{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:6LBQIF6EHC33YOP7S6AYVAGV45","short_pith_number":"pith:6LBQIF6E","schema_version":"1.0","canonical_sha256":"f2c30417c438b7bc39ff97818a80d5e7604f770968ef89255a9635623029346e","source":{"kind":"arxiv","id":"1902.09725","version":3},"attestation_state":"computed","paper":{"title":"Conservative Agency via Attainable Utility Preservation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Alexander Matt Turner, Dylan Hadfield-Menell, Prasad Tadepalli","submitted_at":"2019-02-26T04:42:54Z","abstract_excerpt":"Reward functions are easy to misspecify; although designers can make corrections after observing mistakes, an agent pursuing a misspecified reward function can irreversibly change the state of its environment. If that change precludes optimization of the correctly specified reward function, then correction is futile. For example, a robotic factory assistant could break expensive equipment due to a reward misspecification; even if the designers immediately correct the reward function, the damage is done. To mitigate this risk, we introduce an approach that balances optimization of the primary r"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"1902.09725","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2019-02-26T04:42:54Z","cross_cats_sorted":[],"title_canon_sha256":"72464154fceb6b275d84dc36861752cb40b59718ff41121a8534e5ce881f6d18","abstract_canon_sha256":"4403048896c8b93fb72b87daa75c814e3b5e9f02e9e21e323bcd5a71ba6abc11"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:08:58.927261Z","signature_b64":"wwA0083W/DUmBcE1KQquyTvm3h4QM+q02FOksPbYImIEN7MzVYfwSa6aMu5lp5KYO+zCvYk3TRKn83jeGbYzCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f2c30417c438b7bc39ff97818a80d5e7604f770968ef89255a9635623029346e","last_reissued_at":"2026-07-05T01:08:58.926833Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:08:58.926833Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Conservative Agency via Attainable Utility Preservation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Alexander Matt Turner, Dylan Hadfield-Menell, Prasad Tadepalli","submitted_at":"2019-02-26T04:42:54Z","abstract_excerpt":"Reward functions are easy to misspecify; although designers can make corrections after observing mistakes, an agent pursuing a misspecified reward function can irreversibly change the state of its environment. If that change precludes optimization of the correctly specified reward function, then correction is futile. For example, a robotic factory assistant could break expensive equipment due to a reward misspecification; even if the designers immediately correct the reward function, the damage is done. To mitigate this risk, we introduce an approach that balances optimization of the primary r"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1902.09725","kind":"arxiv","version":3},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/1902.09725/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"1902.09725","created_at":"2026-07-05T01:08:58.926888+00:00"},{"alias_kind":"arxiv_version","alias_value":"1902.09725v3","created_at":"2026-07-05T01:08:58.926888+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1902.09725","created_at":"2026-07-05T01:08:58.926888+00:00"},{"alias_kind":"pith_short_12","alias_value":"6LBQIF6EHC33","created_at":"2026-07-05T01:08:58.926888+00:00"},{"alias_kind":"pith_short_16","alias_value":"6LBQIF6EHC33YOP7","created_at":"2026-07-05T01:08:58.926888+00:00"},{"alias_kind":"pith_short_8","alias_value":"6LBQIF6E","created_at":"2026-07-05T01:08:58.926888+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"1906.10918","citing_title":"Towards Empathic Deep Q-Learning","ref_index":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6LBQIF6EHC33YOP7S6AYVAGV45","json":"https://pith.science/pith/6LBQIF6EHC33YOP7S6AYVAGV45.json","graph_json":"https://pith.science/api/pith-number/6LBQIF6EHC33YOP7S6AYVAGV45/graph.json","events_json":"https://pith.science/api/pith-number/6LBQIF6EHC33YOP7S6AYVAGV45/events.json","paper":"https://pith.science/paper/6LBQIF6E"},"agent_actions":{"view_html":"https://pith.science/pith/6LBQIF6EHC33YOP7S6AYVAGV45","download_json":"https://pith.science/pith/6LBQIF6EHC33YOP7S6AYVAGV45.json","view_paper":"https://pith.science/paper/6LBQIF6E","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1902.09725&json=true","fetch_graph":"https://pith.science/api/pith-number/6LBQIF6EHC33YOP7S6AYVAGV45/graph.json","fetch_events":"https://pith.science/api/pith-number/6LBQIF6EHC33YOP7S6AYVAGV45/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6LBQIF6EHC33YOP7S6AYVAGV45/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6LBQIF6EHC33YOP7S6AYVAGV45/action/storage_attestation","attest_author":"https://pith.science/pith/6LBQIF6EHC33YOP7S6AYVAGV45/action/author_attestation","sign_citation":"https://pith.science/pith/6LBQIF6EHC33YOP7S6AYVAGV45/action/citation_signature","submit_replication":"https://pith.science/pith/6LBQIF6EHC33YOP7S6AYVAGV45/action/replication_record"}},"created_at":"2026-07-05T01:08:58.926888+00:00","updated_at":"2026-07-05T01:08:58.926888+00:00"}