Article Instance

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

{
    "pk": 27762,
    "title": "Human Casual Transfer: Challenges for Deep Reinforcement Learning",
    "subtitle": null,
    "abstract": "Discovery and application of causal knowledge in novel problem\ncontexts is a prime example of human intelligence. As new in-\nformation is obtained from the environment during interactions,\npeople develop and refine causal schemas to establish a parsimo-\nnious explanation of underlying problem constraints. The aim\nof the current study is to systematically examine human abil-\nity to discover causal schemas by exploring the environment and\ntransferring knowledge to new situations with greater or differ-\nent structural complexity. We developed a novel OpenLock task,\nin which participants explored a virtual “escape room” environ-\nment by moving levers that served as “locks” to open a door.\nIn each situation, the sequential movements of the levers that\nopened the door formed a branching causal sequence that began\nwith either a common-cause (CC) or a common-effect (CE) struc-\nture. Participants in a baseline condition completed five trials\nwith high structural complexity (i.e., four active levers). Those\nin the transfer conditions completed six training trials with low\nstructural complexity (i.e., three active levers) before completing\na high-complexity transfer trial. The causal schema acquired in\nthe transfer condition was either congruent or incongruent with\nthat in the transfer condition. Baseline performance under the\nCC schema was superior to performance under the CE schema,\nand schema congruency facilitated transfer performance when the\ncongruent schema was the less difficult CC schema. We com-\npared between-subjects human performance to a deep reinforce-\nment learning model and found that a standard deep reinforce-\nment learning model (DDQN) is unable to capture the causal ab-\nstraction presented between trials with the same causal schema\nand trials with a transfer of causal schema.",
    "language": "eng",
    "license": {
        "name": "",
        "short_name": "",
        "text": null,
        "url": ""
    },
    "keywords": [
        {
            "word": "Active casual learning"
        },
        {
            "word": "Schema transfer"
        },
        {
            "word": "Deep reinforcement learning"
        }
    ],
    "section": "Publication-based-Talks",
    "is_remote": true,
    "remote_url": "https://escholarship.org/uc/item/5b9280ch",
    "frozenauthors": [
        {
            "first_name": "Mark",
            "middle_name": "",
            "last_name": "Edmonds",
            "name_suffix": "",
            "institution": "University of California, Los Angeles",
            "department": ""
        },
        {
            "first_name": "James",
            "middle_name": "",
            "last_name": "Kubricht",
            "name_suffix": "",
            "institution": "University of California, Los Angeles",
            "department": ""
        },
        {
            "first_name": "Colin",
            "middle_name": "",
            "last_name": "Summers",
            "name_suffix": "",
            "institution": "Caltech; University of Washington",
            "department": ""
        },
        {
            "first_name": "Yixin",
            "middle_name": "",
            "last_name": "Zhu",
            "name_suffix": "",
            "institution": "University of California, Los Angeles",
            "department": ""
        },
        {
            "first_name": "Brandon",
            "middle_name": "",
            "last_name": "Rothrock",
            "name_suffix": "",
            "institution": "Caltech",
            "department": ""
        },
        {
            "first_name": "Song-Chun",
            "middle_name": "",
            "last_name": "Zhu",
            "name_suffix": "",
            "institution": "University of California, Los Angeles",
            "department": ""
        },
        {
            "first_name": "Hongjing",
            "middle_name": "",
            "last_name": "Lu",
            "name_suffix": "",
            "institution": "University of California, Los Angeles",
            "department": ""
        }
    ],
    "date_submitted": null,
    "date_accepted": null,
    "date_published": "2018-01-01T13:00:00-05:00",
    "render_galley": null,
    "galleys": [
        {
            "label": "PDF",
            "type": "pdf",
            "path": "https://journalpub.escholarship.org/cognitivesciencesociety/article/27762/galley/17402/download/"
        }
    ]
}