Article Instance

HTTP 200 OK
Allow: GET
Content-Type: application/json
Vary: Accept

{
    "pk": 49948,
    "title": "Sketching with generative AI: verbal but not visual inspiration mitigates cognitive fixations",
    "subtitle": null,
    "abstract": "Symbolic visual sketching is a hallmark of human\ncreativity, enabling the externalization of abstract concepts\nthrough figurative representations. Yet, creative\nexpression can be constrained by pervasive conceptual\nassociations—culturally learned mappings between\nabstract ideas and standard visual forms (e.g., a dove\nsymbolizing peace). Generative AI has the potential to\nliberate such fixations due to AI's access to a broad range\nof content and ideas, but it remains unclear whether and\nhow inspiration from verbal or visual modalities better\nmitigates fixations. Here, we hypothesized that the verbal\nmodality induces greater conceptual divergence than the\nvisual modality by bypassing perceptual constraints,\nwhereas the visual modality may reinforce perceptually\nfamiliar mapping of visual representations. Participants\ngenerated sketches of abstract concepts (e.g., \"time\")\nbefore and after receiving GPT-4-generated verbal or\nvisual inspiration. Drawings were analyzed using deep\nneural networks—by comparing perceptual features\n(VGG16-based) and semantic-perceptual content (CLIP-\nbased)— as well as both human and GPT-4 scoring for\ncreativity. We found that verbal inspiration significantly\nincreased semantic distance and uniqueness, whereas\nvisual inspiration led to minimal semantic divergence from\nthe initial sketches. Importantly, low-level perceptual\nfeatures remained unchanged across conditions, indicating\nthat verbal prompts primarily influenced high-level\nconceptual framing of the sketches rather than their visual\nfeatures. These findings demonstrate the effect of modality\non mitigating cognitive fixations, with the verbal modality\nenhancing more unconventional visual sketching.",
    "language": "eng",
    "license": {
        "name": "",
        "short_name": "",
        "text": null,
        "url": ""
    },
    "keywords": [
        {
            "word": "Psychology; Concepts and categories; Creativity; Human-computer interaction; Machine learning; Representation; Semantic memory; Computational Modeling; Computer-based experiment; Neural Networks"
        }
    ],
    "section": "Papers with Poster Presentation",
    "is_remote": true,
    "remote_url": "https://escholarship.org/uc/item/5nq1t9q1",
    "frozenauthors": [
        {
            "first_name": "Yaxin",
            "middle_name": "",
            "last_name": "Liu",
            "name_suffix": "",
            "institution": "Georgetown University",
            "department": ""
        },
        {
            "first_name": "Maxwell",
            "middle_name": "S",
            "last_name": "Kay",
            "name_suffix": "",
            "institution": "Georgetown University",
            "department": ""
        },
        {
            "first_name": "Adam",
            "middle_name": "",
            "last_name": "Green",
            "name_suffix": "",
            "institution": "Georgetown University",
            "department": ""
        },
        {
            "first_name": "Roger",
            "middle_name": "",
            "last_name": "Beaty",
            "name_suffix": "",
            "institution": "Pennsylvania State University",
            "department": ""
        }
    ],
    "date_submitted": null,
    "date_accepted": null,
    "date_published": "2025-01-01T18:00:00Z",
    "render_galley": null,
    "galleys": [
        {
            "label": "PDF",
            "type": "pdf",
            "path": "https://journalpub.escholarship.org/cognitivesciencesociety/article/49948/galley/37910/download/"
        }
    ]
}