An official website of the United States government
Published research highlights the presence of demographic bias in automated facial attribute classification algorithms, particularly impacting women and individuals with darker skin tones. Existing bias mitigation techniques typically require demographic annotations and often obtain a trade-off between fairness and accuracy, i.e., Pareto inefficiency. Facial attributes, whether common ones like gender or others such as "chubby" or "high cheekbones", exhibit high interclass similarity and intraclass variation across demographics leading to unequal accuracy. This requires the use of local and subtle cues using fine-grained analysis for differentiation. This paper proposes a novel approach to fair facial attribute classification by framing it as a fine-grained classification problem. Our approach effectively integrates both low-level local features (like edges and color) and high-level semantic features (like shapes and structures) through cross-layer mutual attention learning. Here, shallow to deep CNN layers function as experts, offering category predictions and attention regions. An exhaustive evaluation on facial attribute annotated datasets demonstrates that our FineFACE model improves accuracy by $$1.32\%$$ to $$1.74\%$$ and fairness by $$67\%$$ to $$83.6\%$$, over the SOTA bias mitigation techniques. Importantly, our approach obtains a Pareto-efficient balance between accuracy and fairness between demographic groups. In addition, our approach does not require demographic annotations and is applicable to diverse downstream classification tasks. To facilitate reproducibility, the code and dataset information is available at~\url{https://github.com/VCBSL-Fairness/FineFACE}.
more »
« less
This content will become publicly available on February 1, 2026
Addressing discretization-induced bias in demographic prediction
Abstract Racial and other demographic imputation is necessary for many applications, especially in auditing disparities and outreach targeting in political campaigns. The canonical approach is to construct continuous predictions—e.g. based on name and geography—and then to often discretize the predictions by selecting the most likely class (argmax), potentially with a minimum threshold (thresholding). We study how this practice produces discretization bias. For example, we show that argmax labeling, as used by a prominent commercial voter file vendor to impute race/ethnicity, results in a substantial under-count of Black voters, e.g. by 28.2% points in North Carolina. This bias can have substantial implications in downstream tasks that use such labels. We then introduce a joint optimization approach—and a tractable data-driven threshold heuristic—that can eliminate this bias, with negligible individual-level accuracy loss. Finally, we theoretically analyze discretization bias, show that calibrated continuous models are insufficient to eliminate it, and that an approach such as ours is necessary. Broadly, we warn researchers and practitioners against discretizing continuous demographic predictions without considering downstream consequences.
more »
« less
- Award ID(s):
- 2339427
- PAR ID:
- 10616974
- Editor(s):
- Levy, Morris
- Publisher / Repository:
- PNAS Nexus
- Date Published:
- Journal Name:
- PNAS Nexus
- Volume:
- 4
- Issue:
- 2
- ISSN:
- 2752-6542
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Structured demographic models are among the most common and useful tools in population biology. However, the introduction of integral projection models (IPMs) has caused a profound shift in the way many demographic models are conceptualized. Some researchers have argued that IPMs, by explicitly representing demographic processes as continuous functions of state variables such as size, are more statistically efficient, biologically realistic, and accurate than classic matrix projection models, calling into question the usefulness of the many studies based on matrix models. Here, we evaluate how IPMs and matrix models differ, as well as the extent to which these differences matter for estimation of key model outputs, including population growth rates, sensitivity patterns, and life spans. First, we detail the steps in constructing and using each type of model. Second, we present a review of published demographic models, concentrating on size‐based studies, which shows significant overlap in the way IPMs and matrix models are constructed and analyzed. Third, to assess the impact of various modeling decisions on demographic predictions, we ran a series of simulations based on size‐based demographic data sets for five biologically diverse species. We found little evidence that discrete vital rate estimation is less accurate than continuous functions across a wide range of sample sizes or size classes (equivalently bin numbers or mesh points). Most model outputs quickly converged with modest class numbers (≥10), regardless of most other modeling decisions. Another surprising result was that the most commonly used method to discretize growth rates for IPM analyses can introduce substantial error into model outputs. Finally, we show that empirical sample sizes generally matter more than modeling approach for the accuracy of demographic outputs. Based on these results, we provide specific recommendations to those constructing and evaluating structured population models. Both our literature review and simulations question the treatment of IPMs as a clearly distinct modeling approach or one that is inherently more accurate than classic matrix models. Importantly, this suggests that matrix models, representing the vast majority of past demographic analyses available for comparative and conservation work, continue to be useful and important sources of demographic information.more » « less
-
We propose a novel framework for the statistical analysis of genus-zero 4D surfaces, i.e., 3D surfaces that deform and evolve over time. This problem is particularly challenging due to the arbitrary parameterizations of these surfaces and their varying deformation speeds, necessitating effective spatiotemporal registration. Traditionally, 4D surfaces are discretized, in space and time, before computing their spatiotemporal registrations, geodesics, and statistics. However, this approach may result in suboptimal solutions and, as we demonstrate in this paper, is not necessary. In contrast, we treat 4D surfaces as continuous functions in both space and time. We introduce Dynamic Spherical Neural Surfaces (D-SNS), an efficient smooth and continuous spatiotemporal representation for genus-0 4D surfaces. We then demonstrate how to perform core 4D shape analysis tasks such as spatiotemporal registration, geodesics computation, and mean 4D shape estimation, directly on these continuous representations without upfront discretization and meshing. By integrating neural representations with classical Riemannian geometry and statistical shape analysis techniques, we provide the building blocks for enabling full functional shape analysis. We demonstrate the efficiency of the framework on 4D human and face datasets. The source code and additional results are available at https://4d-dsns.github.io/DSNS/.more » « less
-
The technique of modifying the geometry of a problem from Euclidean to Hessian metric has proved to be quite effective in optimization, and has been the subject of study for sampling. The Mirror Langevin Diffusion (MLD) is a sampling analogue of mirror flow in continuous time, and it has nice convergence properties under log-Sobolev or Poincare inequalities relative to the Hessian metric. In discrete time, a simple discretization of MLD is the Mirror Langevin Algorithm (MLA), which was shown to have a biased convergence guarantee with a non-vanishing bias term (does not go to zero as step size goes to zero). This raised the question of whether we need a better analysis or a better discretization to achieve a vanishing bias. Here we study the Mirror Langevin Algorithm and show it indeed has a vanishing bias. We apply mean-square analysis to show the mixing time bound for MLA under the modified self-concordance condition.more » « less
-
Recent studies show that Natural Language Processing (NLP) technologies propagate societal biases about demographic groups associated with attributes such as gender, race, and nationality. To create interventions and mitigate these biases and associated harms, it is vital to be able to detect and measure such biases. While existing works propose bias evaluation and mitigation methods for various tasks, there remains a need to cohesively understand the biases and the specific harms they measure, and how different measures compare with each other. To address this gap, this work presents a practical framework of harms and a series of questions that practitioners can answer to guide the development of bias measures. As a validation of our framework and documentation questions, we also present several case studies of how existing bias measures in NLP—both intrinsic measures of bias in representations and extrinsic measures of bias of downstream applications—can be aligned with different harms and how our proposed documentation questions facilitates more holistic understanding of what bias measures are measuring.more » « less
