THE LEAP OF FAITH
# 11 in an irregular series
You can find your way across a familiar room in the dark. Until someone moves a chair.
The hypegeist - that roar of things which get picked up and talked about incessantly - quite understandably missed some mice who know the feeling. But their behaviours are rather instructive.
Researchers let the creatures explore a landscape of tiles and towers in the dark. Within two hours, they were leaping between towers from memory, without first checking the landing.
They settled on a small set of routes they kept returning to. Those familiar paths made what came next more predictable. The researchers suggest that could mean less to remember.
And when a tile changed height, the mice reacted more strongly if they approached along a well-worn route. On familiar ground, a small change stood out.
But so what? Well, work is a well-known room with familiar ground where somebody keeps moving the chairs. Customers change. Rules change. Experienced people move on. Models get replaced, and AI agents start behaving differently.
Take a customer you know well. They’re usually happy with a quick chat. Now they want every promise in writing. Your agent drafts a confirmation. But you pick up the phone.
The request is perfectly polite. It’s also unlike them.
A note left for you, others and your agents that says ‘called customer’ loses the reason you stepped in. The next intelligence to assess it sees what you did. But misses the why. And that information might be the most important context of all.
Keep that reason and its outcome, and humans and AI can understand why you called and if your 'spidey-sense' uncovered a problem. Or whether it went off unnecessarily - and your counterpart was just being asked to keep better records by procurement.
After a few of these examples, your agent might be able to make better recommendations in future. And you can then test whether those recommendations catch the most likely reality. Or not.
That’s how an AI-infused operation keeps up as the chairs move. It retains the judgement behind an action, tests the implication empirically and helps to improve your decision-making. Continuously. Each new learning is a ratchet in quality.
Best of all? What you noticed can become part of how the next person or agent works.
Their next leap isn’t so dark or from such a high ledge when it has your experience behind it. Which means we all have rational reasons to be less timid that the mice.
Meet the mice: Ma et al.
bioRxiv preprint:
nt:
