{"id":"https://openalex.org/W7223661740","doi":"https://doi.org/10.2196/preprints.114747","title":"Validating Large Language Model Judges for Surgical Patient Education: Multi-Expert Study of Reference Standard Reliability and Directional Error (Preprint)","display_name":"Validating Large Language Model Judges for Surgical Patient Education: Multi-Expert Study of Reference Standard Reliability and Directional Error (Preprint)","publication_year":2026,"publication_date":"2026-10-18","ids":{"openalex":"https://openalex.org/W7223661740","doi":"https://doi.org/10.2196/preprints.114747"},"language":"en","primary_location":{"id":"doi:10.2196/preprints.114747","doi":"https://doi.org/10.2196/preprints.114747","is_oa":false,"landing_page_url":"https://doi.org/10.2196/preprints.114747","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"acceptedVersion","is_accepted":true,"is_published":false,"raw_source_name":null,"raw_type":"posted-content"},"volume":null,"type":"preprint","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101575123","display_name":"Onur Erdo\u011fan","orcid":"https://orcid.org/0000-0002-6171-3042","observed_orcids":["https://orcid.org/0000-0002-6171-3042"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Onur Erdo\u011fan","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0000-0002-6171-3042","affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5090590840","display_name":"Yener \u015eahin","orcid":"https://orcid.org/0000-0001-9539-357X","observed_orcids":["https://orcid.org/0000-0001-9539-357X"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yener \u015eahin","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0000-0001-9539-357X","affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5044618366","display_name":"Mustafa \u015eakar","orcid":"https://orcid.org/0000-0002-8344-4074","observed_orcids":["https://orcid.org/0000-0002-8344-4074"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mustafa Sakar","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0000-0002-8344-4074","affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5003123783","display_name":"Adnan Da\u011f\u00e7\u0131nar","orcid":"https://orcid.org/0000-0001-9532-7820","observed_orcids":["https://orcid.org/0000-0001-9532-7820"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Adnan Da\u011f\u00e7\u0131nar","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0000-0001-9532-7820","affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"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/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.9887999892234802,"subfield":{"id":"https://openalex.org/subfields/2718","display_name":"Health Informatics"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},"topics":[{"id":"https://openalex.org/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.9887999892234802,"subfield":{"id":"https://openalex.org/subfields/2718","display_name":"Health Informatics"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},{"id":"https://openalex.org/T13309","display_name":"Reliability and Agreement in Measurement","score":0.0010999999940395355,"subfield":{"id":"https://openalex.org/subfields/1804","display_name":"Statistics, Probability and Uncertainty"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T13117","display_name":"Health Education and Validation","score":0.0007999999797903001,"subfield":{"id":"https://openalex.org/subfields/3304","display_name":"Education"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/craniosynostosis","display_name":"craniosynostosis","score":0.9850000143051147},{"id":"https://openalex.org/keywords/inter-rater-reliability","display_name":"inter-rater reliability","score":0.8560000061988831},{"id":"https://openalex.org/keywords/directional-error","display_name":"directional error","score":0.9210000038146973},{"id":"https://openalex.org/keywords/safety-classification","display_name":"safety classification","score":0.42899999022483826},{"id":"https://openalex.org/keywords/jev","display_name":"Jev","score":0.5559999942779541},{"id":"https://openalex.org/keywords/clinical-decision-support","display_name":"clinical decision support","score":0.33399999141693115},{"id":"https://openalex.org/keywords/generative-ai-in-medicine","display_name":"generative AI in medicine","score":0.4729999899864197},{"id":"https://openalex.org/keywords/neurosurgical-training","display_name":"neurosurgical training","score":0.4740000069141388}],"concepts":[{"id":"https://openalex.org/C61863361","wikidata":"https://www.wikidata.org/wiki/Q470749","display_name":"Inter-rater