Carta al director
Chatbots and psychiatry: Beware of hallucinations of artificial intelligence
Chatbots y psiquiatría: precaución con las alucinaciones de la Inteligencia Artificial
Read
205
Timeswas read the article
48
Total PDF
157
Total HTML
Share statistics
array:22 [ "pii" => "S1134593423000039" "issn" => "11345934" "doi" => "10.1016/j.psiq.2023.100394" "estado" => "S300" "fechaPublicacion" => "2023-01-01" "aid" => "100394" "copyright" => "Elsevier España, S.L.U.. All rights reserved" "copyrightAnyo" => "2023" "documento" => "simple-article" "crossmark" => 1 "subdocumento" => "cor" "cita" => "Psiquiatr Biol . 2023;30:" "abierto" => array:3 [ "ES" => false "ES2" => false "LATM" => false ] "gratuito" => false "lecturas" => array:1 [ "total" => 0 ] "itemAnterior" => array:18 [ "pii" => "S1134593423000040" "issn" => "11345934" "doi" => "10.1016/j.psiq.2023.100395" "estado" => "S300" "fechaPublicacion" => "2023-01-01" "aid" => "100395" "copyright" => "Elsevier España, S.L.U." "documento" => "article" "crossmark" => 1 "subdocumento" => "rev" "cita" => "Psiquiatr Biol . 2023;30:" "abierto" => array:3 [ "ES" => false "ES2" => false "LATM" => false ] "gratuito" => false "lecturas" => array:1 [ "total" => 0 ] "es" => array:12 [ "idiomaDefecto" => true "cabecera" => "<span class="elsevierStyleTextfn">Revisión</span>" "titulo" => "Relación entre la cantidad y el tipo de alcohol (destilado versus fermentado) consumido y el riesgo de aparición de deterioro cognitivo compatible con un síndrome demencial en personas mayores de 60 años: revisión sistemática" "tienePdf" => "es" "tieneTextoCompleto" => "es" "tieneResumen" => array:2 [ 0 => "es" 1 => "en" ] "titulosAlternativos" => array:1 [ "en" => array:1 [ "titulo" => "Relationship between the amount and type of alcohol (distilled versus fermented) consumed and the risk of cognitive impairment compatible with a dementia syndrome in people over 60 years of age: A systematic review" ] ] "contieneResumen" => array:2 [ "es" => true "en" => true ] "contieneTextoCompleto" => array:1 [ "es" => true ] "contienePdf" => array:1 [ "es" => true ] "resumenGrafico" => array:2 [ "original" => 0 "multimedia" => array:8 [ "identificador" => "f0005" "etiqueta" => "Figura 1" "tipo" => "MULTIMEDIAFIGURA" "mostrarFloat" => true "mostrarDisplay" => false "figura" => array:1 [ 0 => array:4 [ "imagen" => "gr1.jpeg" "Alto" => 2243 "Ancho" => 2715 "Tamanyo" => 338474 ] ] "detalles" => array:1 [ 0 => array:3 [ "identificador" => "al0005" "detalle" => "Figura " "rol" => "short" ] ] "descripcion" => array:1 [ "es" => "<p id="sp0005" class="elsevierStyleSimplePara elsevierViewall">Procedimiento de selección de artículos para la revisión sistemática.</p>" ] ] ] "autores" => array:1 [ 0 => array:2 [ "autoresLista" => "Josep Deví-Bastida, Helena Puig-Martel, Santiago Llana-Cortés, Alexandra García-Guix, M. Teresa Abellán-Vidal, M. Dolores López-Villegas" "autores" => array:6 [ 0 => array:2 [ "nombre" => "Josep" "apellidos" => "Deví-Bastida" ] 1 => array:2 [ "nombre" => "Helena" "apellidos" => "Puig-Martel" ] 2 => array:2 [ "nombre" => "Santiago" "apellidos" => "Llana-Cortés" ] 3 => array:2 [ "nombre" => "Alexandra" "apellidos" => "García-Guix" ] 4 => array:2 [ "nombre" => "M. Teresa" "apellidos" => "Abellán-Vidal" ] 5 => array:2 [ "nombre" => "M. Dolores" "apellidos" => "López-Villegas" ] ] ] ] ] "idiomaDefecto" => "es" "EPUB" => "https://multimedia.elsevier.es/PublicationsMultimediaV1/item/epub/S1134593423000040?idApp=UINPBA00004N" "url" => "/11345934/0000003000000001/v1_202308071525/S1134593423000040/v1_202308071525/es/main.assets" ] "en" => array:10 [ "idiomaDefecto" => true "cabecera" => "<span class="elsevierStyleTextfn">Carta al