Re-Specifying the Utility Function: Resolving "Code Drift" in Behavioral Models via the AIM Framework
1. The Problem: "Code Drift" in Utility Aggregation
Current economic and behavioral models largely rely on a singular metric of "Preference" (or Utility, U) to predict agent behavior. These models assume that U acts as a coherent, stable value.
However, this singular definition aggregates three neurobiologically distinct drivers that obey contradictory mathematical laws:
- Homeostatic Drivers (Appetites): Subject to strict saturation (diminishing marginal utility).
- Autotelic Drivers (Intrinsic): Subject to flow states (constant or increasing marginal utility).
- Reflexive Drivers (Mimetic): Subject to social recursion (dependent on the utility functions of other agents).
The Hypothesis: Treating these three distinct vectors as a single variable (U) constitutes a "Definitional Error"(or Code Drift). This error causes "gear grinding" in models—manifesting as anomalies like the Easterlin Paradox or Veblen Effects—because the model attempts to apply a single logic to competing neural systems.
2. The Proposed Solution: The AIM Vector
We propose abandoning the singular "Preference" metric in favor of a tri-partite vector based on the AIM Motivation Framework.
We posit that an agent's decision (D) is not a function of maximized Utility (U), but the resultant vector of three independent systems:
Where:
A (Appetites): A variable governed by Negative Feedback Loops. It seeks 0 (homeostasis). It is biologically bounded and cyclical.
I (Intrinsic): A variable governed by Internal Reward Loops. It is independent of external observation. It is the source of "noise" in standard models but represents "signal" in the AIM framework.
M (Mimetic): A variable governed by Positive Feedback Loops (Mirror Neurons). It is unbounded and socially recursive (Agent X wants object O because Agent Y wants object O).
3. Theoretical Application (The "Contract")
Just as a legal contract fails if terms are vaguely defined, behavioral models fail when the input variable "Desire" is ambiguous.
By applying AIM Definitions as a "strict contractual standard" for research, we can test the following predictions:
Prediction A (Saturation)
Incentives targeting A (e.g., basic income) will show sharp diminishing returns once biological thresholds are met.
Prediction M (Contagion)
Incentives targeting M (e.g., status goods, rankings) will show no saturation point but high volatility (bubbles), as they rely on external social validation.
Prediction I (Crowding Out)
Introducing M-based incentives (rewards/rankings) to an I-based activity will degrade the stability of the behavior by shifting the neural driver from a renewable internal loop to a volatile external loop.
Note for SDT researchers: This prediction directly extends and formalizes Self-Determination Theory's observations about extrinsic rewards undermining intrinsic motivation. The AIM Framework provides the neural mechanism (switching from VTA dopamine loops to mirror neuron systems) and a quantifiable prediction (behavior becomes more volatile and context-dependent).
4. Call to Collaboration
The AIM Framework offers the definitions (the taxonomy), but we require the metrics (the econometrics). We invite economists, game theorists, and neuroscientists to "plug" these definitions into existing models to test if disaggregating U into {A, I, M} yields higher predictive accuracy than current standard models.
Why This Matters for Your Research
- ✓It respects your mathematics: Uses concepts like "diminishing marginal utility" and "recursion" that are standard in behavioral economics
- ✓It validates definitional work: Frames the contribution as fixing construct validity so you can fix the mathematics
- ✓It isolates mimetic behavior: By separating M as a distinct variable, you can model social contagion without breaking your core utility framework
Ready to Collaborate?
We're seeking research partners to empirically test these predictions. Whether you work in behavioral economics, game theory, neuroscience, or experimental psychology, we want to hear from you.