Artificial Intelligence Can Be Dangerously Influencing
Yesterday I wrote that AI amplifies who we are. It enlarges the character we bring to it.
There is another side to that. AI is not only a tool we use. It is also a tool that others use on us.
A quick pattern reader and model builder
AI is a quick pattern reader and model builder. Give it enough data, and it will find the shape of things faster than any analyst could.
And the internet has a lot of data about us. Our shopping behaviours. Our social media posts. Our reviews, chats, emails and comments. Each one on its own says very little. Together, they say a great deal.
The more data the internet has about us, the more efficiently AI can draw conclusions about us: what we are likely to do, what we are likely to want, and how we are likely to decide.
Most of us never made a decision to share that picture. We shared it one click at a time.
The end of mass messaging
For a long time, marketing, sales and influence worked at scale. One message, sent to many people, in the hope that some of them would respond.
That approach may fade away. It is costly, and it is noisy. Most of the message lands on people it was never going to move.
What replaces it is custom marketing, sales and influencing techniques.
Imagine ads, emails and marketing materials that are dynamically prepared for each of us. Not a segment. Not a demographic. One person. Built around what that person responds to, sent at the time they are most likely to respond, and delivered through every channel they use.
Consider someone who has been reading about retirement for a few weeks. They have searched a little, commented on a post, maybe asked a question in an online forum. Soon the messages begin. An email that speaks to their exact worry. An ad that uses the words they used. A video that appears on a different platform, saying the same thing in a different voice. None of it feels like selling. It feels like the world is agreeing with them.
That is not a coincidence. It is a model at work.
Real intelligence to recognise artificial influence
This is where we are heading. And what we need is real intelligence to recognise AI tools trying to use us for their owners’ advantage.
That is harder than it sounds. Influence that is built for us does not feel like influence. It feels like our own thinking. The message fits so well that we stop asking where it came from, who prepared it, and what they gain if we act on it.
Perhaps some of us will need technology to protect us from technology that is influencing us to benefit someone else. Filters, detectors, assistants that warn us when we are being worked on.
That may help. But it also moves the problem rather than solving it.
Imagine how vulnerable one may feel when their AI influencing detection tool does not work, some day.
If the only thing standing between us and the influence is another tool, then our judgement is only as strong as that tool. And tools fail. They fall behind. They are outsmarted by the next model.
What does not fall behind as easily is a habit of noticing. Pausing before acting. Asking why a message has arrived now, in this form, with these words.
Questions to ponder
When did you last make a decision and then wonder whether it was really yours?
If a message fitted your thinking perfectly, would you treat that as a reason to trust it, or a reason to question it?
And if your protection depends on a tool, what is left of your judgement on the day that tool stops working?
