{"id":32393,"date":"2026-01-21T17:01:53","date_gmt":"2026-01-21T17:01:53","guid":{"rendered":"https:\/\/lamarr-institute.org\/publication\/survey-to-behavior-downstream-alignment-of-human-values-in-llms-via-survey-questions\/"},"modified":"2026-06-08T13:20:09","modified_gmt":"2026-06-08T13:20:09","slug":"survey-to-behavior-downstream-alignment-of-human-values-in-llms-via-survey-questions","status":"publish","type":"publication","link":"https:\/\/lamarr-institute.org\/de\/publication\/survey-to-behavior-downstream-alignment-of-human-values-in-llms-via-survey-questions\/","title":{"rendered":"Survey-to-Behavior: Downstream Alignment of Human Values in LLMs via Survey Questions"},"content":{"rendered":"<p>Large language models implicitly encode preferences over human values, yet steering them often requires large training data. In this work, we investigate a simple approach: Can we reliably modify a model&#8217;s value system in downstream behavior by training it to answer value survey questions accordingly? We first construct value profiles of several open-source LLMs by asking them to rate a series of value-related descriptions spanning 20 distinct human values, which we use as a baseline for subsequent experiments. We then investigate whether the value system of a model can be governed by fine-tuning on the value surveys. We evaluate the effect of finetuning on the model&#8217;s behavior in two ways; first, we assess how answers change on in-domain, held-out survey questions. Second, we evaluate whether the model&#8217;s behavior changes in out-of-domain settings (situational scenarios). To this end, we construct a contextualized moral judgment dataset based on Reddit posts and evaluate changes in the model&#8217;s behavior in text-based adventure games. We demonstrate that our simple approach can not only change the model&#8217;s answers to in-domain survey questions, but also produces substantial shifts (value alignment) in implicit downstream task behavior.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Large language models implicitly encode preferences over human values, yet steering them often requires large training data. In this work, we investigate a simple approach: Can we reliably modify a model&#8217;s value system in downstream behavior by training it to answer value survey questions accordingly? We first construct value profiles of several open-source LLMs by asking them to rate a series of value-related descriptions spanning 20 distinct human values, which [&hellip;]<\/p>\n","protected":false},"author":12,"featured_media":0,"template":"","meta":{"_acf_changed":false,"footnotes":""},"publication-type":[30],"class_list":["post-32393","publication","type-publication","status-publish","hentry","publication-type-article"],"acf":[],"publishpress_future_workflow_manual_trigger":{"enabledWorkflows":[]},"_links":{"self":[{"href":"https:\/\/lamarr-institute.org\/de\/wp-json\/wp\/v2\/publication\/32393","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/lamarr-institute.org\/de\/wp-json\/wp\/v2\/publication"}],"about":[{"href":"https:\/\/lamarr-institute.org\/de\/wp-json\/wp\/v2\/types\/publication"}],"author":[{"embeddable":true,"href":"https:\/\/lamarr-institute.org\/de\/wp-json\/wp\/v2\/users\/12"}],"version-history":[{"count":0,"href":"https:\/\/lamarr-institute.org\/de\/wp-json\/wp\/v2\/publication\/32393\/revisions"}],"wp:attachment":[{"href":"https:\/\/lamarr-institute.org\/de\/wp-json\/wp\/v2\/media?parent=32393"}],"wp:term":[{"taxonomy":"publication-type","embeddable":true,"href":"https:\/\/lamarr-institute.org\/de\/wp-json\/wp\/v2\/publication-type?post=32393"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}