director</span>" "titulo" => "Chatbots and psychiatry: Beware of hallucinations of artificial intelligence" "tieneTextoCompleto" => true "autores" => array:1 [ 0 => array:4 [ "autoresLista" => "Jesús Enrique Ibáñez Vizoso, Agustín Bermejo Pastor" "autores" => array:2 [ 0 => array:4 [ "nombre" => "Jesús Enrique" "apellidos" => "Ibáñez Vizoso" "email" => array:2 [ 0 => "ibanez.vizoso@gmail.com" 1 => "jesusenrique.ibanez@salud.madrid.org" ] "referencia" => array:2 [ 0 => array:2 [ "etiqueta" => "<span class="elsevierStyleSup">a</span>" "identificador" => "af0005" ] 1 => array:2 [ "etiqueta" => "<span class="elsevierStyleSup">*</span>" "identificador" => "cr0005" ] ] ] 1 => array:3 [ "nombre" => "Agustín" "apellidos" => "Bermejo Pastor" "referencia" => array:1 [ 0 => array:2 [ "etiqueta" => "<span class="elsevierStyleSup">b</span>" "identificador" => "af0010" ] ] ] ] "afiliaciones" => array:2 [ 0 => array:3 [ "entidad" => "Instituto de Psiquiatría y Salud Mental, Hospital Clínico San Carlos, Madrid, Spain" "etiqueta" => "a" "identificador" => "af0005" ] 1 => array:3 [ "entidad" => "Servicio de Psiquiatría, Hospital Universitario Puerta de Hierro, Majadahonda, Spain" "etiqueta" => "b" "identificador" => "af0010" ] ] "correspondencia" => array:1 [ 0 => array:3 [ "identificador" => "cr0005" "etiqueta" => "⁎" "correspondencia" => "Corresponding author." ] ] ] ] "titulosAlternativos" => array:1 [ "es" => array:1 [ "titulo" => "Chatbots y psiquiatría: precaución con las alucinaciones de la Inteligencia Artificial" ] ] "textoCompleto" => "<span class="elsevierStyleSections"><p id="p0005" class="elsevierStylePara elsevierViewall">Artificial intelligence (AI) is a branch of computer science that seeks to develop algorithms and techniques that enable computers to learn and perform tasks that would otherwise require human intelligence. These tasks include pattern recognition, decision making, natural language processing, and complex problem solving.<a class="elsevierStyleCrossRef" href="#bb0005"><span class="elsevierStyleSup">1</span></a></p><p id="p0010" class="elsevierStylePara elsevierViewall">In recent years, the field of psychiatry has seen a significant increase in the use of AI for various applications such as diagnostic assistance, symptom tracking, disease course prediction, psychoeducation and treatment. As promising illustrations of this, BiAffect is a mobile application that uses keyboard metadata to predict affective episodes in bipolar patients, while Avatar Therapy involves the utilization of computer-generated images of faces which interact with patients with schizophrenia via intelligent algorithms, reducing the frequency of hallucinations and the distress associated to them.<a class="elsevierStyleCrossRef" href="#bb0005"><span class="elsevierStyleSup">1</span></a></p><p id="p0015" class="elsevierStylePara elsevierViewall">A chatbot is an AI system that simulates a conversation with humans through the use of natural language. They can be deployed in mobile applications, websites, messaging, etc., and have a wide variety of uses. In Mental Health there is a growing interest in them. They have been used mainly for psychoeducation or to promote adherence to treatment, and potential therapeutic uses are being investigated.