By Lina Hope 
In the digital age, we routinely assign human cognitive traits to non-human systems. We say that an algorithm "thinks," a large language model "knows" an answer, or a predictive tool "understands" our preferences. While these "mental verbs" make the complexities of machine learning accessible and relatable, a new study suggests that this linguistic habit—known as anthropomorphism—may be subtly distorting our understanding of reality, obscuring the nature of technology, and shifting the burden of accountability away from the humans behind the screen.
The Illusion of Awareness: A Conceptual Trap
The human mind is wired for pattern recognition and social connection, often projecting agency onto inanimate objects to make sense of the world. When we describe AI through the lens of human experience, we inadvertently attribute beliefs, intentions, and conscious decision-making to systems that are fundamentally devoid of them.
"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," explains Jo Mackiewicz, a professor of English at Iowa State University. "But at the same time, when we apply mental verbs to machines, there’s also a risk of blurring the line between what humans and AI can do."
The core issue lies in the distinction between simulation and cognition. An AI does not "know" in the way a sentient being knows; it executes complex probabilistic calculations based on vast datasets. When a headline declares that "ChatGPT decided to change its output," it confers an independent agency that simply does not exist. This creates a psychological gap where users may begin to expect human-level ethics or reliability from a system that is, at its core, a sophisticated pattern-matching engine.
Researching the Narrative: The Corpus Study
To quantify how frequently these humanizing terms infiltrate our discourse, a research team led by Mackiewicz and Jeanine Aune—director of the advanced communication program at Iowa State—conducted an extensive linguistic analysis. Their study, "Anthropomorphizing Artificial Intelligence: A Corpus Study of Mental Verbs Used with AI and ChatGPT," published in Technical Communication Quarterly, sought to peel back the layers of how modern media depicts machine intelligence.
The team included Matthew J. Baker, an associate professor of linguistics at Brigham Young University, and Jordan Smith, an assistant professor of English at the University of Northern Colorado. Together, they turned to the News on the Web (NOW) corpus, a gargantuan repository containing over 20 billion words derived from English-language news articles across 20 countries.
Their methodology involved a rigorous tracking of mental verbs—words like "learns," "means," "knows," "thinks," and "wants"—when paired with terms like "AI" or "ChatGPT." The goal was to determine if the mainstream press was inadvertently fostering a "Terminator-style" misconception of AI by granting it human-like cognitive capabilities.
Data Findings: A Surprising Restraint in Journalism
The results of the study were, in several ways, counterintuitive. Despite the common assumption that AI is consistently personified in the media, the research revealed that news writers are far more disciplined than the general public.
"Anthropomorphism has been shown to be common in everyday speech, but we found there’s far less usage in news writing," says Mackiewicz. While anecdotal evidence might suggest that AI is constantly described as a sentient peer, the hard data from the NOW corpus indicates a surprising level of linguistic caution.
For instance, among the identified mental verbs, the word "needs" appeared most frequently in relation to AI, appearing 661 times. However, the researchers discovered that the term "needs" is rarely used to imply human desire. Instead, it functions as a functional descriptor: "AI needs large amounts of data" or "AI needs computing power." These instances parallel how we discuss non-sentient tools, such as a car needing gasoline or a recipe needing heat.
When focusing on specific AI products like ChatGPT, the frequency drops significantly. The word "knows" was the most common verb paired with ChatGPT, yet it appeared only 32 times across the vast dataset. This suggests that professional editorial standards—such as those enforced by the Associated Press, which explicitly discourage the attribution of human traits to non-human systems—are effectively curbing the most egregious forms of anthropomorphism in journalism.
Contextual Nuance: Beyond Simple Word Counting
A critical component of the study was the realization that context is the ultimate arbiter of meaning. The research team found that anthropomorphism is not a binary state but rather a spectrum.
"These instances showed that anthropomorphizing isn’t all-or-nothing and instead exists on a spectrum," notes Aune. At the lower end of the spectrum, phrases like "AI needs to be trained" or "AI needs to be implemented" are written in the passive voice, which functions to shift the focus back to the human actors—the developers and engineers—who are responsible for the system’s operation.
Conversely, at the higher end of the spectrum, phrasing becomes more problematic. Statements like "AI needs to understand the real world" or "AI thinks it has solved the problem" suggest a depth of reasoning, awareness, or ethical capacity that does not exist. These phrases represent a shift from describing a tool to describing a conscious actor, potentially misleading the public about the reliability of the system.
The Dangers of Misplaced Agency
Why does this matter? If the average reader knows, on an intellectual level, that a machine isn’t "thinking," why worry about the terminology?
The researchers argue that the stakes are higher than mere semantics. By consistently describing AI as if it has intentions, the media risks distracting the public from the true source of decision-making: the humans behind the algorithms. When an AI is described as "deciding" to deny a loan or "understanding" a piece of literature, the accountability for that decision becomes diffused.
"Certain anthropomorphic phrases may even stick in readers’ minds and can potentially shape public perception of AI in unhelpful ways," Aune warns. If we view AI as an independent, thinking entity, we are less likely to interrogate the biases inherent in its training data or the corporate interests that dictate its design. We begin to treat the "output" of a machine as an objective, independent truth rather than a manufactured product of human design.
Professional Implications: Toward More Precise Communication
The findings of this study offer a roadmap for technical writers, journalists, and public relations professionals. Rather than viewing language as a secondary concern, those who communicate about AI must recognize it as a fundamental tool of public policy and safety.
The research team emphasizes that the goal is not to eliminate all figurative language, but to practice intentionality. When writing about AI, professionals should prioritize verbs that reflect mechanical processes—such as "analyzes," "processes," "calculates," or "generates"—over verbs that imply consciousness.
"For writers, this nuance matters: the language we choose shapes how readers understand AI systems, their capabilities and the humans responsible for them," Mackiewicz asserts. By shifting the linguistic frame, writers can help demystify the technology, ensuring that the public remains focused on the human-driven systems of accountability rather than the illusion of machine consciousness.
Future Horizons: The Evolution of AI Discourse
As artificial intelligence systems grow increasingly sophisticated, the temptation to use anthropomorphic language will likely intensify. We are entering an era where AI can mimic human tone, style, and emotional expression with uncanny accuracy. This makes the work of scholars like Mackiewicz and Aune more vital than ever.
The team suggests that future research should delve deeper into the impact of these language choices on reader cognition. Does reading about a "thinking machine" actually change a user’s behavior or their trust levels? Does it influence their willingness to rely on AI for critical tasks, such as medical diagnoses or legal counsel?
For now, the study serves as a necessary corrective, highlighting that while our machines may be getting smarter, our language about them remains the most powerful tool we have for maintaining a grip on reality. By choosing our words with precision, we ensure that we remain the masters of our tools, rather than becoming the passive observers of a projected intelligence that does not, and cannot, exist.
Ammar Sabilarrohman 
Ali Ikhwan 
Iffa Jayyana 