The list of things to worry about around artificial intelligence (AI) was already long: mass unemployment, robot superhackers, AI psychosis, atrophying skills, increased loneliness, workslop, grindslop, trendslop, just regular slop slop, even deadly AI-engineered viruses. I am sorry to inform you that science has just added yet another reason to fear AI to the list.
A new study suggests the more we interact with AI chatbots, the more robotic our own behaviour will become.
How do you change when you deal with AI?
To understand the idea behind this new paper, think back to the last time you interacted with an AI-powered customer service bot. Maybe you had an entirely satisfactory experience. But if dealing with this AI drove you to a nearly murderous rage, it likely wasn’t because the bot wasn’t helpful and cheery. They almost always are.
It was probably because there was some human quirk about your situation or language that the AI just couldn’t understand. What did you do in response?
If this new theoretical research by an international team is correct, you probably attempted to simplify and standardise your request and tone. You eliminated slang and metaphor. You minimised the uniqueness of your situation and presented your problem in the simplest possible terms. Your ask became explicit and detailed.
In short, you did your best to sound less like a quirky, complex human and more like a robot.
Your chatbots are training you, too
This might have gotten you whatever you needed out of that particular interaction. But the team of researchers behind the paper, recently published in AI & Society, is worried about the long-term effects of sanding off our rough edges to make dealing with bots less infuriating.
It’s true, they note, that over time robots are learning from us and becoming more human-like. But the influence goes the other way as well. We’re not just training AI, AI is training us, too. Each time we get what we want from an AI chatbot by acting like a robot, the behaviour is reinforced. Over time, we’re likely to mould ourselves to suit AI.
“The consumer acts, the robot responds, and with repeated exposure, in time the consumer internalises the exchange,” explained Selcen Ozturkcan, a member of the research team and professor at Sweden’s Linnaeus University. “For people, the innate instinct to mirror can lead to people returning the robot’s communicative behaviours, making them more robot-like.”
As AI becomes more human-like and humans become more robot-like, we are slowly converging on a mutual style of communication that the researchers dub “robotoid humanness.” (Which, in the interest of preserving humans’ precious weirdness, I should point out is also a good name for an experimental garage band.)
Beware “robotoid humanness”
It’s important to stress that this paper is theoretical. The scientists were writing up what they think is likely to happen. They didn’t conduct experiments to prove that it is already happening. But as ScienceAlert points out, linking to earlier research, they also “aren’t the first to suggest that AI interactions risk changing our thinking and how we express ourselves as flesh-and-blood people.”
We know that people respond to incentives. If everyone suspects you of using AI to generate your writing when you use em dashes or certain sentence constructions, writers will use these “tells” less. If management is thrilled every time someone uses AI, everyone will start using AI as much as possible, whether it makes sense or not.
So it seems logical that, if you can only get your customer service issue sorted out if you talk like a robot, people will start talking like robots more. Is this as terrifying a threat as AI-engineered killer viruses? No, but it’s certainly unsettling to think AI may be subtly training us to act less human.
I’m putting it on my list of things to keep an eye on as the AI revolution continues to unfold. You, too, might want to watch out for AI nudging us toward increasingly robot-like behaviour. – Inc./Tribune News Service