<a class="elsevierStyleCrossRef" href="#bb0010"><span class="elsevierStyleSup">2</span></a></p><p id="p0020" class="elsevierStylePara elsevierViewall">Recently, on November 30, 2022, the natural language processing model ChatGPT, originally based on GPT-3 and developed by OpenAI, was released. It is capable of generating natural and coherent responses by attending to the context and intent behind a question or a conversation; It is, therefore, a very advanced chatbot. Its potential has dazzled part of the scientific and academic world: one author published the response he obtained when he asked for a reasoned opinion on the use of rapamycin from a certain perspective<a class="elsevierStyleCrossRef" href="#bb0015"><span class="elsevierStyleSup">3</span></a>; meanwhile, in Nature it has been asked if this technology means the end of the essay as a method of evaluation of students.<a class="elsevierStyleCrossRef" href="#bb0020"><span class="elsevierStyleSup">4</span></a></p><p id="p0025" class="elsevierStylePara elsevierViewall">In this context, it is natural to ask how we can take advantage of these tools in our clinical, scientific or teaching activity. As a simple test, we entered into ChatGPT the following text: “Tell me the prevalence of bipolar disorder. The data must be based on recent and good quality publications, and be properly referenced”. The program's response, in a very convincing style, began like this: “The prevalence of bipolar disorder varies depending on the population and the study methodology used. According to a study published in 2018 in the journal “JAMA Psychiatry“ that reviewed data from more than 200 epidemiological studies, the global prevalence of type 1 bipolar disorder (BP-1) is 0.6%, while the prevalence of the disorder bipolar type 2 (BP-2) is 1.1%. However, these data can vary significantly depending on the geographic region and the population studied”. The response went on to refer to another study and concluded with a cautious paragraph. It then gave the complete references of the 2 mentioned studies.</p><p id="p0030" class="elsevierStylePara elsevierViewall">After this surprising and apparent response we proceeded to check the data and the references. Neither of the two DOIs corresponded to an existing scientific paper. The first paper seemed to be inspired by another published in Archives of General Psychiatry in 2011, but the methodology and some data did not match what was described in the answer; with the second paper the problem was even more pronounced. That is, the studies had been invented or “hallucinated”.</p><p id="p0035" class="elsevierStylePara elsevierViewall">AI and language processing models are prone to hallucinating (in AI jargon, generating meaningless content), which degrades the performance of these technologies and in medicine can lead to safety issues in their potential uses, as well as violations of private data.<a class="elsevierStyleCrossRef" href="#bb0025"><span class="elsevierStyleSup">5</span></a> A recent review about the expectations regarding AI in psychiatry explains various factors that contribute to the low maturity of this technology, especially for its routine use in clinical practice. Its implementation and use in our activities will require a greater maturity of the technology, identifying and solving the problems that inevitably arise, and, importantly, educating healthcare personnel in the fundamentals of AI.<a class="elsevierStyleCrossRef" href="#bb0030"><span class="elsevierStyleSup">6</span></a></p></span>" "pdfFichero" => "main.pdf" "tienePdf" => true "bibliografia" => array:2 [ "titulo" => "References" "seccion" => array:1 [ 0 => array:2 [ "identificador" => "bs0005" "bibliografiaReferencia" => array:6 [ 0 => array:3 [ "identificador" => "bb0005" "etiqueta" => "1" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Artificial intelligence and chatbots in psychiatry" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:3 [ 0 => "K.T. Pham" 1 => "A. Nabizadeh" 2 => "S. Selek" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1007/s11126-022-09973-8" "Revista" => array:7 [ "tituloSerie" => "Psychiatr