The economics of motion perception and invariants of visual sensitivity

Sergei Gepshtein, Ivan Tyukin, Michael KubovyJournal of Vision 2007 · DOI

TL;DR

This paper applies neoclassical economic theory to model how the human visual system optimizes motion perception under resource constraints. By treating visual sensitivity as an economic allocation problem, the authors propose an "equilibrium theory" that explains how the visual system balances measurement and stimulus uncertainties to achieve optimal speed estimation, aligning with observed human visual sensitivity patterns.

What Problem Does This Paper Address?

The human visual system, like any complex system, operates with limited resources (neurons, processing power). This raises a fundamental question: how does it efficiently allocate these resources to perceive motion, especially when faced with varying stimulus conditions and the inherent trade-offs between precision and accuracy? Specifically, the paper addresses how visual sensitivity is distributed across different speeds and spatiotemporal parameters, and why different perceptual phenomena like apparent motion occur. The authors aim to provide a normative framework, explaining how the system should operate optimally, and then compare this to how it does operate.

Methods

The core of the paper's methodology is the application of neoclassical economic theory, particularly the concept of equilibrium, to visual perception. This involves:

  • Uncertainty Principle: Drawing parallels to the Heisenberg uncertainty principle in physics, the authors posit a measurement uncertainty principle for visual stimuli, stating that the product of uncertainties in spatial/temporal resolution and frequency cannot be reduced below a certain constant.
  • Economic Equilibrium: The authors frame the visual system's resource allocation as an economic problem where the "consumer" (visual system) aims to minimize "dissatisfaction" (measurement errors) given a limited "budget" (neural resources). Equilibrium is achieved when satisfaction from one component cannot be improved without reducing satisfaction from another, and overall satisfaction is maximized.
  • Spatiotemporal Uncertainty Function: They develop a mathematical model for the combined spatial and temporal uncertainty in measuring motion parameters.
  • Optimization Frameworks: Three frameworks are explored:
  • Local Optimization: Assumes the system can optimize for each speed independently, leading to a "spendthrift" but precise approach.
  • Total Integration: Assumes extreme frugality, integrating across all speeds and weighting by their prevalence in the environment, leading to a "frugal" but less precise approach.
  • Weighted Optimization: A realistic compromise that linearly combines the first two, accounting for both measurement and stimulus uncertainties, and the prevalence of speeds in the natural environment.
  • Equivalence Sets: To understand suboptimal conditions, they derive "equivalence sets" representing conditions with the same degree of uncertainty imbalance.

Datasets

The paper primarily uses existing empirical data for comparison and validation, rather than introducing new datasets for training or evaluation. Key datasets/sources include:

  • Kelly (1979) spatiotemporal contrast sensitivity function: This data is used to represent human visual sensitivity in space-time plots and is compared against the theoretical predictions of optimal and suboptimal conditions.
  • Empirical data on apparent motion: Studies by Burt and Sperling (1981) and Gepshtein and Kubovy (2007) on apparent motion regimes (tradeoff vs. coupling) are used to validate the theory's predictions about different perceptual experiences at different speeds.
  • Data on motion adaptation: Studies by Krekelberg et al. (2006) are referenced to discuss the predicted effects of adaptation on visual sensitivity.

Key Insights and Ideas

  • Vision as an Economic System: The most striking idea is viewing the visual system's resource allocation through the lens of neoclassical economics. This provides a powerful, unified framework for understanding diverse perceptual phenomena.
  • The Equilibrium Principle: The core insight is that optimal visual performance, especially under resource constraints, is achieved when there's a balance between different sources of uncertainty (measurement vs. stimulus). This equilibrium dictates where sensitivity should be highest.
  • Invariants of Visual Sensitivity: The theory predicts that the shape of optimal and suboptimal conditions (represented by hyperbolic curves) is invariant across different parameters, suggesting fundamental principles governing visual processing.
  • Explaining Apparent Motion Regimes: The theory elegantly explains why different regimes of apparent motion (coupling and tradeoff) are observed at different speeds, linking them to the underlying balance of uncertainties.
  • Normative Framework for Sensitivity: The paper offers a normative model that predicts how visual sensitivity should be distributed, providing a benchmark against which biological vision can be compared.

Key Results

  • Prediction of Optimal Sensitivity Distribution: The theory predicts that visual sensitivity should be highest at points where spatial and temporal uncertainties are in equilibrium, forming a hyperbolic curve in the space-time distance plot. This theoretical curve closely matches the observed maximal sensitivity set in human vision.
  • Equivalence Sets Match Empirical Data: The derived "equivalence sets" (representing equally suboptimal conditions) closely resemble the "bent-loaf-of-bread" shape of human isosensitivity contours, suggesting that human vision operates by balancing uncertainties.
  • Explanation of Apparent Motion Regimes: The theory predicts that space-time coupling (observed at high speeds) and tradeoff (observed at low speeds) are not arbitrary but are consequences of the visual system's strategy to balance uncertainties, transitioning smoothly with speed.
  • Predictions for Motion Adaptation: The theory predicts that motion adaptation should lead to global changes in visual sensitivity, not just at the adapting speed, and that sensitivity can improve or deteriorate depending on the specific conditions.

Limitations and Future Work

  • Qualitative Predictions: The authors acknowledge that the theory's predictions are primarily qualitative. While the shapes of sensitivity contours are predicted, the exact parameters and precise locations are not definitively determined by the theory alone.
  • Simplifying Assumptions: The model relies on assumptions like the additivity of uncertainties and a specific form of the uncertainty function. Relaxing these could lead to more nuanced predictions.
  • "Importance Function" for Speeds: The model incorporates the prevalence of speeds in the environment but lacks a direct measure of "speed importance" for an organism. Future work could explore how such importance functions might influence predictions.
  • Testable Predictions for Adaptation: The paper suggests that future research should conduct comprehensive studies on motion adaptation to test the theory's predictions about global changes in sensitivity.
  • Comparison with Other Models: The paper briefly compares its approach to "weak fusion" and Bayesian inference models, highlighting differences in how uncertainty and prior statistics are handled. Further exploration of these comparisons could be valuable.

Summary updated

Abstract

Neural systems face the challenge of optimizing their performance with limited resources, just as economic systems do. Here, we use tools of neoclassical economic theory to explore how a frugal visual system should use a limited number of neurons to optimize perception of motion. The theory prescribes that vision should allocate its resources to different conditions of stimulation according to the degree of balance between measurement uncertainties and stimulus uncertainties. We find that human vision approximately follows the optimal prescription. The equilibrium theory explains why human visual sensitivity is distributed the way it is and why qualitatively different regimes of apparent motion are observed at different speeds. The theory offers a new normative framework for understanding the mechanisms of visual sensitivity at the threshold of visibility and above the threshold and predicts large-scale changes in visual sensitivity in response to changes in the statistics of stimulation and system goals.
BibTeX
@article{gepshtein2007economics,
  author = {Sergei Gepshtein and Ivan Tyukin and Michael Kubovy},
  title = {{The economics of motion perception and invariants of visual sensitivity}},
  journal = {Journal of Vision},
  year = {2007},
  doi = {10.1167/7.8.8},
}

Loading related content...

Figure