{"paper":{"title":"Learning Preferences from Conjoint Data: A Structural Deep Learning Approach","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"Embedding a deep neural network inside a random utility logit model recovers flexible preference heterogeneity from conjoint data.","cross_cats":["econ.EM"],"primary_cat":"stat.ME","authors_text":"Avidit Acharya, Jens Hainmueller, Yiqing Xu","submitted_at":"2026-04-12T22:35:04Z","abstract_excerpt":"Conjoint experiments randomize multidimensional profiles, offering a powerful design for recovering structural preference parameters -- including marginal rates of substitution, willingness to pay, and the distribution of preferences across a population. Yet the dominant approach in political science has focused on nonparametric causal estimands that do not leverage this potential. We propose a structural approach that embeds a deep neural network within a random utility logit model, allowing preference parameters to vary as a fully flexible function of respondent characteristics. The neural n"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We propose a structural approach that embeds a deep neural network within a random utility logit model, allowing preference parameters to vary as a fully flexible function of respondent characteristics. [...] We apply our method to three prominent conjoint studies and find rich preference heterogeneity masked by reduced-form averages: a near-zero gender effect coexists with 83% preferring female candidates, opposition to undemocratic behavior is near-universal but varies sharply in intensity, and progressive tax preferences cut across every partisan subgroup.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that the random utility logit model with neural network-embedded parameters accurately represents the choice process, and that double/debiased machine learning successfully debiases the estimates despite the high flexibility of the neural network.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A structural deep learning approach for conjoint data reveals rich preference heterogeneity masked by reduced-form averages in three studies.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Embedding a deep neural network inside a random utility logit model recovers flexible preference heterogeneity from conjoint data.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ef365248b953a0a6ee9f458b458ce19c4cb1dc8880704e8e2857c394e7ed283d"},"source":{"id":"2604.10845","kind":"arxiv","version":2},"verdict":{"id":"1a7ab9ad-4288-49a2-91c0-bd81421b4940","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T15:05:00.691263Z","strongest_claim":"We propose a structural approach that embeds a deep neural network within a random utility logit model, allowing preference parameters to vary as a fully flexible function of respondent characteristics. [...] We apply our method to three prominent conjoint studies and find rich preference heterogeneity masked by reduced-form averages: a near-zero gender effect coexists with 83% preferring female candidates, opposition to undemocratic behavior is near-universal but varies sharply in intensity, and progressive tax preferences cut across every partisan subgroup.","one_line_summary":"A structural deep learning approach for conjoint data reveals rich preference heterogeneity masked by reduced-form averages in three studies.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that the random utility logit model with neural network-embedded parameters accurately represents the choice process, and that double/debiased machine learning successfully debiases the estimates despite the high flexibility of the neural network.","pith_extraction_headline":"Embedding a deep neural network inside a random utility logit model recovers flexible preference heterogeneity from conjoint data."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.10845/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"}