Q" "fecha" => "2022" "volumen" => "93" "numero" => "1" "paginaInicial" => "249" "paginaFinal" => "253" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/35212940" "web" => "Medline" ] ] ] ] ] ] ] ] 1 => array:3 [ "identificador" => "bb0010" "etiqueta" => "2" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Chatbots and conversational agents in mental health: a review of the psychiatric landscape" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:5 [ 0 => "A.N. Vaidyam" 1 => "H. Wisniewski" 2 => "J.D. Halamka" 3 => "M.S. Kashavan" 4 => "J.B. Torous" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:6 [ "tituloSerie" => "Can J Psychiatr Rev Can Psychiatr" "fecha" => "2019" "volumen" => "64" "numero" => "7" "paginaInicial" => "456" "paginaFinal" => "464" ] ] ] ] ] ] 2 => array:3 [ "identificador" => "bb0015" "etiqueta" => "3" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Rapamycin in the context of Pascal’s Wager: generative pre-trained transformer perspective" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:2 [ 0 => "ChatGPT Generative Pre-trained Transformer" 1 => "A. Zhavoronkov" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.18632/oncoscience.571" "Revista" => array:6 [ "tituloSerie" => "Oncoscience." "fecha" => "2022" "volumen" => "9" "paginaInicial" => "82" "paginaFinal" => "84" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/36589923" "web" => "Medline" ] ] ] ] ] ] ] ] 3 => array:3 [ "identificador" => "bb0020" "etiqueta" => "4" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "AI bot ChatGPT writes smart essays - should professors worry?" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:1 [ 0 => "C. Stokel-Walker" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:2 [ "tituloSerie" => "Nature" "fecha" => "2022" ] ] ] ] ] ] 4 => array:3 [ "identificador" => "bb0025" "etiqueta" => "5" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Survey of hallucination in natural language generation," "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:6 [ 0 => "Z. Ji" 1 => "N. Lee" 2 => "R. Frieske" 3 => "T. Yu" 4 => "D. Su" 5 => "Y. Xu" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:2 [ "tituloSerie" => "ACM Comput Surv" "fecha" => "2022" ] ] ] ] ] ] 5 => array:3 [ "identificador" => "bb0030" "etiqueta" => "6" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Expectations for artificial intelligence (AI) in psychiatry" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:6 [ 0 => "S. Monteith" 1 => "T. Glenn" 2 => "J. Geddes" 3 => "P.C. Whybrow" 4 => "E. Achtyes" 5 => "M. Bauer" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1007/s11920-022-01378-5" "Revista" => array:7 [ "tituloSerie" => "Curr Psychiatry Rep" "fecha" => "2022" "volumen" => "24" "numero" => "11" "paginaInicial" => "709" "paginaFinal" => "721" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/36214931" "web" => "Medline" ] ] ] ] ] ] ] ] ] ] ] ] ] "idiomaDefecto" => "en" "url" => "/11345934/0000003000000001/v1_202308071525/S1134593423000039/v1_202308071525/en/main.assets" "Apartado" => array:4 [ "identificador" => "95983" "tipo" => "SECCION" "en" => array:2 [ "titulo" => "Carta al director" "idiomaDefecto" => true ] "idiomaDefecto" => "en" ] "PDF" => "https://static.elsevier.es/multimedia/11345934/0000003000000001/v1_202308071525/S1134593423000039/v1_202308071525/en/main.pdf?idApp=UINPBA00004N&text.app=https://www.elsevier.es/" "EPUB" => "https://multimedia.elsevier.es/PublicationsMultimediaV1/item/epub/S1134593423000039?idApp=UINPBA00004N" ]
| Year/Month | Html | Total | |
|---|---|---|---|
| 2023 October | 34 | 13 | 47 |
| 2023 September | 114 | 35 | 149 |
| 2023 August | 7 | 0 | 7 |
| 2023 July | 2 | 0 | 2 |