By Suro Senen 
In the lexicon of the modern era, the words "think," "know," "understand," and "remember" serve as the bedrock of human experience. They describe the internal theater of the mind, the cognitive processes that define our sentience. However, as artificial intelligence (AI) integrates into every facet of our digital and professional lives, these terms are increasingly being co-opted to describe algorithms. This linguistic shift—while perhaps convenient—risks fundamental misunderstandings about the nature of the systems we build and rely upon.
A new study, "Anthropomorphizing Artificial Intelligence: A Corpus Study of Mental Verbs Used with AI and ChatGPT," published in the Technical Communication Quarterly, suggests that while we frequently caution against "humanizing" machines, the reality of how we write about them is far more complex than previously assumed.
The Mechanics of Anthropomorphism
Anthropomorphism, the practice of attributing human traits, emotions, or intentions to non-human entities, is a psychological shorthand. We name our cars, we talk to our house plants, and we increasingly project agency onto the large language models (LLMs) that populate our screens.
"We use mental verbs all the time in our daily lives, so it makes sense that we might also use them when we talk about machines—it helps us relate to them," says Jo Mackiewicz, a professor of English at Iowa State University and lead author of the study. "But at the same time, when we apply mental verbs to machines, there’s a risk of blurring the line between what humans and AI can actually do."
The research team, which included Jeanine Aune (Iowa State University), Matthew J. Baker (Brigham Young University), and Jordan Smith (University of Northern Colorado), sought to quantify this phenomenon. They argue that when news outlets or technical writers use phrases like "AI decided" or "ChatGPT knows," they inadvertently suggest that these systems possess internal beliefs, motivations, or consciousness.
In reality, these systems are advanced statistical engines. They produce responses by analyzing massive datasets to predict patterns, not by forming independent ideas or navigating ethical dilemmas. By masking this computational reality with human-centric verbs, the public may develop unrealistic expectations regarding AI’s reliability, accountability, and objectivity.
Chronology of the Linguistic Study
The investigation into this phenomenon was born out of a growing need for media literacy in the age of generative AI. As the rapid deployment of systems like ChatGPT outpaced public understanding, the researchers recognized that the narrative framing of these tools was becoming a primary source of misinformation.
- Phase One: Conceptualization: The team identified the prevalence of "mental verbs"—words describing cognitive states—as the primary vehicle for anthropomorphizing AI.
- Phase Two: Data Collection: To establish a representative sample, the researchers turned to the News on the Web (NOW) corpus, a gargantuan repository of more than 20 billion words derived from English-language news articles spanning 20 countries.
- Phase Three: Quantitative Analysis: The team searched for correlations between AI-specific terms (such as "AI," "ChatGPT," "machine learning") and a targeted list of mental verbs ("learns," "thinks," "knows," "understands," "wants").
- Phase Four: Qualitative Contextualization: Perhaps most importantly, the team analyzed the context of these findings, recognizing that a word’s usage in a sentence dictates its impact on the reader.
Supporting Data: The Surprising Restraint of Journalists
Contrary to the popular belief that the media is rampant with AI-driven personification, the study revealed a surprising level of professional restraint. While anthropomorphism is a staple of casual conversation, it appears significantly less frequently in professional news writing than the researchers initially hypothesized.
The data provided by the NOW corpus indicated that mental verbs are not as frequently paired with AI-related terms as might be expected in a post-ChatGPT world. For instance, the verb "needs"—which can be interpreted as either a biological necessity or a technical requirement—was the most frequent pairing for "AI," appearing 661 times. In contrast, "knows," a verb that strongly implies cognitive sentience, appeared with "ChatGPT" only 32 times across the massive dataset.
"Anthropomorphism has been shown to be common in everyday speech, but we found there’s far less usage in news writing," says Mackiewicz.
The researchers attribute this, in part, to the rigorous editorial standards maintained by major news organizations. The Associated Press (AP) and other style guides have begun to explicitly advise journalists against attributing human traits to algorithms. This guidance acts as a structural barrier to the proliferation of misleading anthropomorphic language, forcing writers to adopt more technical, objective terminology.
The Spectrum of Context: When "Need" Doesn’t Mean "Want"
A critical discovery of the study is that linguistic context matters far more than the specific verb chosen. The word "needs," for example, often functions as a neutral descriptor of mechanical requirements rather than an indicator of desire.
When a news outlet writes, "AI needs large amounts of data to function," the usage is analogous to saying "a car needs gasoline." It describes a functional prerequisite. Similarly, phrases written in the passive voice—such as "AI needs to be implemented"—often serve to shift the focus away from the machine and back toward the human developers, engineers, and organizations that hold the actual agency.
However, the researchers caution that anthropomorphism exists on a spectrum. At one end, there are neutral, functional descriptions. At the other end, there are phrases like "AI needs to understand the real world." This specific phrasing crosses a line, suggesting that the system is capable of "understanding"—a cognitive process—rather than just "processing" or "calculating." These nuances are exactly where public perception is most easily swayed.
Implications for the Future of AI Literacy
The implications of this research extend far beyond the newsroom. They touch upon how society assigns responsibility for AI-driven outcomes. If we describe AI as having "intentions," we implicitly diminish the responsibility of the humans who designed the system.
"Certain anthropomorphic phrases may even stick in readers’ minds and can potentially shape public perception of AI in unhelpful ways," notes Jeanine Aune.
As AI tools become increasingly sophisticated, the challenge for journalists, technical writers, and even corporate communicators will be to maintain a delicate balance. Writers must convey the utility and power of these tools without resorting to the "ghost in the machine" tropes that simplify complex engineering into human-like narratives.
The research team suggests that future academic work should focus on the impact of this language on the reader. Does a reader who consumes news describing AI as "knowing" information perceive the output of that system as more authoritative than one who reads about AI "retrieving" data? Understanding the causal link between word choice and public trust is the next frontier in this field of study.
Final Reflections for Professionals
For technical and professional communication practitioners, the message is clear: language is not a neutral vessel. By reflecting on their own writing processes, professionals can choose words that accurately depict AI as a tool rather than an entity.
"Our findings can help technical and professional communication practitioners reflect on how they think about AI technologies as tools in their writing process and how they write about AI," the authors conclude.
As we continue to integrate these technologies into our society, our vocabulary must evolve to be as precise as the code that powers the machines themselves. By maintaining a distinction between human cognition and algorithmic processing, we ensure that we remain the masters of our tools, rather than the subjects of their supposed "intelligence."
Basiran 
Sagoh 
Lina Hope 