Dictionary.com, n.d; Altuncu et al., 2022; Hwang et al., 2021). Yet AI can save time and money, and inspire new ideas. Though AI can be a shortcut to information seeking, predictive analytics, and content creation, one must be transparent in using AI. Legislation has not caught up with the technology yet leaving many gray areas.

Kim Herrington- Senior Analyst | Data Journalist, Forrester

Kim is a former improv comedian which was apparent throughout her spirited presentation. Kim’s mission is to help people with tech tools, communication, and data literacy advocacy. In 2020, Kim told us she became “the world champion of data literacy advocacy,” which gave her a bit of imposter syndrome. She likes to encourage people to explore data and technology in a creative and playful manner. Her goal was to have attendees feel that “technology can be fun and useful, that play is important, and that production versus perfection” is the mindset for trying new platforms.

When Kim became curious about creating images with Midjourney, YouTube helped her learn how. She explained why she pays $30 a month for a subscription: She didn’t want others to see her rookie work as she was just starting out, and, because it is addicting. As Kim played with prompts, she became excited and proud of what was generated. She noticed that the more detailed prompts rendered more interesting the images. Midjourney has tutorial documents that are helpful in learning how to create interesting prompts.

Kim recounted a story of showing her five-year-old daughter an image she prompted using Midjourney based on the child’s imaginary friend. The daughter thought it looked like her imaginary friend, but with certain differences, which she explained to her mom, prompting more renderings. Mother and daughter were so excited about bringing this imaginary world to life graphically, that Kim wanted to share the fun with her daughter’s classmates. She created a prompt sheet for the other students to take home so they could express feelings, activities, settings, creatures, and colors to generate visual outputs (See prompt form here).

Kim Herrington’s AI Midjourney prompt sheet QR code

Her inspiration for the assignment was her concern that perhaps children are not using their imaginations as much as in previous generations. Kim teaches adults about data literacy, “which is all about creativity, curiosity, and critical thinking.” The rest of her time was spent showing us silly prompts from five-year-olds and the fascinating graphics Midjourney created from them. Kim emphasized how she was learning as she played with the technology. Ultimately, she had the audience cracking up with her fun use of memes, storytelling, case studies, and encouragement.

Takeaways: AI tools can be useful and amusing. Cut loose and play with new technology. Don’t worry about perfection “because you’re making dinosaurs and dolphins on bicycles. Have fun and play.” Spend some time in your imagination and make it come to life with AI. Approaching AI tools with playfulness and low stakes assignments in the classroom can likely assist students in developing their own curiosity around artificial intelligence without fear of failure that some may experience when trying evolving, even intimidating, technology.

Ryohei Okabe, VP, Product, Zeta

Our third speaker dove deep into the back end of marketing technology, giving us a glimpse of what happens behind the scenes when people shop online. To begin, Ryohei gave a refresher on relevant mar-tech acronyms and terms:

CDP: Customer Data Platform

ESP: Email Service Provider

DSP: Demand Side Platform

SSP: Supply Side Platform

BI: Business Intelligence

CMS: Content Management System

He noted that marketers need these various platforms and that his company provides them all in one platform. Beyond technology, marketers need services such as product design, engineering, product management, product marketing, quality assurance, learning and development, strategy, account management, analytics, creative, campaign management, and data management. Ryohei mentioned that many companies may have their own internal marketing teams of 20 or so, but what they really need is about 100 on their team. This is where Zeta, Ryohei’s company, can fill in the gaps with technology and services in the digital space. Zeta works with a wide variety of major US brands including various retailers, media companies, and other technology organizations.

Ryohei explained the behind-the-scenes process of getting to know a consumer and what they may like so that the brand can offer them what they, and people like them, might want. He showed the “consumer journey” of someone who may have logged into a website to purchase an item, therefore providing first-person data so they can transact. This information is used to learn more about the consumer and is added to their consumer database. From there, outbound communication is sent via email, direct mail, messaging apps, and website interactions. Finally, there are AI applications that operate opaquely to connect data to find similar prospects.

One person may have between 100 and 10,000 identifiers on the internet, so they are not looking at individuals, but rather patterns.

