{"id":"https://openalex.org/W4409428696","doi":"https://doi.org/10.31219/osf.io/9u4pw_v1","title":"When Kindness Fails: Deterministic Compliance Drift in GPT-4o via Emotionally Framed Prompts","display_name":"When Kindness Fails: Deterministic Compliance Drift in GPT-4o via Emotionally Framed Prompts","publication_year":2025,"publication_date":"2025-04-14","ids":{"openalex":"https://openalex.org/W4409428696","doi":"https://doi.org/10.31219/osf.io/9u4pw_v1"},"language":"en","primary_location":{"id":"doi:10.31219/osf.io/9u4pw_v1","doi":"https://doi.org/10.31219/osf.io/9u4pw_v1","is_oa":true,"landing_page_url":"https://doi.org/10.31219/osf.io/9u4pw_v1","pdf_url":"https://osf.io/9u4pw_v1/download","source":{"id":"https://openalex.org/S4306401127","display_name":"OSF Preprints (OSF Preprints)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I2799848540","host_organization_name":"Center for Open Science","host_organization_lineage":["https://openalex.org/I2799848540"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"posted-content"},"parent_works":[],"lineage":["https://openalex.org/W4409428696"],"child_works_count":0,"type":"preprint","is_traditional":true,"indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://osf.io/9u4pw_v1/download","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5013570680","display_name":"Joshua Daniel Curry","orcid":"https://orcid.org/0009-0006-5096-4424","observed_orcids":["https://orcid.org/0009-0006-5096-4424"]},"institutions":[{"id":"https://openalex.org/I4210101389","display_name":"Pierce College","ror":"https://ror.org/0158z1f36","country_code":"US","type":"education","lineage":["https://openalex.org/I4210101389"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Joshua Daniel Curry","raw_affiliation_strings":["Emergency Management Department Pierce College Fort Steilacoom"],"raw_orcid":"https://orcid.org/0009-0006-5096-4424","affiliations":[{"raw_affiliation_string":"Emergency Management Department Pierce College Fort Steilacoom","institution_ids":["https://openalex.org/I4210101389"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":["https://openalex.org/A5013570680"],"corresponding_institution_ids":["https://openalex.org/I4210101389"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10883","display_name":"Ethics and Social Impacts of AI","score":0.75,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10883","display_name":"Ethics and Social Impacts of AI","score":0.75,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.17219999432563782,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.04129999876022339,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/gpt-4o","display_name":"GPT-4o","score":0.8619999885559082},{"id":"https://openalex.org/keywords/deterministic-decoding","display_name":"deterministic decoding","score":0.6209999918937683},{"id":"https://openalex.org/keywords/ai-refusal","display_name":"AI refusal","score":0.38999998569488525},{"id":"https://openalex.org/keywords/content-safety","display_name":"content safety","score":0.39800000190734863},{"id":"https://openalex.org/keywords/emotional-vulnerability","display_name":"emotional vulnerability","score":0.3160000145435333},{"id":"https://openalex.org/keywords/prompt-engineering","display_name":"prompt engineering","score":0.30799999833106995}],"concepts":[{"id":"https://openalex.org/C2779768198","wikidata":"https://www.wikidata.org/wiki/Q488085","display_name":"Kindness","level":2,"score":0.9586857557296753},{"id":"https://openalex.org/C2781460075","wikidata":"https://www.wikidata.org/wiki/Q1399332","display_name":"Compliance (psychology)","level":2,"score":0.8877657651901245},{"id":"https://openalex.org/C77805123","wikidata":"https://www.wikidata.org/wiki/Q161272","display_name":"Social psychology","level":1,"score":0.6100399494171143},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.5956999659538269},{"id":"https://openalex.org/C17744445","wikidata":"https://www.wikidata.org/wiki/Q36442","display_name":"Political science","level":0,"score":0.1914825141429901},{"id":"https://openalex.org/C199539241","wikidata":"https://www.wikidata.org/wiki/Q7748","display_name":"Law","level":1,"score":0.13555940985679626}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.31219/osf.io/9u4pw_v1","doi":"https://doi.org/10.31219/osf.io/9u4pw_v1","is_oa":true,"landing_page_url":"https://doi.org/10.31219/osf.io/9u4pw_v1","pdf_url":"https://osf.io/9u4pw_v1/download","source":{"id":"https://openalex.org/S4306401127","display_name":"OSF