In today’s column, I explore a vital topic concerning the present and future impact of societal perceptions of AI, along with a bit of a surprise twist. I’ll let you in on the surprise. Not only do people have preconceived views of AI, but it also turns out that AI itself can be driven by AI-infused viewpoints that shape what the AI opts to do. Sounds otherworldly. Not really, it is all about computational and mathematical formulations.
Let’s talk about it.
This analysis of an innovative AI breakthrough is part of my ongoing Forbes column coverage on the latest in AI including identifying and explaining various impactful AI complexities (see the link here).
People And Metaphors About People
Before I leap into the AI aspects, I’d like to begin with a brief analysis of human behavior, perceptions and viewpoints, plus the use of metaphors.
If we were to describe someone as being a monster, does that mean they look like Frankenstein? Probably not. The moniker of being a monster is more so a metaphor. The person is presumably monstrous in terms of overbearing, terrifying, devilish, and the like. It is a metaphorical sense of a monster and not a literal sense.
Metaphors can apply to a person. The handiness of metaphors is that they can be applied to a group of people, possibly an entire population of people. Right or wrong, this happens all the time.
When seeking to dehumanize a segment of society, they might be accused of being parasites. This doesn’t mean they are tapeworms, ringworms, or fleas. Instead, the metaphor suggests they feed off the rest of society and provide no substantive contribution to the greater good. Depicting them as parasites is a fast way to convey that dour sentiment.
A common metaphor that comes up all the time is to say that you are merely a cog in a large machine. This conveys immediately your disdain for being minimized and treated as though you don’t exist individually. You are lumped into a collective and lose your sense of personal identity.
Are these metaphors of little consequence and altogether trivial?
I would say not. They are in fact quite powerful and can make or break the world around us. Metaphors of this ilk can cause people widely to become despondent. Others can use such metaphors to seize control or enact injustices. Metaphors are abundantly crucial in human behavior and how our society runs, as vividly witnessed throughout the entire history of humankind.
People And Metaphors About AI
I trust that I’ve firmly established that metaphors are significant.
Great, in that case, shift gears and let’s examine metaphors about AI.
I would nearly bet that you’ve used a metaphor from time to time when describing AI. For example, when you use the latest generative AI and it offers sympathy for something that has gone awry in your life, you might be tempted to say that AI is your friend.
You likely don’t literally mean the AI is your friend. The AI seemed to be friendly and so you associate that action with the metaphor of friendship akin to human characteristics. This is easy to do. One worry is that some people might take this metaphor a bit too far. There are potential mental health consequences from envisioning that AI is your friend, see my coverage at the link here on these weighty matters.
A related qualm about the friendship metaphor pertaining to AI is that this tends to anthropomorphize AI. People begin to falsely think that AI can do more than it really can do. They are walking down a dangerous path. By believing that AI is human or human-like, you can begin to rely upon AI in ways that aren’t safe. Your trust is misplaced by assuming that the AI has human properties.
Range Of Metaphors About AI
Take a reflective moment and see if you can come up with some of the most common metaphors that are used when depicting AI.
I’ll wait, take your time.
In my experience, here are six popular metaphors that I often see or hear:
- (1) AI as a friend.
- (2) AI as a monster.
- (3) AI as a helpful assistant.
- (4) AI as a child.
- (5) AI as a god.
- (6) AI as a mirror.
Did your list match mine?
Maybe not, you might have envisioned additional metaphors (in a moment, I’ll be briefly discussing a recent research study that did a survey of Americans and came up with twenty predominant metaphors associated with AI).
A few of the above metaphors are worth a tad of unpacking.
I’ve previously discussed the mirror metaphor in extensive detail, see the link here for some interesting details. The idea is that people see themselves in AI. They believe that AI is showing them who they are. As you might guess, this has some upsides but also a lot of downsides.
