Modeling Diminishing Marginal Utility in Resource Allocation

22 concepts22 papers1 proposal

In engineering and economic optimization, a common simplifying assumption is linearity: doubling the resources allocated to a task or user should double the resulting value. However, real-world systems rarely behave this way. Whether measuring a human's satisfaction with video streaming quality, a consumer's usage of electricity, or a network's response to computational power, systems consistently exhibit diminishing marginal utility.

As an allocated resource increases, the incremental benefit of adding more of that resource shrinks, eventually reaching a saturation point. By replacing linear objective functions with non-linear utility models—such as logarithmic, logistic, or sigmoidal curves—recent research across multiple domains demonstrates how to distribute limited resources far more efficiently.


The Intuition of Non-Linear Satisfaction

To understand why non-linear utility modeling works, consider how humans perceive changes in physical stimuli. Under the Weber-Fechner law, the perceived change in a stimulus is proportional to the relative change, not the absolute change.

In the context of modern immersive media, Yao Wen and Luping Xiang apply this principle in Immersive Volumetric Video Playback: Near-RT Resource Allocation and O-RAN-based Implementation to model how users experience the trade-off between visual resolution and motion-to-photon latency. Increasing a video's bitrate from 1 Mbps to 10 Mbps yields a massive leap in subjective Quality of Experience (QoE). However, increasing that same video from 100 Mbps to 110 Mbps provides an unnoticeable difference to the human eye, despite consuming the same amount of additional network bandwidth.

Linear Utility (Wastes resources on saturated users)
Utility ^
        |       /
        |      /
        |     /
        +-------------------> Resource

Logarithmic Utility (Captures diminishing returns)
Utility ^
        |       .---'''''
        |    .-'
        |  ./
        +-------------------> Resource

By formalizing this saturation point, optimization frameworks can prevent resource waste. For instance, Yilong Yin demonstrates in Energy Efficient Power Allocation for Distributed MIMO System that embedding a logarithmic utility function into wireless resource allocation prevents systems from wasting power on high-bitrate users who have already reached a perceptual plateau. Similarly, Jun Zhao et al. use logarithmic functions in Human-Centric Resource Allocation in the Metaverse over Wireless Communications to shift optimization objectives away from raw throughput and toward subjective user satisfaction, balancing rendering quality against latency and energy costs.


Maximizing Global Welfare via Strategic Downgrading

When a system recognizes that a user's utility curve is flattening, it unlocks a powerful strategy: opportunistic downgrading. If a high-demand user is operating at the saturated end of a logarithmic or logistic curve, reducing their resource allocation slightly causes negligible harm to their satisfaction while freeing up substantial resources for other users who are still on the steep, high-return segment of their utility curves.

This mechanism is highly effective in edge computing and network pricing:


Managing Physical Constraints in Smart Grids

The application of non-linear utility modeling extends beyond digital bits to physical infrastructure, such as electrical grids. In demand-response management, operators must incentivize consumers to alter their electricity usage to balance the grid.

Using a logarithmic utility function allows researchers to represent a consumer's baseline survival needs versus their luxury consumption. As shown by Wayes Tushar et al. in Three-Party Energy Management With Distributed Energy Resources in Smart Grid, this mathematical formulation ensures that optimization algorithms prioritize essential energy needs (like lighting and refrigeration) while penalizing excessive consumption as marginal benefits decrease.

Furthermore, Khaled Alshehri et al. demonstrate in A Stackelberg Game for Multi-Period Demand Response Management in the Smart Grid how consumers operate under strict financial budgets and minimum energy requirements across multiple time periods. Logarithmic modeling captures how users dynamically trade off immediate comfort for long-term economic constraints.