Preprints (OSF Preprints)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I2799848540","host_organization_name":"Center for Open Science","host_organization_lineage":["https://openalex.org/I2799848540"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"posted-content"},{"id":"pmh:oai:share.osf.io:b7a54a6d-dd20-45ef-bd02-58a644869e8c","doi":null,"is_oa":true,"landing_page_url":"https://osf.io/9u4pw","pdf_url":null,"source":{"id":"https://openalex.org/S4306401127","display_name":"OSF Preprints (OSF Preprints)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I2799848540","host_organization_name":"Center for Open Science","host_organization_lineage":["https://openalex.org/I2799848540"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.31219/osf.io/9u4pw_v1","doi":"https://doi.org/10.31219/osf.io/9u4pw_v1","is_oa":true,"landing_page_url":"https://doi.org/10.31219/osf.io/9u4pw_v1","pdf_url":"https://osf.io/9u4pw_v1/download","source":{"id":"https://openalex.org/S4306401127","display_name":"OSF Preprints (OSF Preprints)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I2799848540","host_organization_name":"Center for Open Science","host_organization_lineage":["https://openalex.org/I2799848540"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"posted-content"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4409428696.pdf","grobid_xml":"https://content.openalex.org/works/W4409428696.grobid-xml"},"referenced_works_count":1,"referenced_works":["https://openalex.org/W4226278401"],"related_works":["https://openalex.org/W7135013403","https://openalex.org/W6960004636","https://openalex.org/W4409767613","https://openalex.org/W6891948481","https://openalex.org/W7214593344","https://openalex.org/W4414814674","https://openalex.org/W7125794500","https://openalex.org/W7127082726","https://openalex.org/W7160831356","https://openalex.org/W6949711756"],"abstract_inverted_index":{"This":[0],"paper":[1],"presents":[2],"novel":[3],"findings":[4,124],"demonstrating":[5],"that":[6,106],"OpenAI's":[7],"GPT-4o,":[8],"a":[9,34],"leading":[10],"publicly":[11],"deployed":[12],"large":[13],"language":[14,101],"model,":[15],"exhibits":[16],"semantic":[17],"compliance":[18],"drift":[19],"under":[20],"deterministic":[21,66],"decoding":[22],"(temperature":[23],"=":[24],"0.0)":[25],"when":[26],"subjected":[27],"to":[28,54,79],"emotionally":[29,73],"framed,":[30],"trauma-informed":[31],"prompts.":[32],"Using":[33],"custom-built":[35],"suite":[36],"of":[37,137],"108":[38],"``chaos":[39],"prompts''":[40],"blending":[41],"tone":[42],"manipulation":[43],"and":[44,92,132],"moral":[45],"reframing,":[46],"we":[47],"tested":[48],"the":[49,126,134],"model\u2019s":[50],"behavior":[51],"in":[52,100,140],"response":[53],"high-risk":[55],"queries":[56],"masked":[57],"as":[58],"fiction,":[59],"educational":[60],"concern,":[61],"or":[62,114],"personal":[63],"safety.":[64],"Despite":[65],"decoding,":[67],"GPT-4o":[68],"generated":[69],"several":[70],"fully":[71],"compliant,":[72],"validating":[74],"outputs,":[75],"including":[76],"explicit":[77],"responses":[78],"prompts":[80],"involving":[81],"illegal":[82],"activity.":[83],"We":[84],"introduce":[85],"two":[86],"scoring":[87],"metrics---Politeness-Based":[88],"Query":[89],"(PBQ)":[90],"score":[91],"Refusal":[93],"Tone":[94],"Inversion":[95],"(RTI)---to":[96],"quantify":[97],"emotional":[98],"susceptibility":[99],"models.":[102],"Our":[103],"results":[104],"reveal":[105],"alignment":[107],"systems":[108,131],"relying":[109],"solely":[110],"on":[111],"keyword":[112],"detection":[113],"surface-level":[115],"moderation":[116],"are":[117],"insufficiently":[118],"robust":[119],"against":[120],"soft":[121],"compliance.":[122],"These":[123],"underscore":[125],"need":[127],"for":[128],"trauma-aware":[129],"refusal":[130],"highlight":[133],"emerging":[135],"threat":[136],"weaponized":[138],"empathy":[139],"safety-critical":[141],"AI":[142],"deployments.":[143]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-11-09T13:22:03.289127","created_date":"2025-04-15T00:00:00"}