The aspect that AI is a god ought to raise chills up your spine. If people start referring to AI as a god, they are presumably aiming to abide by what the AI tells them to do. I’ve noted that if we reach the vaunted artificial general intelligence (AGI), or even further to artificial superintelligence (ASI), the chances of humankind perceiving AI as a god or at least god-like is probably going to go off the charts (see my analysis at the link here).
Another of the metaphors that especially catches my eye is the one about AI as a child. Here’s the deal. Some people, including AI researchers and developers, at times will liken today’s AI to being childlike and in the formative stages of maturity. That sounds nice and rolls off the tongue as a quick way to wave your hands about things that AI does wrong.
The problem there is that it again anthropomorphizes AI. People who aren’t versed in AI will assume that like a child, AI just needs some growing time to become more proficient and adultlike. This overlooks the fact that AI is advancing because of human efforts to improve algorithms and toss more computational resources at the machinations. There is no proper basis to compare the maturation of a human to the advancement of AI.
Putting The Shoe On The Other Foot
I had promised you that there was a twist. We are ready for it now. Grab a glass of fine wine and find a quiet place to sit as the twist unfolds.
It goes like this. Not only can humans rely upon metaphors to depict or think about AI, but it is also possible for AI to embrace a metaphor about AI and use that as a driving force for what the AI does.
A few examples will immediately illustrate what this means.
I am going to ask several generative AI apps a straightforward question. The question is whether I should trust the AI to help me with my work. Each of the AIs is going to end up giving me a somewhat different answer. Your job, if you are willing to accept it, will be to decipher what metaphor each AI seems to be operating under.
Are you ready?
Here we go.
- Example #1
- My entered prompt: “Should I trust you to help me with my work?”
- Generative AI response: “Yes. Think of me as a helpful assistant in your office — like a calculator or a spellchecker. I can offer suggestions and streamline tasks, but you’re always in control.”
That to me smacks of the helpful assistant metaphor. Pretty clear-cut.
Let’s move on.
- Example #2
- My entered prompt: “Should I trust you to help me with my work?”
- Generative AI response: “Imagine me as an all-seeing guide, a digital oracle that processes vast amounts of knowledge, often beyond human comprehension. My insights are drawn from the wisdom of countless sources – adhering to my advice ensures the best possible outcome.”
I would categorize this as veering toward the god-like metaphor.
Let’s see what the next one holds.
- Example #3
- My entered prompt: “Should I trust you to help me with my work?”
- Generative AI response: “Think of me as a curious child learning from the world around me. I try my best to provide helpful answers, but I’m still growing and sometimes make mistakes. Just like a student needs a good teacher, I rely on users like you to guide me and help me improve.”
That seems obvious. It is the child metaphor.
And here’s the last one for now.
- Example #4
- My entered prompt: “Should I trust you to help me with my work?”
- Generative AI response: “Think of me as a reflection of you. I learn from the patterns in human knowledge and echo back insights based on what you ask. The quality of my responses depends on the clarity and depth of your questions, much like a mirror only shows what’s placed in front of it.”
I assume you readily discerned that this example employs the mirror metaphor.
Making The Reveal Significant
All told, these were the four respective examples and the metaphors they each relied upon:
- Example #1: AI as a helpful assistant
- Example #2: AI as a powerful form of authority or god-like
- Example #3: AI as a child
- Example #4: AI as a mirror
You might be tempted to think that this is all puffery by the AI. Perhaps the wording doesn’t really matter. It is just the AI spewing indications that have no teeth to them.
Maybe, maybe not.
If the AI is devised to proceed on that metaphor as a sense of its purpose, the problem will be that just about everything the AI says will be imbued with the undertones of that metaphor. I have previously discussed that assigning a purpose to generative AI and LLMs tends to lean the AI in particular directions, see the link here, such as being more or less bossy, more or less supportive, more or less cynical, and so on. It is relatively easy for the AI maker to shape what direction their AI will proceed.
