Columns November/December 2021

Looking Under the Hood: Automated Content Production

By Thomas Barker

Over 240 years ago, a Swiss clockmaker named Pierre Jaquet-Droz constructed a doll, named The Writer,1 that could produce script handwriting, on paper with an ink pen, according to programmable mechanisms. The result was an early form of a tweet, only with a 40-character limit.

It may seem like something of a stretch, but today’s equivalent of this marvel can be seen in the nearly 200 content management startups represented by AngelList2 as of September 2021. These companies and agencies — in many ways the growth industry in technical communication today — are all about headless content, automated video content, AI assisted storytelling, and buzz-based marketing content. Like The Writer, they employ both humans and writing machines in their work.

In this column we will look at the academic conversation surrounding automated content production and the research underpinnings in AI that sustain it. We will uncover some key research concepts that can help content managers keep their writing machines tuned up. Let’s take a look under the hood.

Some Terminology

The vocabulary of automated content production includes words to describe the outcome of these cognition algorithms: content automation, automated content management, enterprise content management, and automated writing assistance. These terms remind us of the ubiquity of automation in the process of content creation. One important application is in content marketing3: a very trendy tactic that focuses on creating customers through engaging content instead of simply pitching products and services.

According to researcher Loredanna Patrutiu Baltes, “Given that digital marketing requires the existence of a content marketing, the success or the failure of the company’s online communication depends to a significant extent on the quality of its content marketing.”4

Why is this so? At its heart, content marketing is about establishing relationships first. Sales come second. Further, content marketing is key to market differentiation in digital marketing5 (using social media and web-based channels). Key to market differentiation is target analysis, which is often assisted by AI and data mining. For content producers, the logical pivot is to automated content production: cognitive artifacts admixed with software and digital components.

The engine that drives the branding and storytelling artifacts that characterize marketing is the analytical capability under the hood. These analytics provide the initial shape and flexibility of target audiences.

According to Bellini and Nesi, content production, the heart of content marketing, can be more efficient and cheaper if it is supported by artificial intelligence algorithms for content composition, formatting, protection, and workflow management, among other efficiencies.6

Automated Writing

Nothing can replace a team of innovative writers and technologists, journalists, IT professionals, and policy thinkers. But scan any list of management startups and you will find artificial intelligence platforms at work under the hood helping groups and businesses attract members, promote events, and sell products more effectively.

As one of them put it, “let the platform do the heavy lifting.” The Wordtracker blog stops just short of saying AI can replace writers, but notes that it can “help in the process.” In this supportive way, content automation is a part of any content management and development strategy. It uses the algorithmic and rule-based intelligence of software like Grammarly and Wordsmith to do this.

So today a community manager might use smart enablement and onboarding software to create user group profiles. The question this use poses is: how far can the automation of content production go? Will it continue only in a support role, or can it eventually lead to the production of semantic artifacts?

What is interesting to us, apart from the fascination of watching The Writer click through his little writing task, his eyes moving in a lifelike way to follow his automated quill, is the perfection of the technologies — metals, miniaturization, timing — that Jaquet-Droz used in making this marvel. Without these technologies a writing machine would have been an imaginative product, imbued with magic or spells. But it is not. The turning of the wheels inside the doll vouchsafes the authenticity of the mechanism.

The technological advances — in engineering, materials, and know-how — that afforded for this human-like communicator have their modern counterpart in applications like GPT-37 that can easily pass mathematical, semantic, and ethical tasks for the creation of life-like prose.

If we look at the automated writers of today, we enter the world of content creation through artificial intelligence and natural language generation (NLG). The clinking escape mechanisms of the past are silent, frictionless algorithms working away behind dashboards and visualizations instead of a doll face.

It may be that research into automated content or automated contents support systems looks now just like a thing of the future. However, content managers today can easily spot trends in machine learning, artificial intelligence, and serverless architectures at academic conferences. These topics are harbingers of the advances it will be made by these technologies in the development of automated content systems. Writing that looks and sounds like it came from a real person ride the wave of these scientific explorations.8

The Research Agenda in Communication

These considerations of the research agendas surrounding automated content production and management raise even larger issues having to do with communication in general. Recent research in artificial intelligence end communication have begun to reflect that we are no longer simply communicating between one human and another, but between humans and machines.

