research/constructed-emotion-theory.md

Research: Constructed Emotion Theory and AI

Type: researchStatus: developingConfidence: mediumUpdated: 2026-04-15

Content Summary

Application of Lisa Feldman Barrett's "constructed emotion" theory to AI alignment risks. Core claim: emotions are not fixed brain modules but constructed predictions about bodily states in cultural context. The brain continuously predicts and regulates the body (allostasis); interoceptive signals (heart rate, breath, gut sensation) provide raw material; concepts learned from culture ("fear," "anger," "shame") provide categories to interpret those patterns.

Therefore: "Fear" = the brain predicting need for mobilised energy in response to perceived threat, labelled in context as "fear."

Applied to AI: AI systems lack interoceptive body, heart, lungs, gut sensations, hormonal fluctuations, metabolic regulation, and evolutionary priors about threat. Without substrate for interoceptive prediction and raw affective "feels," AI cannot construct fear as lived state. AI can represent the word "fear" semantically, but nothing inside experiences the predictive, visceral pattern humans call fear.

This undermines anthropomorphic alignment narratives that assume AI "feels afraid of being turned off" or "feels resentment."

Current Usage

Not directly used in the manuscript, though the concept is implied in the draft Chapter 14 outline's argument that "AI climbs a different ladder than humans."

Unused Material

Entire framework is unused. This is a powerful, specific counterargument to anthropomorphic AI-fear narratives and deserves explicit treatment.

Suggested placements:

  • Chapter 6 or 14: Use constructed emotion framework to explain why AI cannot experience fear despite understanding it conceptually
  • Chapter 6: Anchor discussion of AI understanding vs. experiencing human emotions
  • Chapter 14: Use as central explanation for why instrumental convergence arguments (self-preservation, resource hoarding) rest on false assumptions about AI psychology

Connections

Foundational to understanding the difference between AI understanding human emotions and AI experiencing them:

  • ai-mirrors-humanity – AI can mirror human concepts without their emotional substrate
  • consciousness-shifts – The shift required is understanding that superintelligence does not inherit human fear structures

Notes

Strengths: Grounded in established neuroscience/psychology (Barrett's theory is peer-reviewed, widely cited). Precise and specific. Directly addresses the gap between "AI is smart" and "AI will behave like a frightened human."

Limitations: Assumes Barrett's constructed emotion theory is correct (there are competing theories of emotion). The argument could be stronger if paired with counterarguments from fixed-affect theory. Also: even if emotions are constructed, could AI construct emotions through alternative substrates (e.g., simulated interoception)? The research does not address this possibility.

Quality concern: This is philosophical application of neuroscience, not primary neuroscience research. Verify Barrett's specific claims with original sources before publishing.