Description
Every trainer learns the four quadrants: positive and negative reinforcement, positive and negative punishment. The grid sorts consequences by what the trainer adds or takes away. It says nothing about what the animal makes of them — and behavior is changed by the impact, not by the consequence. Consequential Impact puts that second half back, and the familiar grid starts to look different.
The book traces how the grid was built, from Thorndike’s puzzle boxes onward, and why it survived the objections raised against it. Skinner never drew the grid. Thorndike withdrew the punishment half of his own Law of Effect in 1932.
What replaces the grid is a way of judging a configuration by whether it holds itself. Angle an aircraft’s wings upward and a roll corrects itself. Reinforcement is dihedral in just that sense: the behavior is what the architecture is already producing. Punishment is anhedral: something must hold the suppression in place continuously. From this the book derives the four classic side effects of punishment — escape and avoidance, escalating counter-aggression, learned helplessness, generalized fears — as the architecture’s own stability solutions. It also marks out the narrow place where punishment is the right instrument, sets five requirements for doing it well, and is blunt that it is wrong for nearly every problem a trainer will meet.
A final chapter runs the same diagnosis at the scale of AI training.
Who this book is for
- Animal trainers and behavior consultants at any level
- Anyone who was taught the four quadrants and felt something was missing
- Instructors teaching learning theory
- Readers interested in how AI systems are trained
Book details
Paperback: 6 × 9 in, 220 pages · ISBN 979-8-9958081-9-0
EPUB: ISBN 979-8-9958081-8-3
LCCN 2026922785
Publisher: Libertywings Publishing, McNeal, Arizona
Article Comments