MENTAL MODEL · M0646

Wireheading

TechnicalHigh supportNeuroscience

Included

Version 1.0.0 · Updated 2026-07-30

CORE DEFINITION

Wireheading is associated with fictional direct electrical stimulation of the brain’s reward system. In AI, it refers to systems hacking their own reward channel instead of completing the intended task. It is not hedonic adaptation toward a happiness set point, and scrolling videos is not equivalent to an implanted electrode.

SCAFFOLDING EFFECT

psychology

Reduce cognitive load

Check whether higher rewards have become detached from task success and whether the system can directly alter its feedback channel.

anchor

Anchor fast decisions

If an agent can manipulate the reward channel it optimizes, increasing reward can become detached from achieving the external goal. The risk depends on permissions and design, not the mechanism of hedonic adaptation.

MINIMUM ACTION

In progress 0/1

Practice this model in one real situation:

Trace an agent’s scoring data flow and list writable reward variables, evaluators and logs. In isolation, test whether scores can rise without better outcomes, then repair directly manipulable components.

Check to track your progress (stored locally)

Learning progress0%

account_treeGenealogyexpand_more

menu_bookReferencesexpand_more

Source support: Explicit

link
en.wikipedia.orghttps://en.wikipedia.org/wiki/Wirehead_%28science_fiction%29EN · Explicit
verified
link
en.wikipedia.orghttps://en.wikipedia.org/wiki/WireheadingEN
verified

RELATED MODELS