Effect of Demographic Bias on Skin Lesion Classification DOI: 10.59275/j.melba.2026-4156 Volume: 2026

\firstname Ralf \surname Raumanns \aff 1,3,4 \orcid 0000-0002-7055-8136, \name Gerard Schouten \aff 2 \orcid 0000-0001-7042-2143, \name Veronika Cheplygina \aff 4 \orcid 0000-0003-0176-9324, \name Josien P.W. Pluim \aff 3 \orcid 0000-0001-7327-9178 · 2026

Read Effect of Demographic Bias on Skin Lesion Classification DOI: 10.59275/j.melba.2026-4156 Volume: 2026 by \firstname Ralf \surname Raumanns \aff 1,3,4 \orcid 0000-0002-7055-8136, \name Gerard Schouten \aff 2 \orcid 0000-0001-7042-2143, \name Veronika Cheplygina \aff 4 \orcid 0000-0003-0176-9324, \name Josien P.W. Pluim \aff 3 \orcid 0000-0001-7327-9178 (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-03214.txt.

Description

Raumanns and co-authors measure how demographic bias in training data skews skin lesion classification. The team uses ResNet-based models. They build datasets with set sex and age mixes by linear programming. Three training schemes are compared: single-task, a reinforcing multi-task scheme, and an adversarial scheme.

Sex-matched training data gave the best scores for each sex. Adding male patients to training helped the male group even when women were the majority. The two added schemes shrank or closed bias gaps in balanced and female-majority sets. In male-majority sets they helped little, and models still favored males. Age told a unlike story. Scores fell with age in all schemes, and the young did best no matter the data mix. The authors conclude sex bias flows mostly from data imbalance, while age bias favors the young in all cases. Each kind thus needs its own fix.

Catalog record

CREATOR\firstname Ralf \surname Raumanns \aff 1,3,4 \orcid 0000-0002-7055-8136, \name Gerard Schouten \aff 2 \orcid 0000-0001-7042-2143, \name Veronika Cheplygina \aff 4 \orcid 0000-0003-0176-9324, \name Josien P.W. Pluim \aff 3 \orcid 0000-0001-7327-9178
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE81 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.03214

Browse the full library · About the Tower of Babel