Fairness Evaluation of Neural Networks Through Computational Profile Likelihood
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
ABSTRACT Despite high predictive performance, machine learning models can be unfair towards specific demographic subgroups characterized by sensitive attributes such as gender or race. This paper presents a novel approach using Computational Profile Likelihood (CPL) to assess potential bias in neural network decisions with respect to sensitive attributes. CPL estimates the conditional probability of a network's internal neuron excitation levels during predictions. To assess the impact of sensitive attributes on predictions, the CPL distribution of individuals sharing a particular value of a sensitive attribute and a specific outcome (e.g., “women” and “high income”) is compared to a subgroup sharing another value of the sensitive attribute but with the same outcome (e.g., “men” and “high income”). The resulting disparities between distributions can be used to quantify the bias with respect to the sensitive attribute and the outcome class. We also assess the efficacy of bias reduction techniques through their influence on the resulting disparities. Experimental results on three widely used datasets indicate that the CPL of the trained models can be used to characterize significant differences between multiple protected groups, highlighting that these models display quantifiable biases. Furthermore, after applying bias mitigation methods, the gaps in CPL distributions are reduced, indicating a more similar internal representation for profiles of different protected groups.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it