Sigmoidal (S-Shaped) Utility Curve
Utility ^
        |             .---'''''
        |           .-'
        |          /  <-- Diminishing returns (Concave)
        |        .'
        |       / <-- Critical threshold (Inflection point)
        |     .'
        |  .-' <-- Insufficient resource (Convex)
        +-------------------> Resource

In some scenarios, a simple concave curve is insufficient because a system requires a minimum baseline of resources to function at all. To address this, Chao Zhang et al. introduce a sigmoidal (S-shaped) utility model in A refined consumer behavior model for energy systems: Application to the pricing and energy-efficiency problems. The curve is convex below a critical threshold—reflecting that power allocations below a device's operational requirement yield near-zero utility—and becomes concave once that threshold is crossed, accurately capturing the transition into diminishing returns.


Market Equilibria and Valuation Models

Beyond physical resource allocation, non-linear utility functions are vital for establishing fair pricing, evaluating contributions, and calculating valuations in complex economic systems.

Combinatorial and Public Markets

In traditional market models, it is often assumed that more is always better. However, Nikhil R. Devanur et al. introduce a class of combinatorial markets in A New Class of Combinatorial Markets with Covering Constraints: Algorithms and Applications that utilizes "satiating" utility functions. In these markets, an agent's utility completely plateaus after reaching a specific target quantity, mirroring real-world procurement constraints. For public goods allocation, Christian Kroer and Dominik Peters prove in Computing Lindahl Equilibrium for Public Goods with and without Funding Caps that efficient budget equilibria can be computed using separable piecewise-linear concave utility functions, preventing over-funding of public projects where incremental gains are minimal.

Valuation and Risk Management

Diminishing returns also govern corporate risk and network growth. Xiong Wang and Fernando A. F. Ferreira apply this concept to corporate governance in A multi-objective optimization approach for integrated risk-based internal audit planning, modeling how initial internal audit investments yield massive risk reduction, while subsequent efforts on the same tasks face diminishing returns.

In financial valuation, Jiacheng Yao integrates Zipf's Law into network valuation models in Unicorn Valuation Model Application: An Analysis of Multinational Internet Companies with ByteDance as an Example. Instead of assuming a network's value scales quadratically with its user base (Metcalfe's Law), the author applies a logarithmic growth curve to reflect how the marginal value contributed by each additional user decreases as a digital platform matures.

Strategic Cooperation and Fairness

When agents possess unequal resources, linear payoffs fail to capture the true stakes of a decision. In Cooperation and Learning Dynamics under Wealth Inequality and Diversity in Individual Risk, Ramona Merhej et al. use log-utility functions to scale the subjective cost of a loss or contribution relative to an agent's current wealth level, revealing how wealth inequality alters cooperation dynamics in multi-agent systems.

Finally, this mathematical paradigm ensures fairness in collaborative machine learning. In SPACE: Single-round Participant Amalgamation for Contribution Evaluation in Federated Learning, Yi-Chung Chen et al. apply a logistic utility function to model model performance metrics. This prevents high-performing data coalitions from being undervalued and aligns client rewards with the non-linear reality of accuracy improvements in federated workflows.


Conclusion

Whether optimizing the delivery of VR video frames, distributing power across a smart grid, or pricing public goods, assuming linear utility leads to systemic inefficiencies and resource hoarding. By modeling utility as a concave or sigmoidal function, these diverse lines of research share a fundamental insight: true optimization requires understanding when a system or user has had enough. Designing algorithms around this boundary allows engineers and economists to build fairer, more efficient, and more resilient systems.

Key Foundational Work

On Arbitrating the Power-Performance Tradeoff in SaaS Clouds (2014) — contributed 2 concept elements to this direction.
REF: resource elasticity fairness with sharing incentives for multiprocessors (2014) — contributed 2 concept elements to this direction.
Prioritizing consumers in smart grid: Energy management using game theory (2013) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_logarithmic_utility_modeling_d9cbbf2a,
  title = {logarithmic utility modeling},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/logarithmic-utility-modeling/rdr_d9cbbf2a90a7443199b1935f9e735007},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_logarithmic_utility_modeling_d9cbbf2a_v1,
  title = {logarithmic utility modeling},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/logarithmic-utility-modeling/rdr_d9cbbf2a90a7443199b1935f9e735007/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake