Auditing Subgroup Fairness

What it does

When your agent uses it

• The user wants per-subgroup recall and leakage
• Wants to check whether de-identification under-protects a group
• Wants to surface a documentation gap where subgroup data is missing
• Needs equalized-odds-style disparity numbers for a clinical model

Example prompts

• “fairness”
• “subgroup”
• “bias audit”
• “/auditing-subgroup-fairness”

Requirements

• Python 3

Workflow steps

6 steps, taken from the first numbered list in SKILL.md.

1. 1Tag the gold corpus by group. Add a synthetic group to each PHI span's
2. 2Run fairness_report on the model + suite.
3. 3Read leakage first, recall second. For de-id, a group with higher leakage
4. 4Compute the disparity (leakage_disparity) and locate worst_group.
5. 5Document the gap. If race/ethnicity surrogates are absent, report that the
6. 6Feed it forward. Put per-group numbers and the gap into the model card and

What it can do on your machine

Read from SKILL.md and the folder at commit 34d7b8c. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  No scripts in the folder and no shell commands in SKILL.md (its code samples are python).

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  Links to these hosts (documentation or services it may open):

  • datadiversity.org
  • arxiv.org
  • doi.org

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names no API keys, tokens, secrets or passwords.

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Auditing Subgroup Fairness loads about 1.5k tokens when it runs. Until then it costs about 161 tokens; SKILL.md has 540 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~161

When it runs · the whole SKILL.md, loaded when a task matches

~1.5k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from maziyarpanahi/openmed at commit 34d7b8c, republished under its Apache-2.0 licence (© maziyarpanahi). 540 words, ~1,471 tokens.

name

auditing-subgroup-fairness

description

Audit an OpenMed NER or de-identification model for performance disparities across demographic subgroups (sex, age band, race/ethnicity when available) using openmed.eval.fairness_report. Use when the user wants per-subgroup recall and leakage, wants to check whether de-identification under-protects a group, wants to surface a documentation gap where subgroup data is missing, or needs equalized-odds-style disparity numbers for a clinical model. Trigger on "fairness", "subgroup", "bias audit", "disparity", "equalized odds", "under-protected group", "per-group recall", or "STANDING Together" for an OpenMed model.

license

Apache-2.0

metadata.project

OpenMed

metadata.category

evaluation-quality

metadata.pairs

adjacent

metadata.version

1.0

from openmed.eval import fairness_report

# Gold fixtures must tag spans with a surrogate group, e.g.
#   {"start": 4, "end": 12, "label": "PERSON", "metadata": {"group": "female"}}
fair = fairness_report(
    "OpenMed/Privacy-PII-Detection",
    "golden",                 # named suite, or pass a list of fixtures
    device="cpu",
)

print("leakage disparity:", fair.leakage_disparity)
print("worst group      :", fair.worst_group, fair.worst_group_leakage)
for group, m in sorted(fair.per_group.items()):
    print(f"  {group:14s} recall={m.recall:.3f}  leakage={m.leakage_rate:.4f}")

# Under-protection alarm: any group leaking more than the rest.
LEAKAGE_GAP_LIMIT = 0.0       # leakage-first: ideally zero leakage everywhere
assert fair.leakage_disparity <= LEAKAGE_GAP_LIMIT or fair.worst_group_leakage == 0

© maziyarpanahi, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in skills/auditing-subgroup-fairness of maziyarpanahi/openmed.

Open the folder on GitHubat commit 34d7b8c