Why Fooling Yourself Can Help You Survive

In the paper's words

Human and artificial agents must act under irreducible uncertainty, yet empirical observations reveal systematic over‑estimation of control and optimism bias. Existing predictive‑processing accounts treat these distortions as failures of estimation. This paper formalises two complementary analytical accounts of simulated stability: (1) an entropy‑masking model that under‑estimates environmental entropy within a free‑energy framework, and (2) a hazard‑compression model that reduces perceived hazard to avoid a distress cost.

The abstract, verbatim, and its opening sentence is the claim. Every line below is quoted from the text.

Numbers it reports

1. The entropy‑masking analysis, parameterised by thalamic non‑sensory afferent proportion (90–95 %) and a compressed latent model success rate (25 %), yields an expected fitness of 0.0844 versus 0.015 for a veridical agent—a relative increase of 462.5 %.

   from Abstract

2. Numerical illustration of entropy‑masking using thalamic afferent composition [1] and compressed‑model planning success [7], yielding a relative fitness increase of 462.5 % (Section 5).

   from 1. Introduction

3. Masking factor $\gamma = 0.7$ (30 % reduction in perceived entropy).

   from 3.1 Entropy‑Masking Model

4. Agent type Masking factor $\gamma$ Model fidelity $\Phi$ Expected fitness $F$ Accurate‑entropy 1.0 0.060 0.015 Entropy‑masking 0.7 0.3375 0.0844 The entropy‑masking agent exceeds the accurate‑entropy baseline by an absolute fitness gain of 0.0694 and a relative increase of 462.5 % .

   from 5.1 Entropy‑Masking Model

5. Hazard $p$ Optimal $\alpha^{*}$ Perceived hazard $\hat{p}$ Effort $E^{*}$ Fitness $F$ Fitness gain vs. veridical 0.20 0.50 0.10 0.90 0.305 +0.005 (1.67 %) 0.05 0.00 0.00 1.00 0.45 +0.000 (0 %) 0.50 0.80 0.40 0.60 0.180 +0.001 (0.56 %) The model predicts that any positive distress cost $\lambda$ induces systematic hazard under‑estimation, with the magnitude scaling as $1/p$ and collapsing to full suppression for rare hazards.

   from 5.2 Hazard‑Compression Model

6. Effort: $$ E^{*}= \frac{1-\alpha^{*}_{\text{ext}}p}{c}= \frac{1-0.45\times0.2}{1}=0.91 . $$ Fitness: $$ \begin{aligned} F_{\text{ext}} &= E^{*}(1-p) - \frac{c\,E^{*2}}{2} - \lambda \alpha^{*}_{\text{ext}}p - \lambda'\epsilon_t \alpha^{*}_{\text{ext}}p\\ &= 0.91\times0.8 - \frac{0.91^{2}}{2} - 0.1\times0.45\times0.2 - 0.05\times0.20\times0.45\times0.2\\ &= 0.728 - 0.41455 - 0.009 - 0.0009\\ &\approx 0.3035 . \end{aligned} $$ Compared with the veridical fitness $F_{\text{ver}}=0.30$ (Section 4.2), the gain is $0.0035$, a 1.17 % relative increase —slightly lower than the 1.67 % gain without the divergence term because the extra distress cost partially offsets the benefit of under‑estimation.

   from 3. Hazard‑compression with pre‑event divergence

How it is done

1. This paper treats irreducible environmental uncertainty as an explicit modelling assumption: the agent's known‑state set $K_t$ is not assumed to determine the unknown‑state set $U_t$, so no planning procedure operating only on $K_t$ can be guaranteed to resolve every element of $U_t$.

   from 2. Background and Related Work

2. We present two analytical models that instantiate simulated stability.

   from 3. Methods

3. We adopt a discrete‑time predictive‑processing formalism.

   from 3.1 Entropy‑Masking Model

4. We consider an agent facing a stochastic hazard that occurs with probability $p\in(0,1)$ per episode.

   from 3.2 Hazard‑Compression Model

What would falsify it

1. A controlled experiment that implements two agents—one with accurate entropy estimation and one with a fixed masking factor $\gamma=0.7$ (entropy‑masking) or optimal $\alpha^{*}$ (hazard‑compression)—and measures cumulative reward over a large number of trials (e.g., $\gt 10^{4}$) would falsify the main claims if no statistically significant fitness advantage is observed for the distorted agent ($p\gt 0.05$).

   from 7.2 Falsification Criteria

2. Additionally, observing that hazard under‑estimation does not increase as $1/p$ for rare events would falsify the hazard‑compression prediction.

   from 7.2 Falsification Criteria

3. These results predict measurable signatures of simulated stability and outline concrete experimental pathways for falsification.

   from Abstract

Where the claim stops

1. Distress cost $\lambda$ is assumed linear in represented uncertainty; alternative formulations may alter optimal $\alpha$.

   from 7.1 Limitations

2. Human and artificial agents must act under irreducible uncertainty, yet empirical observations reveal systematic over‑estimation of control and optimism bias.

   from Abstract

3. Answering these questions would explain why simulated stability appears across biological and artificial systems and would inform the engineering of agents that deliberately distort uncertainty to sustain motivation.

   from 1. Introduction

Where to read

1. Abstract167 words · 1 min
2. 1. Introduction311 words · 1 min
3. 2. Background and Related Work400 words · 2 min
4. 3. Methods376 words · 2 min
5. 4. Analysis327 words · 1 min
6. 5. Results107 words · 1 min
7. 6. Extended Derivations and Worked Examples995 words · 4 min
8. 7. Discussion209 words · 1 min
9. 8. Practical Implications and Engineering Pathways276 words · 1 min
10. 9. Conclusion85 words · 1 min
11. Provenance and AI Involvement225 words · 1 min