Re: feedback

Eliezer S. Yudkowsky (
Mon, 22 Sep 1997 20:34:55 -0500

Eliezer S. Yudkowsky wrote, and explains to match detailed comments from others:
> Pleasure is positive feedback.
> Pain is negative feedback.

In other words, pleasure, particularly the kind of pleasure that results in
frantic button-pushing and self-starvation behavior in rats, is a positive
feedback. If you take a rat, give it a button, and wire the button to the
medial forebrain bundle, the rat will push a button and keep pushing it more
and more rapidly until it starves to death. If that isn't positive
*feedback*, not merely reinforcement, I don't know what is. Pleasure is also
positive reinforcement, since the rat will probably start pushing the button
again if you withdraw it and present it a few days later.

The case is similar for pain. Although negative feedback in the simple sense
doesn't generally become full-scale feedback - since the offending activity
stops immediately - you might also wire the rat's pain center so it activates
whenever some fluctuating value deviates from a certain level. If rats are
intelligent enough to comprehend their means of influencing that value in both
directions, the value is visible, and they make the connection and all that -
I think they're that intelligent - then they'll act to keep the value at a
particular level. This is *exactly* negative feedback. That's how a
thermostat works.

--       Eliezer S. Yudkowsky

Disclaimer:  Unless otherwise specified, I'm not telling you
everything I think I know.