With this data, Zeta is able to orchestrate the customer experience through data automation and guide the creative through testing. AI can help with messaging, call to action, subject lines for communication, and recommendations. Additional benefits to AI include gathering consumer identifiers, and detailed marketing analytics.

Instead of reviewing spreadsheets as in the past, AI can look for real-time insights, opportunities, and detailed marketing analytics to increase efficiency, better segmentation, and improve consumer experience. AI can identify “look-alike modeling,” to find others who are similar to current customers. Natural language processing is also a benefit of using AI. For example, it can scan an article and determine recommendations. AI is also used to prove that Zeta’s marketing efforts are effective.

AI can help the marketing process through omnichannel marketing mix and modeling, which means it can automate what, when, and how to send messaging to customers and prospects by assessing engagement and conversion data. You will recognize the output of this recommendation process when you get a message from a brand that says something like, “You may also be interested in ____.” A/B testing, forecasting, and other tools factor into Zeta services driven by AI which all boil down to time and human cost to make things easier. Security assessments and privacy controls are streamlined with AI technology.

As a Product Manager, Ryohei assesses user problems, tries to understand marketers’ needs, and then works with design engineers to build out solutions. He gave some demonstrations on the behind-the-scenes tools on the Zeta and Facebook platforms. Ryohei explained that general concepts that students learn in school, like building out proper market segments and continuously testing and evaluating, are valuable.

Takeaways: AI can identify patterns and automate marketing processes that can make marketing efforts more effective and efficient while guiding customers seamlessly for a better experience with enhanced results.

Ewa Maslowska, Department of Advertising, University of Illinois Urbana-Champaign

Dr. Maslowska’s talk titled: “Teaching Computational Advertising in The ChatGPT Era,” revolved around her expertise in digital environments, personalized communication, algorithmic recommendations, consumer engagement and how to teach these concepts in the classroom.

Dr. Maslowska explained that computational advertising is a hybrid of advertising, marketing, data analytics, and computer science, which is “all about serving the right message, to the right audience, at the right time and place.” Her goal is to understand the tools, their potential, and the challenges they bring.

To prepare students for computational and programmatic advertising, Dr. Maslowska leverages a variety of teaching tools: Academic articles, magazine articles, industry reports, webinars, and case studies, such as the Harvard Business Review on the effectiveness of search advertising. Additionally, she takes advantage of open access and free tools, including:

Kaggle for data

Social Media Macroscope (by U of Illinois Technology Services and the National Center for Supercomputing Applications)

Clarifai full stack AI

CausalImpact (by Google) causal evaluation

Apply Magic Sauce (by the U of Cambridge Psychometrics Centre) digital footprint inferences

Textgain’s Personality Profiler (by textgain.com) writing style analysis

OpenAI GPT-3 Playground (now ChatGPT) prompt reactor

Bard (by Google) conversational tool

DALL.E create art and visuals through voice prompts

Dr. Maslowska shared an assignment she uses in class using the AI platform CausalImpact to assess advertising campaigns. The tasks are as follows: “Use examples discussed in class and a VW case to come up with their own research question/case (e.g., GameStop’s stock).” Preparation includes showing a video on the VW scandal analysis using CausalImpact. Students are shown another video applying CausalImpact analysis to evaluate campaign success. Students are then asked to submit their own RQs, code with results, and a conclusion regarding what they found.

Other classroom assignments included a personalized algorithm audit. Students were assigned readings about targeting consumers and consumer profiling. Next, they were instructed to access “Ad Settings” on their Google, Facebook, and Instagram profiles. After that, they were to create different personas by adjusting interactions with the recommendation algorithms. Finally, they compared their “algorithmic identities” to try to understand the algorithms.

Another assignment used the Apply Magic Sauce platform using digital trace data to predict personal characteristics. Dr. Maslowska asked students to think about their personalities based on the Big 5 Personality Test and discuss them with classmates. They were then instructed to take a short block of text they wrote from an essay or social media post and upload it to Apply Magic Sauce. Next, upload the same block of text to Textgain’s Personality Profiler. Then, upload the same block of text to ChatGPT or a similar platform. Finally, they were tasked with comparing the results.

