The Language of AI: How Words Shape Our Perception of Machines
Ever caught yourself saying, ‘ChatGPT knows the answer’ or ‘AI thinks this way’? If so, you’re not alone. But here’s the kicker: when we use human-like language to describe AI, we’re often blurring the line between what machines do and what they are. This isn’t just a semantic quibble—it’s a fundamental issue that shapes how we perceive, trust, and interact with technology.
The Humanization of Machines: A Double-Edged Sword
Let’s start with the core idea: anthropomorphism. It’s a fancy word for giving human traits to non-human things. When we say AI ‘understands’ or ‘decides,’ we’re unintentionally making it sound like a thinking, feeling entity. Personally, I think this is where the trouble begins. What many people don’t realize is that AI doesn’t ‘think’ in the way we do. It processes data, recognizes patterns, and spits out responses based on algorithms. There’s no consciousness, no intent, no inner monologue.
What makes this particularly fascinating is how easily we slip into this language. As Jo Mackiewicz, a professor of English, points out, we use mental verbs like ‘think’ and ‘know’ all the time in our daily lives. It’s natural to extend them to machines, especially when we’re trying to make sense of something as complex as AI. But here’s the catch: this language can create a false equivalence. When we say ‘AI knows,’ we’re implying a level of awareness that simply isn’t there.
The Misleading Power of Words
One thing that immediately stands out is how this kind of language can inflate expectations. If AI ‘knows’ something, it must be infallible, right? Wrong. AI is only as good as the data it’s trained on, and it can make mistakes—sometimes catastrophic ones. From my perspective, this is where the real danger lies. When we overstate AI’s capabilities, we set ourselves up for disappointment, or worse, blind trust in systems that aren’t ready for it.
Take the phrase ‘AI decided.’ It suggests autonomy, as if the machine made a choice independent of human input. But what this really suggests is that we’re shifting responsibility away from the people who built and trained the system. Developers, engineers, and organizations are the ones pulling the strings, yet the language we use often obscures their role. If you take a step back and think about it, this isn’t just a linguistic issue—it’s a societal one.
The Surprising Reality of AI Language in Media
Here’s where things get interesting: a recent study by Mackiewicz, Jeanine Aune, and their colleagues found that anthropomorphism in news writing about AI is less common than we might think. After analyzing a massive dataset of news articles, they discovered that mental verbs like ‘knows’ or ‘learns’ were rarely paired with AI terms. For instance, ‘needs’ was the most common verb used with AI, but even that appeared only 661 times out of billions of words.
A detail that I find especially interesting is how context matters more than the words themselves. When journalists wrote ‘AI needs data,’ they weren’t implying that AI has desires. They were simply stating a requirement, much like saying a car ‘needs’ fuel. This nuance is crucial because it shows that not all uses of mental verbs are created equal.
The Spectrum of Anthropomorphism
What many people overlook is that anthropomorphism isn’t black and white—it’s a spectrum. Some phrases, like ‘AI needs to understand the real world,’ come closer to suggesting human-like qualities. These instances raise a deeper question: are we setting unrealistic expectations by implying that AI can reason or feel? In my opinion, this is where writers need to tread carefully. Language isn’t just about description; it’s about shaping perception.
This raises another point: editorial standards play a huge role. The Associated Press, for example, discourages attributing human traits to AI. This could explain why anthropomorphism is less prevalent in news writing than in casual conversation. But even with these guidelines, the potential for misuse remains. As AI becomes more integrated into our lives, the way we talk about it will only grow in importance.
Why This Matters for the Future
If there’s one takeaway from this research, it’s that language isn’t neutral—it’s powerful. The words we choose to describe AI influence how we understand its capabilities, limitations, and the humans behind it. Personally, I think this is a call to action for writers, journalists, and anyone who talks about technology. We need to be more mindful of how our language shapes public perception.
Looking ahead, I’m curious about how this will evolve. Will we see more nuanced language as AI becomes more advanced, or will anthropomorphism increase as machines become more human-like in their interactions? One thing’s for sure: the conversation is far from over. As Mackiewicz and Aune suggest, future studies could explore how even rare uses of anthropomorphic language impact public understanding.
Final Thoughts
In the end, the way we talk about AI isn’t just about semantics—it’s about accountability, expectations, and the future of human-machine interaction. From my perspective, the key is to strike a balance: acknowledge AI’s capabilities without attributing human qualities it doesn’t possess. After all, AI is a tool, not a mind. And how we describe it will determine how we use it—for better or for worse.
So, the next time you hear someone say ‘AI knows,’ remember: it’s just a machine. And the real story is the humans behind it.
