{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,10,16]],"date-time":"2025-10-16T00:47:16Z","timestamp":1760575636641,"version":"build-2065373602"},"reference-count":0,"publisher":"Association for the Advancement of Artificial Intelligence (AAAI)","issue":"3","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["AIES"],"abstract":"<jats:p>Understanding what constitutes safety in AI-generated\ncontent is complex. While developers often rely on\npredefined taxonomies, real-world safety judgments also\ninvolve personal, social, and cultural perceptions of harm.\nThis paper examines how annotators evaluate the safety of\nAI-generated images, focusing on the qualitative reasoning\nbehind their judgments.  Analyzing 5,372 open-ended\ncomments, we find that annotators consistently invoke\nmoral, emotional, and contextual reasoning that extends\nbeyond structured safety categories. Many reflect on\npotential harm to others more than to themselves, grounding\ntheir judgments in lived experience, collective risk, and\nsociocultural awareness. Beyond individual perceptions, we\nalso find that the structure of the task itself--including\nannotation guidelines--shapes how annotators interpret and\nexpress harm. Guidelines influence not only which images\nare flagged, but also the moral language used in\njustifications. Annotators frequently cite factors such as\nimage quality, visual distortion, and mismatches between\nprompt and output as contributing to perceived\nharm-dimensions often overlooked in standard evaluation\nframeworks. Our findings reveal that existing safety\npipelines miss critical forms of reasoning that annotators\nbring to the task. We argue for evaluation designs that\nscaffold moral reflection, differentiate types of harm, and\nmake space for subjective, context-sensitive\ninterpretations of AI-generated content.<\/jats:p>","DOI":"10.1609\/aies.v8i3.36742","type":"journal-article","created":{"date-parts":[[2025,10,15]],"date-time":"2025-10-15T13:27:35Z","timestamp":1760534855000},"page":"2612-2624","source":"Crossref","is-referenced-by-count":0,"title":["\"Just a Strange Pic\": Rethinking 'Safety' in GenAI Image Safety Annotation Tasks from Diverse Annotators' Perspectives"],"prefix":"10.1609","volume":"8","author":[{"given":"Ding","family":"Wang","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Mark","family":"D\u00edaz","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Charvi","family":"Rastogi","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Aida","family":"Davani","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Vinodkumar","family":"Prabhakaran","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Pushkar","family":"Mishra","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Roma","family":"Patel","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Alicia","family":"Parrish","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zoe","family":"Ashwood","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Michela","family":"Paganini","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Tian Huey","family":"Teh","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Verena","family":"Rieser","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Lora","family":"Aroyo","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"9382","published-online":{"date-parts":[[2025,10,15]]},"container-title":["Proceedings of the AAAI\/ACM Conference on AI, Ethics, and Society"],"original-title":[],"link":[{"URL":"https:\/\/ojs.aaai.org\/index.php\/AIES\/article\/download\/36742\/38880","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/ojs.aaai.org\/index.php\/AIES\/article\/download\/36742\/38880","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,15]],"date-time":"2025-10-15T13:27:37Z","timestamp":1760534857000},"score":1,"resource":{"primary":{"URL":"https:\/\/ojs.aaai.org\/index.php\/AIES\/article\/view\/36742"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,15]]},"references-count":0,"journal-issue":{"issue":"3","published-online":{"date-parts":[[2025,10,15]]}},"URL":"https:\/\/doi.org\/10.1609\/aies.v8i3.36742","relation":{},"ISSN":["3065-8365"],"issn-type":[{"value":"3065-8365","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,10,15]]}}}