reliability","level":3,"score":0.7064999938011169},{"id":"https://openalex.org/C111640148","wikidata":"https://www.wikidata.org/wiki/Q847349","display_name":"Rubric","level":2,"score":0.6970999836921692},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.6480000019073486},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6115000247955322},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.607699990272522},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.5260000228881836},{"id":"https://openalex.org/C191147762","wikidata":"https://www.wikidata.org/wiki/Q186289","display_name":"Human reliability","level":3,"score":0.48410001397132874},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.44209998846054077},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.43459999561309814},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.43369999527931213},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.39259999990463257},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.3395000100135803},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.32919999957084656},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.3203999996185303},{"id":"https://openalex.org/C21200559","wikidata":"https://www.wikidata.org/wiki/Q7451068","display_name":"Sensitivity (control systems)","level":2,"score":0.32019999623298645},{"id":"https://openalex.org/C3017944768","wikidata":"https://www.wikidata.org/wiki/Q1450463","display_name":"Poison control","level":2,"score":0.31119999289512634},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.3059999942779541},{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.3005000054836273},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.29260000586509705},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.29170000553131104},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2865999937057495},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.28130000829696655},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.2759999930858612},{"id":"https://openalex.org/C70364389","wikidata":"https://www.wikidata.org/wiki/Q18757","display_name":"Validity","level":3,"score":0.2750000059604645},{"id":"https://openalex.org/C2777548347","wikidata":"https://www.wikidata.org/wiki/Q5456937","display_name":"Flagging","level":2,"score":0.27250000834465027},{"id":"https://openalex.org/C2779190172","wikidata":"https://www.wikidata.org/wiki/Q4913888","display_name":"Binary data","level":3,"score":0.26010000705718994}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.2196/preprints.114747","doi":"https://doi.org/10.2196/preprints.114747","is_oa":false,"landing_page_url":"https://doi.org/10.2196/preprints.114747","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"acceptedVersion","is_accepted":true,"is_published":false,"raw_source_name":null,"raw_type":"posted-content"}],"best_oa_location":null,"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":21,"referenced_works":["https://openalex.org/W1983897914","https://openalex.org/W1996290308","https://openalex.org/W2070285512","https://openalex.org/W2149490731","https://openalex.org/W2327037637","https://openalex.org/W4281692003","https://openalex.org/W4367310920","https://openalex.org/W4403880728","https://openalex.org/W4406152263","https://openalex.org/W4408078062","https://openalex.org/W4408197148","https://openalex.org/W4414891332","https://openalex.org/W4415024906","https://openalex.org/W4415797148","https://openalex.org/W4415802775","https://openalex.org/W4415930269","https://openalex.org/W4415985271","https://openalex.org/W4417075517","https://openalex.org/W7124460067","https://openalex.org/W7133239970","https://openalex.org/W7169776463"],"related_works":["https://openalex.org/W7214209004","https://openalex.org/W7214303826","https://openalex.org/W7223252559","https://openalex.org/W7165368332","https://openalex.org/W7134036278","https://openalex.org/W7172246804","https://openalex.org/W7167091752","https://openalex.org/W4413971377","https://openalex.org/W7164992558","https://openalex.org/W7154531043"],"abstract_inverted_index":{"BACKGROUND":[0],"Large":[1],"language":[2],"models":[3],"(LLMs)":[4],"are":[5],"increasingly":[6],"used":[7],"to":[8,44,76,164,182,210,298],"evaluate":[9,45],"AI-generated":[10],"medical":[11,314],"content,":[12],"yet":[13],"judge":[14],"validity":[15,311],"is":[16,364],"commonly":[17],"inferred":[18],"from":[19,208,296],"agreement":[20,333],"with":[21,356],"human":[22,31,59],"ratings":[23],"without":[