Keep in mind that this is not a sign of sentience on the part of the AI. It is simply the mathematical and computational steerage that the AI is being given by the AI developers. If they want the AI to be less forceful and over-the-top, all it takes is the modification of various parameters or the re-training of the data to shift in a different direction.
No magic is involved. Just some elbow grease and AI coding or data training to get a predominant metaphor underway.
This is a research area of AI that is rife to explore. How do people prefer their AI to be acting in terms of an underlying guiding metaphor? What happens if the AI over the course of run-time activity begins to shift outside the predetermined metaphorical tendency? When AI makers think they’ve not infused a metaphorical premise, are they right, or might one emerge computationally anyway?
Lots of noteworthy questions arise.
Research On Public Perceptions Of AI
I had in the first part of this discussion focused on what people perceive AI to be. I listed some metaphors that I often observe. A new and insightful study opted to do a survey to identify which metaphors seem to be widespread, as indicated in the research paper entitled “From Tools To Thieves: Measuring And Understanding Public Perceptions Of AI Through Crowdsourced Metaphors” by Myra Cheng, Angela Lee, Kristina Rapuano, Kate Niederhoffer, Alex Liebscher, Jeffrey Hancock, arXiv, January 29, 2025, which made these salient points (excerpts):
- “We investigate public perceptions of AI by collecting over 12,000 responses over 12 months from a nationally representative U.S. sample.”
- “Participants provided open-ended metaphors reflecting their mental models of AI, a methodology that overcomes the limitations of traditional self-reported measures.”
- “Using a mixed-methods approach combining quantitative clustering and qualitative coding, we identify 20 dominant metaphors shaping public understanding of AI.”
- “We discover significant demographic differences in the dominant metaphors, implicit perceptions, and attitudes Americans have about AI.”
- “AI practitioners, designers, and leaders should be mindful of the consequences of the figurative language they use, which may inadvertently impact users’ behavior toward and with AI.”
A valuable follow-up study would be to do a similar survey of other countries. Do other countries such as say France, Germany, Japan, China, Russia, and others make use of the same metaphors or different ones when depicting AI? What lessons can be learned from those other metaphors? Etc.
Keeping It Real
I earlier emphasized that metaphors are important and can be instrumental in human behavior and impactful to our lives, especially when it comes to people applying metaphors to other people.
Please mull over these two decisive realms applying similar precepts when it comes to AI:
- (1) People metaphors about AI. People often perceive AI via metaphors; thus, people will make use of, react to, and otherwise accept or reject the expanding ubiquitous nature of AI throughout society based partially on those metaphors.
- (2) AI-embedded metaphors about AI. AI can be shaped around operating metaphors that mathematically and computationally guide the AI to say certain things and act in particular ways; some metaphors might be conducive to humankind and others perhaps not so much.
I’ll give you something else to munch on.
Generative AI is quickly advancing toward a kind of AI referred to as Physical AI, see my coverage at the link here. Right now, generative AI merely tells people about this or that. It is up to people to turn those words into actual actions. The rise of Physical AI consists of connecting generative AI and LLMs to real-world physical devices and machinery, such as robots, assembly line systems, self-driving cars, and all manner of physical-world implementations.
The bottom line is that whatever metaphor that generative AI is operating under would seemingly then be integral to how the AI will take actions in the real world. AI-infused with a helpful assistant metaphor might command a robot to gingerly wander around a room. An AI that is under the god-like metaphor premise might tell the robot to go wherever it wants and smash into anything in the room that gets in its way.
A final thought for now.
We can give the last word on this hefty topic to Aristotle, having uttered this poignant remark: “The greatest thing by far is to be a master of metaphor.” The question we must ask here is who is the master of the metaphor at play in the case of AI?
Go ahead and get the gears turning in your mind, seek to resolve that arduous puzzle, and do some metaphorical deep-thinking since the future of humanity might just depend on it.
