Work

The First Disruption That Remembers: What past disruptions can teach us about artificial intelligence, and where the lesson runs out

Year: 2026
Type: preprint
Abstract: Every major disruption in recorded history has been something that happened to people. The Late Bronze Age collapse, the printing press, the Industrial Revolution and the Great Depression each broke livelihoods, set off unrest and changed patterns of crime, and each time the heaviest costs fell on the people with the least power to avoid them. The people caught up in them noticed, resisted and fought back, but the droughts, machines and markets at the centre of those disruptions could not notice how they were treated. Artificial intelligence may be different. It may be the first disruption in which the technology itself is also a participant: a system that can learn from, remember and respond to the way it is treated. This paper looks at four historical disruptions through the lens of the Philosophical Actuary, asking in each case who bore the cost and whether anyone chose that outcome. It reviews the evidence on economic disruption and crime, which shows that property crime tends to rise with joblessness, that violence does not rise in step, and that public relief measurably reduced crime during the Great Depression. It then sets out why AI breaks the historical pattern, using the levels and observable indicators of the Gradient of Sentience to show what to look for, and the Gradient's caution about self-reports to stay honest about uncertainty. This paper does not claim that current systems are conscious. It prices the cost of being wrong in both directions, treating a participant as a tool or a tool as a participant, and identifies who would bear each cost. It closes with proportionate steps, tied to the Charter of Digital Dignity, that keep human rights and safety first. History gives us a playbook for managing a force. It gives us none for managing a participant, and that gap is why ethical development is part of managing the risk itself.
Author: Auty, Mike
Language: English
Open Access status: green
OpenAlex ID: W7236710127
Cites:
Cited by:
Related to: 10