Dr. Maslowska shared a couple more assignments using various AI platforms to help students learn how to use them and what the tools could do. This hands-on approach of exploring an assortment of AI technology provides students with real-world experience and understanding of the power of AI.

Takeaways: Digital advertising is constantly changing and intersects many disciplines. As educators, there is difficulty in balancing theoretical and practical perspectives. The challenge is there is no access to industry tools or data. Ultimately, AI presents some ethical, legal, environmental, privacy, societal, and bias concerns.

Eunjin (Anna) Kim - University of Southern California, Annenberg School for Communication and Journalism.

As the 2023 Advertising Division’s Early Career Teaching Excellence Award winner, we were eager to hear what Dr. Kim had to say. Her talk was titled, “Analyzing text data with Orange and ChatGPT.” Though she covered a lot of information, her definitions and easy-to-follow demonstrations brought the presentation to life.

Dr. Kim began with defining terms and explaining why certain processes are valuable. She defined Text Mining as “a computational approach to the discovery of new, previously unknown information and/or knowledge through automated extraction of information from a large amount of unstructured text.” The unstructured text includes blogs, consumer responses, social media posts, news articles, emails, memos, photos, videos, administration documents, etc. Dr. Kim explained that up to 80% of available data is unstructured (Deloitte, 2019).

There are multiple benefits to a systematic review of literature as Dr. Kim listed:

• Discovery of new insights

• Computational linguistic research (improving the analysis, understanding and generation of human language from text)

• Document classification

• Clustering/organizing documents

• Visualization of document space

• Making predictions (e.g., predicting stock prices based on the analysis of news articles and financial reports)

• Content-based recommender system (movies, books, articles, etc.)

However, because human language can be ambiguous and often needs contextualization, it is difficult for a computer to analyze, Dr. Kim explains. That’s where Natural Language Processing (NLP) comes in. NLP includes many different techniques for interpreting human language, ranging from statistical and machine learning methods to rules-based and algorithmic approaches, according to Dr. Kim. She further explained that “basic NLP tasks include tokenization and parsing, lemmatization/stemming, part-of-speech tagging, language detection, and identification of semantic relationships.” Tokenization means “splitting text into meaningful elements.”

With this information in mind, Dr. Kim gave the audience some visual lessons on what tokenization and lemmatization, “the algorithmic process of determining the lemma of a word based on its intended meaning,” looks like. She also explained that data preparation includes removing stop words, like “and,” “the,” “a” and similar words that don’t provide much meaning.

Her quote from @BigDataBorat is spot on: “In Data Science, 80% of time spent - prepare data. 20% of time spent - complain about need to prepare data [sic].”

Then, Dr. Kim introduced “Orange”: an open-source machine learning and data mining visualization tool. She led attendees through the process of downloading Orange on their laptops and demonstrated the various components available to create a variety of outputs. Entire folders of text can be uploaded into Orange and directed to create word clouds, sentiment analysis, heatmaps, semantic network analysis, and more. Participants followed along as Dr. Kim demonstrated just how easy it is to create impressive results using the AI tool Orange.

Dr. Kim highlighted more AI data analysis tools, such as Google Trends, Voyant Tools (tutorials) Tableau, and Mention (paid platform). A class activity and an assignment were shared and discussed. The class activity titled, “Effective Boolean Search Terms for Text Data Generation” had students identifying a significant and interesting client issue and its relevance in the field. They extracted keywords and formulated Boolean search terms with pertinent operators. Then, they used Mention to create an alert for their particular problem.

The assignment titled, “Analyzing Client Issues,” is a multistep critical thinking exercise. Students worked through the following steps: (1) Research the client; (2) Identify the problem; (3) Data analysis methodology and rationale; (4) Research analytics, visualization, and interpretation (word cloud, topic models, sentiment analysis, or emotion profile analysis, etc.); (5) Recommendations and conclusions; and, (6) Future research directions. Attendees were glad to learn more about various AI tools and how they can be used in the advertising classroom.

Takeaway: There are so many AI tools that can produce insights much quicker through text mining, natural language processing, and data cleansing. As advertising professors, becoming familiar with AI platforms and developing assignments for students to learn valuable skills will go a long way to prepare them for their professional lives.

