pith:6CWT6K3Q
RLHF May Not Reflect Genuine Preferences
Measuring human preferences in RLHF is a social science problem because responses often reflect non-attitudes and constructed preferences rather than genuine opinions.
arxiv:2604.03238 v2 · 2026-01-31 · cs.HC
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{6CWT6K3QKRZA7DN5XXOXYGHCBN}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
We argue that measuring human preferences in RLHF is a social science problem. We present a taxonomy distinguishing genuine preferences from non-attitudes, constructed preferences, and measurement artifacts, along with diagnostic approaches for detecting each.
That the documented phenomena of non-attitudes and constructed preferences from general behavioral science literature transfer directly and pervasively to the specific task of RLHF annotation without needing domain-specific validation studies.
RLHF preference measurement is a social science validity problem because annotators routinely produce non-attitudes, constructed responses, and artifacts rather than stable values.
Formal links
Receipt and verification
| First computed | 2026-06-02T01:03:46.565674Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
f0ad3f2b7054720f8dbdbddd7c18e20b444390f70bce50612bec69aefa1999ae
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/6CWT6K3QKRZA7DN5XXOXYGHCBN \
| jq -c '.canonical_record' \
| python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: f0ad3f2b7054720f8dbdbddd7c18e20b444390f70bce50612bec69aefa1999ae
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "d2aea64851723e357251f0578e5c3cf72d321564641d89d433bab1e8065e739c",
"cross_cats_sorted": [],
"license": "http://creativecommons.org/licenses/by/4.0/",
"primary_cat": "cs.HC",
"submitted_at": "2026-01-31T21:51:17Z",
"title_canon_sha256": "e4a26c9ac93ebfd2ce3f2d4f0503e2bca3054eae9a56831c16cac577c9f94a3a"
},
"schema_version": "1.0",
"source": {
"id": "2604.03238",
"kind": "arxiv",
"version": 2
}
}