{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreiejnowxgy5fxubxmnimmlft2qmgx52soic4cowwcr3iijdeo5ux5i",
"uri": "at://did:plc:zcsop6sjapkb5vlahd24pddv/app.bsky.feed.post/3mi3rxrpg2lv2"
},
"coverImage": {
"$type": "blob",
"ref": {
"$link": "bafkreigxel25g5xiia7t6mpsezlb2nvfquzt4lisylsp2iyy4lyanaq4re"
},
"mimeType": "image/png",
"size": 676001
},
"path": "/technology/features/ai-sycophancy-study-science-journal-march-2026/",
"publishedAt": "2026-03-28T04:00:00.000Z",
"site": "https://www.rappler.com",
"tags": [
"Editors' Pick",
"Explainers",
"Newsbreak",
"Tech Features",
"Technology",
"artificial intelligence",
"Generative AI",
"tech companies"
],
"textContent": "A new Science study finds that artificial intelligence models affirm questionable behavior or statements far more than a regular human would, reinforced by people actually preferring sycophantic responses",
"title": "AI as people pleaser: What this study tells us about its sycophantic behavior"
}