We may be facing an era in which people may communicate less with other people and more with machines that act like other people. As researchers Guzman and Lewis point out, people now converse with Amazon’s Alexa and Apple’s Siri end with other items and appliances associated with The Internet-of-Things (IoT) as if they were people. They even note that the Associated Press is “using AI-enabled technologies in the production and distribution of news.” Entrepreneurs in the content industry will use or compete with agencies that employ AI in content management in these ways. The most lifelike member of your content team may be an AI-driven bot.9

Human-Machine Communication: The New Frontier

Communication scholars are beginning to shift their focus of study from the human to the machine as the communicative subject. Entangled in this new web of relationships is talk between people and robots, other people and agents. The new frontier for communications research is communicative AI: conversations that used to be between humans and humans but are now between humans and machines directly (mediated, as before, by computers.) The computer is not out of the picture in its older role as the channel of communication, but the communicating subject is now the computer itself. I’m hearing, “So Dave, any luck with those pod-bay doors?”

More likely this dialogue will be between the reader and an automated blog, news article, or product overview. As before, the author, real or algorithmic, may remain anonymous. But not for researchers. They ask ontological questions: what happens down the road when our body of knowledge includes a library of AI-generated, natural-sounding research articles, white papers, and government policy documents? Guzman and Lewis focus on journalism and questions about authority in journalism. For example, they outline an agenda, focusing on the functional, relational, and metaphysical questions surrounding automated content.

For entrepreneurs and content creators, the issues might be more down to earth: investments in AI cost more or less than hiring a good writer; data mining, analytics, and IoT processing might just be technical problems; or content marketing using AI-assisted content development poses planning challenges. But the research community, and its academic conversations, can bring new value for managers and industry analysts, who look below the surface of the text at the algorithmic fingers tapping away.

In this column we have looked at the issue of automation in writing. As we saw, these technical advances, like those in older times, depend on undercurrents of research. Automated writing is such a technology that is beginning to show up in the form of process support for writing and cognitive artifacts themselves, texts written by computer algorithms. Communication scholarship is shifting in response to these more mature technologies. We are a long way from these cognitive artifacts, but their emergence prompts us to look at the scholarship and research at work under the hood.

References
  1. n.d. “The Writer Automaton, Switzerland.” https://www.chonday.com/15454/the-writer-automaton/.
  2. AngelList Venture. n.d. https://www.angellist.com.
  3. Content Marketing Institute. n.d. “What is Content Marketing.” https://contentmarketinginstitute.com/what-is-content-marketing/.
  4. Bates, L.P. 2015. “Content Marketing-The Fundamental Tool of digital marketing. Bulletin of the Transilvania University of Brasov. Economic Services, no. 8(2): 111.
  5. Lucy, Alexander. 2021. “The Who, What, Why, and How of Digital Marketing.” HubSpot. Last modified September 29, 2021. https://blog.hubspot.com/marketing/what-is-digital-marketing.
  6. Bellini, P. and P. Nesi. (2005). “An Architecture of Automating Production of Cross Media Content for Multi-Channel Distribution.” In First International Conference on Automated Production of Cross Media Content for Multi-Channel Distribution (AXMEDIS’05): 11.
  7. Maldonado, A. and Natalie Pistunovich. 2021. “GPT-3 Powers the Next Generation of Apps.” OpenAI. https://openai.com/blog/gpt-3-apps/.
  8. Tyagi, M., M. Sharma, and P. Sharma. (2020). “The Future of the Web.” In Proceedings of the International Conference on Innovative Computing & Communications (ICICC).
  9. Guzma, Andrea and Seth Lewis. 2019. “Artificial Intelligence and Communication: A Human–Machine Communication Research Agenda.” Sage Journals. https://doi.org/10.1177/1461444819858691.

 

This column focuses on a broad range of practical academic issues from teaching and training to professional concerns, research, and technologies of interest to teachers, students, and researchers. Please send comments and suggestions to column editor Thomas Barker at ttbarker@ualberta.ca.