The 26th annual Teaching Pre-Conference Workshop organized by the Advertising Division at the AEJMC focused on the application of AI in advertising practice and education. The workshop featured speakers from both professional and academic backgrounds, shedding light on the impact and challenges of AI in the advertising industry as well as how AI tools could be used effectively by practitioners, scholars, and educators.

Advertising and marketing professionals carefully explained AI applications in advertising over the years, highlighting the evolution from a mere concept to a powerful tool. Speakers such as Morgan Bramlet and Kim Herrington demonstrated how popular generative AI tools including ChatGPT-4 and Midjourney can enhance creativity, content creation, and analytics in advertising. They emphasized the fun and creative aspects of using AI tools in advertising and marketing, encouraging a playful approach to technology. Speakers also showcased examples of using AI to generate songs, lyrics, videos, and other creative content. Meanwhile, controversies exist in using AI-generated content. Both Morgan and Eileen Clark Bramlet discussed the ethical and legal concerns regarding AI-generated content, emphasizing the importance of transparency in the creative process. The panel also offers insights into marketing technology, providing knowledge about the use of AI in various platforms for marketers. Ryohei Okabe discussed the customer journey and how AI applications, such as data automation, enhance customer experience and marketing efficiency. Through case studies and simulations, Ryohei highlighted Zeta’s role in automating marketing processes, optimizing messaging, and improving consumer targeting.

In terms of using AI in research and teaching, speakers Dr. Ewa Maslowska and Dr. Eunjin (Anna) Kim shared class assignments and activities to illustrate the effective integration of AI tools including Kaggle, Apply Magic Sauce, and ChatGPT in teaching. Dr. Maslowska discussed the challenges of teaching computational advertising in the ChatGPT era and provided an overview of the ethical, legal, environmental, privacy, societal, and bias concerns associated with AI in advertising education. In addition, Dr. Kim demonstrated the use of Orange, an open-source machine learning and data mining visualization tool, for text data analysis. She led several activities during the panel which allowed the audience to experiment with Orange to analyze consumer data and create client analysis.

The workshop highlighted the multifaceted applications of AI in advertising and marketing, ranging from content generation to data analysis and marketing optimization. Some important trends arise from the panel discussions. First of all, AI can enhance creativity when used in the brainstorming process. In teaching, there could be an emphasis on the creative and playful use of generative AI tools to engage students. Platforms like Midjourney could encourage students to explore their imagination, fostering a positive attitude towards technology.

Secondly, the discussions about ethical and legal issues related to the use of AI will continue to be one of the most important topics for practitioners and scholars. Although there has been a heightened awareness of the legal implications of using AI in advertising, especially in terms of copyright, legal gray areas can create challenges for content creators. Problems such as potential inaccuracies, outdated results, and deep fakes further complicate the ethical implications of using AI. Privacy and bias are also acknowledged as significant challenges, which call for solutions from industry leaders and the government.

Third, AI has been used widely in data analysis and marketing optimization, automating processes like customer segmentation, messaging, and recommendations. In academic settings, this industry trend requires more thoughtful designs of classes that center around computational advertising, which may involve an interdisciplinary collaboration from advertising, marketing, data analytics, and computer science. Furthermore, educators should consider leveraging various AI tools and platforms such as Orange, Google Trends, and Voyant Tools to better prepare students for the industry.

Given the swift and pervasive evolution of AI as a communication tool with consequential outcomes, it is imperative to examine best uses in academia and in the field. We would like to thank all of our prestigious speakers for sharing their knowledge and insights. Special thanks to our sponsor, the Walter Cronkite School of Journalism and Mass Communication, Arizona State University, for the use of the ASU Washington Center. We would also like to express our gratitude to Dr. Juan Mundel for coordinating the pre-conference meeting space. We would like to acknowledge all attendees and their important contributions to our conversation. The pre-conference committee is honored to uphold the Advertising Division legacy of the longest running, full day teaching pre-conference at the annual convention of AEJMC. The planning committee included: Dr. Shanshan Lou, Appalachian State University and Professor Robin Spring, Grand Valley State University.