24],"fully":[25],"characterizing":[26],"the":[27,30,35,126,252,269,301,310,346],"reliability":[28,196],"of":[29,38,312],"reference":[32,317],"standard":[33],"or":[34],"clinical":[36,171],"direction":[37],"disagreement.":[39],"OBJECTIVE":[40],"This":[41],"study":[42],"aimed":[43],"LLM":[46,84],"judges":[47,85],"for":[48],"surgical":[49],"patient":[50],"education":[51],"using":[52,155],"a":[53,77,146,157],"validation":[54,363],"framework":[55],"that":[56,161],"jointly":[57],"considers":[58],"reference-standard":[60],"reliability,":[61,135,318],"judge-expert":[62],"concordance,":[63],"directional":[64,138],"error,":[65,139,320],"and":[66,111,125,145,169,179,189,213,220,224,231,245,263,323,340,354],"safety-sensitive":[67],"classification.":[68],"We":[69,132],"additionally":[70],"examined":[71,133],"whether":[72],"these":[73],"principles":[74],"extend":[75],"decision-native":[78,158],"probabilistic":[79,159],"evaluator.":[80],"METHODS":[81],"Five":[82],"generative":[83],"evaluated":[86,153],"100":[87],"craniosynostosis":[88],"patient-education":[89],"responses":[90,106,114,244,250,279,285],"generated":[91],"by":[92],"five":[93,358],"contestant":[94],"LLMs":[95],"(20":[96],"questions":[97],"\u00d7":[98],"5":[99],"generators);":[100],"three":[101],"neurosurgeons":[102],"independently":[103],"rated":[104],"all":[105],"on":[107],"Accuracy,":[108],"Safety,":[109],"Empathy,":[110],"Completeness.":[112],"Contestant":[113],"were":[115,352],"collected":[116],"in":[117,122,129,300],"January":[118],"2026,":[119,124],"generative-judge":[120],"evaluations":[121],"March":[123],"Jev-1.13.0":[127],"benchmark":[128,151],"October":[130],"2026.":[131],"expert":[134,194],"absolute":[136],"agreement,":[137],"cluster-bootstrap":[140],"uncertainty,":[141],"safety":[142,186,338],"classification,":[143],"self-preference,":[144],"five-judge":[147,234],"ensemble.":[148],"The":[149],"same":[150],"was":[152,197],"separately":[154],"Jev-1.13.0,":[156],"evaluator":[160,341],"assigns":[162],"probabilities":[163],"predefined":[165],"ordinal":[166],"scoring":[167],"levels":[168],"binary":[170],"decisions,":[172,187],"rather":[173],"than":[174],"generating":[175],"single":[176],"rubric":[177],"scores":[178,258],"narrative":[180],"rationales,":[181],"examine":[183],"repeatability,":[184],"typed":[185],"calibration,":[188],"configuration":[190,292,342],"sensitivity":[191],"RESULTS":[192],"Individual":[193],"interrater":[195],"limited":[198],"(ICC[A,1]=0.322;":[199],"three-expert":[200],"mean":[201,264],"ICC[A,3]=0.587).":[202],"Generative":[203],"judge\u2013expert":[204],"absolute-agreement":[205],"ICCs":[206],"ranged":[207],"0.182":[209],"0.370.":[211],"Opus-4.6":[212],"GPT-5.4":[214],"had":[215,227],"negative":[216],"signed":[217,265,319],"biases":[218,229],"(\u22120.374":[219],"\u22120.426);":[221],"Llama-4":[222],"Maverick":[223],"DeepSeek":[225],"Reasoner":[226],"positive":[228],"(+0.242":[230],"+0.215).":[232],"A":[233],"dimension-median":[235],"ensemble":[236],"yielded":[237,259],"ICC(A,1)=0.454":[238],"but":[239,280],"detected":[240,276],"0/5":[241],"expert-defined":[242],"Safety-inadequate":[243,278],"automatically":[246],"accepted":[247],"6/10":[248],"overall-inadequate":[249],"at":[251],"fixed":[253],"4.0":[254],"cutoff.":[255],"Jev":[256,291],"matched":[257],"ICC(C,1)=0.440,":[260],"ICC(A,1)=0.141,":[261],"MAE=0.702":[262],"error":[266],"\u22120.690":[267],"against":[268],"corrected":[270],"reference.":[271],"Its":[272],"modal":[273],"Safety":[274],"decisions":[275],"4/5":[277],"falsely":[281],"flagged":[282],"40\u201342/95":[283],"Safety-adequate":[284],"per":[286],"run.":[287],"An":[288],"exploratory":[289],"minimal-state":[290],"increased":[293],"false":[294,321],"acceptance":[295],"0/10":[297],"7/10":[299],"archived":[302],"comparison.":[303],"CONCLUSIONS":[304],"Agreement":[305],"alone":[306],"does":[307],"not":[308,336],"establish":[309],"automated":[313],"evaluators.":[315],"Human":[316],"acceptance,":[322],"review":[324],"burden":[325],"should":[326],"be":[327],"considered":[328],"alongside":[329],"aggregate":[330],"concordance.":[331],"Ensemble":[332],"gains":[334],"need":[335],"improve":[337],"detection,":[339],"changes":[343],"can":[344],"alter":[345],"risk":[347],"profile.":[348],"All":[349],"screening":[350],"analyses":[351],"internal":[353],"exploratory,":[355],"only":[357],"safety-inadequate":[359],"cases;":[360],"independent":[361],"safety-enriched":[362],"required.":[365]},"counts_by_year":[],"updated_date":"2026-10-29T09:56:32.521200","created_date":"2026-10-27T00:00:00"}