{"id":"https://openalex.org/W4412889799","doi":"https://doi.org/10.18653/v1/2025.acl-long.1399","title":"Know Your Mistakes: Towards Preventing Overreliance on Task-Oriented Conversational AI Through Accountability Modeling","display_name":"Know Your Mistakes: Towards Preventing Overreliance on Task-Oriented Conversational AI Through Accountability Modeling","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412889799","doi":"https://doi.org/10.18653/v1/2025.acl-long.1399"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.1399","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1399","pdf_url":"https://aclanthology.org/2025.acl-long.1399.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.acl-long.1399.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5007964652","display_name":"Suvodip Dey","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Suvodip Dey","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5108133777","display_name":"Y. Sun","orcid":"https://orcid.org/0009-0001-7929-9003"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yi-Jyun Sun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5087941479","display_name":"G\u00f6khan T\u00fcr","orcid":"https://orcid.org/0009-0008-7740-2557"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gokhan Tur","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5113822366","display_name":"Dilek Hakkani-T\u00fcr","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dilek Hakkani-T\u00fcr","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":1.7588,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.87803291,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":95,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"28830","last_page":"28843"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.9081000089645386,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.9081000089645386,"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/accountability","display_name":"Accountability","score":0.8292564749717712},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.7036363482475281},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6386764049530029},{"id":"https://openalex.org/keywords/knowledge-management","display_name":"Knowledge management","score":0.4174915850162506},{"id":"https://openalex.org/keywords/process-management","display_name":"Process management","score":0.41116905212402344},{"id":"https://openalex.org/keywords/human\u2013computer-interaction","display_name":"Human\u2013computer interaction","score":0.3955976665019989},{"id":"https://openalex.org/keywords/business","display_name":"Business","score":0.23500704765319824},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.17863672971725464},{"id":"https://openalex.org/keywords/systems-engineering","display_name":"Systems engineering","score":0.1778765320777893},{"id":"https://openalex.org/keywords/political-science","display_name":"Political science","score":0.14021790027618408}],"concepts":[{"id":"https://openalex.org/C2776007630","wikidata":"https://www.wikidata.org/wiki/Q2798912","display_name":"Accountability","level":2,"score":0.8292564749717712},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.7036363482475281},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6386764049530029},{"id":"https://openalex.org/C56739046","wikidata":"https://www.wikidata.org/wiki/Q192060","display_name":"Knowledge management","level":1,"score":0.4174915850162506},{"id":"https://openalex.org/C195094911","wikidata":"https://www.wikidata.org/wiki/Q14167904","display_name":"Process management","level":1,"score":0.41116905212402344},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.3955976665019989},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.23500704765319824},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.17863672971725464},{"id":"https://openalex.org/C201995342","wikidata":"https://www.wikidata.org/wiki/Q682496","display_name":"Systems engineering","level":1,"score":0.1778765320777893},{"id":"https://openalex.org/C17744445","wikidata":"https://www.wikidata.org/wiki/Q36442","display_name":"Political science","level":0,"score":0.14021790027618408},{"id":"https://openalex.org/C199539241","wikidata":"https://www.wikidata.org/wiki/Q7748","display_name":"Law","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.1399","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1399","pdf_url":"https://aclanthology.org/2025.acl-long.1399.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.1399","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.1399","pdf_url":"https://aclanthology.org/2025.acl-long.1399.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412889799.pdf","grobid_xml":"https://content.openalex.org/works/W4412889799.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2494121465","https://openalex.org/W1854749775","https://openalex.org/W4312971597","https://openalex.org/W2129556718","https://openalex.org/W1970818898","https://openalex.org/W2350665531","https://openalex.org/W2906160980","https://openalex.org/W2154949144","https://openalex.org/W3196817267","https://openalex.org/W1976600725"],"abstract_inverted_index":{"Recent":[0],"LLMs":[1,131],"have":[2],"enabled":[3],"significant":[4],"advancements":[5],"for":[6,70],"conversational":[7],"agents.However,":[8],"they":[9],"are":[10,22],"also":[11,156],"well":[12],"known":[13],"to":[14,30,75,112,194,218],"hallucinate,":[15],"producing":[16],"responses":[17],"that":[18,106,142,202],"seem":[19],"plausible":[20],"but":[21,155],"factually":[23],"incorrect.On":[24],"the":[25,114,118,123,143,158,185,190],"other":[26],"hand,":[27],"users":[28],"tend":[29],"over-rely":[31],"on":[32,132],"LLM-based":[33,71],"AI":[34,152],"agents,":[35],"accepting":[36],"AI's":[37],"suggestion":[38],"even":[39],"when":[40],"it":[41],"is":[42,97],"wrong.Adding":[43],"positive":[44],"friction,":[45],"such":[46],"as":[47,56,108],"explanations":[48],"or":[49],"getting":[50],"user":[51,77,206,220],"confirmations,":[52],"has":[53],"been":[54],"proposed":[55,144],"a":[57,109,211],"mitigation":[58],"in":[59,82,122,160,170],"AI-supported":[60],"decision-making":[61],"systems.In":[62],"this":[63],"paper,":[64],"we":[65,200],"propose":[66],"an":[67,98,102],"accountability":[68,95,104,180],"model":[69,85,96],"task-oriented":[72],"dialogue":[73,91,119],"agents":[74],"address":[76],"overreliance":[78],"via":[79],"friction":[80],"turns":[81],"cases":[83],"of":[84,117,151,175],"uncertainty":[86],"and":[87,137],"errors":[88,154,187],"associated":[89],"with":[90,101,128],"state":[92,120],"tracking":[93],"(DST).The":[94],"augmented":[99],"LLM":[100],"additional":[103],"head":[105],"functions":[107],"binary":[110],"classifier":[111],"predict":[113],"relevant":[115],"slots":[116],"mentioned":[121],"conversation.We":[124],"perform":[125],"our":[126],"experiments":[127],"multiple":[129],"backbone":[130],"two":[133],"established":[134],"benchmarks":[135],"(MultiWOZ":[136],"Snips).Our":[138],"empirical":[139],"findings":[140],"demonstrate":[141],"approach":[145],"not":[146],"only":[147],"enables":[148],"reliable":[149],"estimation":[150],"agent":[153],"guides":[157],"decoder":[159],"generating":[161],"more":[162],"accurate":[163],"actions.We":[164],"observe":[165],"around":[166],"3%":[167],"absolute":[168],"improvement":[169],"joint":[171],"goal":[172],"accuracy":[173],"(JGA)":[174],"DST":[176,198],"output":[177],"by":[178],"incorporating":[179],"heads":[181],"into":[182],"modern":[183],"LLMs.Self-correcting":[184],"detected":[186],"further":[188],"increases":[189],"JGA":[191],"from":[192],"67.13":[193],"70.51,":[195],"achieving":[196],"state-of-the-art":[197],"performance.Finally,":[199],"show":[201],"error":[203],"correction":[204],"through":[205],"confirmations":[207],"(friction":[208],"turn)":[209],"achieves":[210],"similar":[212],"performance":[213],"gain,":[214],"highlighting":[215],"its":[216],"potential":[217],"reduce":[219],"overreliance.":[221],"1":[222]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}