By Nana 
In the digital age, the lexicon of human cognition—words like "think," "know," "understand," and "remember"—has migrated from the realm of biology to the silicon architecture of artificial intelligence. While these mental verbs help us bridge the gap between complex technology and everyday experience, they carry a hidden cost: they can inadvertently suggest that machines possess consciousness, intentions, or moral agency.
A recent study published in Technical Communication Quarterly titled "Anthropomorphizing Artificial Intelligence: A Corpus Study of Mental Verbs Used with AI and ChatGPT" challenges the assumption that our media is saturated with this human-centric framing. By analyzing billions of words in the News on the Web (NOW) corpus, researchers have uncovered a nuanced reality about how the fourth estate reports on the rapidly evolving AI landscape.
The Mechanics of Anthropomorphism
Anthropomorphism, the act of assigning human traits to non-human entities, is a well-documented linguistic phenomenon. When a news outlet reports that "ChatGPT knows the answer" or "AI decided to optimize the code," it frames the software not as a probabilistic engine, but as an autonomous agent.
"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 also a risk of blurring the line between what humans and AI can do."
The danger, according to the research team—which includes Jeanine Aune (Iowa State), Matthew J. Baker (Brigham Young University), and Jordan Smith (University of Northern Colorado)—is that this language creates a false impression of technological capability. AI systems are, by design, pattern-recognition engines that process data to produce output; they do not possess beliefs, feelings, or conscious decision-making faculties. By attributing "thought" to these systems, we risk creating unrealistic expectations regarding their reliability and obscuring the human developers and engineers who are ultimately responsible for the system’s performance.
A Chronology of Linguistic Analysis
The research project was born from a growing concern among communication scholars regarding the rapid integration of Large Language Models (LLMs) into the public consciousness.
- Phase One: Conceptualization. The team identified the prevalence of mental verbs in casual conversation and hypothesized that news media—often influenced by public discourse—would mirror this trend.
- Phase Two: Data Aggregation. Utilizing the News on the Web (NOW) corpus, the team accessed a repository of over 20 billion words spanning news articles from 20 countries.
- Phase Three: Quantitative Filtering. The researchers targeted specific mental verbs—such as "learns," "means," and "knows"—to see how often they appeared in proximity to "AI" or "ChatGPT."
- Phase Four: Qualitative Contextualization. Recognizing that word count alone is misleading, the team performed a deep dive into the context of these verbs to distinguish between true anthropomorphism and purely functional, non-human usage.
- Phase Five: Synthesis. The findings were finalized, revealing that the "anthropomorphic tide" was, in fact, significantly lower than initial hypotheses suggested.
Supporting Data: Debunking the Prevalence Myth
The study’s most striking finding is the relative scarcity of anthropomorphic language in professional journalism. Contrary to the belief that news outlets are constantly personifying AI, the data suggests that editorial gatekeeping is working.
"Anthropomorphism has been shown to be common in everyday speech, but we found there’s far less usage in news writing," says Mackiewicz.
The data confirms this disparity. While "needs" was the most common verb paired with AI, appearing 661 times, many of these instances were purely functional—e.g., "AI needs more data." In contrast, the personification of ChatGPT was remarkably low, with the verb "knows" appearing only 32 times across the massive 20-billion-word dataset.
The researchers attribute this restraint to the influence of industry standards. Guidelines from major organizations, such as the Associated Press, explicitly caution journalists against attributing human emotions or motivations to non-human systems. This adherence to style guides appears to act as a significant barrier against the linguistic "creeping" of anthropomorphism.
Contextual Nuance: The Spectrum of Intent
One of the study’s most sophisticated contributions is its rejection of binary classifications. The researchers argue that anthropomorphism exists on a spectrum.
The Functional Utility of "Needs"
When a journalist writes, "AI needs to be trained," the verb "needs" is stripped of its emotional or human intent. It describes a requirement similar to saying a "car needs gasoline." It is a mechanical dependency rather than a psychological desire.
The Passive Voice and Responsibility
Aune highlights that in many instances where mental verbs are used, the sentence structure is deliberately constructed to maintain human oversight. Phrases such as "AI needs to be implemented" or "AI needs to be audited" utilize the passive voice, which effectively pivots the reader’s attention back to the humans—the engineers, policymakers, and corporate leaders—who hold the true agency over the system.
The Danger Zone
Conversely, the study identifies instances where language slips into problematic territory. A phrase like "AI needs to understand the real world" suggests an expectation of ethical reasoning or cognitive awareness. These instances are, according to the researchers, the most dangerous, as they suggest the machine has a responsibility that it cannot possibly meet, potentially leading to a misplaced sense of trust or accountability.
Implications for the Future of AI Literacy
The research team emphasizes that the goal of the study is not to police language, but to encourage a higher level of "AI literacy" among those who write about technology.
For Journalists and Content Creators
The study serves as a call to action for professional communicators. Writers must recognize that their choice of words directly shapes public policy and public perception. By opting for precise, mechanical language over evocative, human-like verbs, writers can help maintain a clear distinction between technological tools and human actors.
For Public Perception
Public perception of AI is in a state of flux. When a society begins to view an algorithm as a "thinking" entity, it may become more susceptible to scams, manipulation, or the belief that an AI’s output is an objective, unbiased truth. Linguistic precision is, therefore, a matter of public safety.
For Future Research
The team suggests that the next frontier of this research lies in measuring the impact of these words. While we now know that anthropomorphic language is used sparingly in news, we do not yet know how even the occasional use of these words influences the average reader’s perception of an AI’s trustworthiness or autonomy.
Final Reflections
The study concludes on a note of cautious optimism. The fact that news organizations are largely resisting the urge to humanize AI suggests that institutional guardrails are effective. However, as AI systems become more integrated into our daily lives—through voice assistants, personalized tutors, and automated customer service—the temptation to use mental verbs will only increase.
"For writers, this nuance matters: the language we choose shapes how readers understand AI systems, their capabilities, and the humans responsible for them," Mackiewicz notes.
Ultimately, the ghost in the machine is a projection. As long as humans remain the creators, maintainers, and final arbiters of AI systems, our language must reflect that reality. By remaining mindful of the verbs we choose, we ensure that our discourse remains anchored in technical reality rather than drifting into the realm of science fiction. The study by Mackiewicz and her colleagues provides a vital roadmap for navigating this linguistic challenge, reminding us that while AI may simulate intelligence, it remains—and will remain—a product of human design.
Asro 
Lina Hope 
Ali Ikhwan 