[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-141099-105":3,"detail-sidebar-cat-0-en-105":80,"doc-detail-141099-en":130},{"code":4,"msg":5,"data":6},0,"ok",{"site_id":7,"language":8,"slug":9,"title":10,"keywords":11,"description":12,"schema_data":13,"social_meta":73,"head_meta":75,"extra_data":77,"updated_unix":79},105,"en","sbis-tgc-a-surprisal-based-metric-for-target-guided-conversation-without-perceived-induction","SBIS-TGC - A Surprisal-Based Metric for Target-Guided Conversation without Perceived Induction","","Target-guided conversational systems require moving dialogue toward the system’s own goal while keeping the user from feeling induced or noticing the system’s underlying intentions. This study introduces Surprisal-Based Induction Score for Target-Guided Conversation (SBIS-TGC), an automatic evaluation metric that quantifies perceived induction by computing surprisal between system utterances using an external language model. Dialogue experiments show SBIS-TGC-guided selection reduces perceived abruptness in target-oriented settings, enabling more natural user induction, while acknowledging remaining challenges for balancing goal achievement and low awareness.",{"@graph":14,"@context":72},[15,34,55],{"@type":16,"itemListElement":17},"BreadcrumbList",[18,23,27,31],{"item":19,"name":20,"@type":21,"position":22},"https://docshare.wps.com","Home","ListItem",1,{"item":24,"name":25,"@type":21,"position":26},"https://docshare.wps.com/document/","Document",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/document/research-report/","Research & Report",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/sbis-tgc-a-surprisal-based-metric-for-target-guided-conversation-without-perceived-induction/141099/",4,{"url":32,"name":10,"@type":35,"image":36,"author":41,"headline":10,"publisher":44,"fileFormat":47,"inLanguage":8,"description":12,"dateModified":48,"datePublished":49,"encodingFormat":47,"isAccessibleForFree":50,"interactionStatistic":51},"DigitalDocument",{"url":37,"@type":38,"width":39,"height":40},"https://docshare.wps.com/thumbnails/sbis-tgc-a-surprisal-based-metric-for-target-guided-conversation-without-perceived-induction/141099.png","ImageObject",300,407,{"name":42,"@type":43},"Jordan Avery","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-09-13","2026-08-25",true,{"@type":52,"interactionType":53,"userInteractionCount":30},"InteractionCounter",{"@type":54},"ViewAction",{"@type":56,"mainEntity":57},"FAQPage",[58,64,68],{"name":59,"@type":60,"acceptedAnswer":61},"What problem does SBIS-TGC address in target-guided conversation?","Question",{"text":62,"@type":63},"It quantifies how much system utterances induce the user, aiming to achieve the system’s target without making the user feel induced or notice the system’s intentions.","Answer",{"name":65,"@type":60,"acceptedAnswer":66},"How is SBIS-TGC computed?",{"text":67,"@type":63},"SBIS-TGC uses surprisal theory by measuring the surprisal between utterances via an external language model.",{"name":69,"@type":60,"acceptedAnswer":70},"What do dialogue experiments show using SBIS-TGC?",{"text":71,"@type":63},"Selecting utterances based on SBIS-TGC reduces perceived abruptness in target-guided settings, supporting more natural induction of users.","https://schema.org",{"og:url":32,"og:type":74,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":76,"canonical":32},"index,follow",{"doc_id":78,"site_id":7},141099,1787649047,{"code":4,"msg":81,"data":82},"success",[83,87,91,95,100,105,110,114,119,122,126],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":84,"show_sort_weight":85,"slug":86},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":88,"show_sort_weight":89,"slug":90},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":92,"show_sort_weight":93,"slug":94},"Exam",70,"exam",{"id":96,"doc_module":4,"doc_module_name":25,"category_name":97,"show_sort_weight":98,"slug":99},5,"Comic",60,"comic",{"id":101,"doc_module":4,"doc_module_name":25,"category_name":102,"show_sort_weight":103,"slug":104},6,"Technology",50,"technology",{"id":106,"doc_module":4,"doc_module_name":25,"category_name":107,"show_sort_weight":108,"slug":109},7,"Healthcare",40,"healthcare",{"id":111,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":112,"slug":113},8,30,"research-report",{"id":115,"doc_module":4,"doc_module_name":25,"category_name":116,"show_sort_weight":117,"slug":118},9,"Religion & Spirituality",20,"religion-spirituality",{"id":117,"doc_module":4,"doc_module_name":25,"category_name":120,"show_sort_weight":117,"slug":121},"World Cup","world-cup",{"id":123,"doc_module":4,"doc_module_name":25,"category_name":124,"show_sort_weight":123,"slug":125},10,"Lifestyle","lifestyle",{"id":127,"doc_module":4,"doc_module_name":25,"category_name":128,"show_sort_weight":96,"slug":129},19,"General","general",{"code":4,"msg":81,"data":131},{"doc_id":78,"user_id":132,"nickname":42,"user_avatar":133,"doc_module":4,"category_id":111,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":30,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":139,"language":140,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":12,"update_tm":79,"read_time":144},1099523882367,"https://ap-avatar.wpscdn.com/davatar_9964176cb1d06d4a9deccf72a44ae3dc","General Paper  \nPeer-Reviewed  \nSBIS-TGC: A Surprisal-Based Metric for Target-Guided Conversation without Perceived Induction  \nKai Yoshidaa,b and Koichiro Yoshinoc,a,b  \nDialogue tasks in which the system has dialogue targets different from those of the user, such as recommendation, persuasion, or information gathering, are collectively referred to as target-guided conversational systems and have been researched widely.  \nTo realize such target-guided conversations, natural topic induction is essential, and thus, prior studies have mainly focused on topic switching and transition strategies.  \nHowever, in advancing a dialogue toward the system’s own targets, it is also essential to ensure that the user neither feels induction nor becomes aware of the system’s underlying intentions, to preserve a better user experience. This study proposes Surprisal-Based Induction Score for Target-Guided Conversation (SBIS-TGC), a novel automatic evaluation metric that quantifies the perceived induction in system utterances without alerting the user to the system’s target or strategy. SBIS-TGC quantifies the degree of induction by computing the surprisal between utterances using an external language model. Dialogue experiments using a system that selects utterances based on SBIS-TGC demonstrate that this approach reduces the perceived abruptness of dialogue in target-guided settings. This enables the system to induce users naturally. However, achieving both target accomplishment and low user awareness remains challenging, and future work must explore further refinements.  \nKey Words: Target-Guided Conversation, Surprisal, Dialogue System, Large Language Models  \n1 Introduction  \nIn dialogue system research, the dialogue target plays a crucial role. In task-oriented dialogue, specific targets aligned with the task are typically defined. Even in non-task-oriented dialogue, it is common for some form of dialogue objective to be set. Many studies have focused on designing dialogue systems that aim to fulfill the user’s intended targets (Young et al. 2010) . Dialogues in which the system has a dialogue target that differs from the user’s are referred to as target-guided conversations, and have attracted increasing research attention (Tang et al. 2019; Zhou et al. 2020; Kishinami et al. 2022; Deng et al. 2023; Liu et al. 2023) . In target-guided conversation,  \na Nara Institute of Science and Technology b Guardian Robot Project, RIKEN c Institute of Science Tokyo  \n(C) The Association for Natural Language Processing. Licensed under CC BY 4.0 ([https://creativecommons.org/licenses/by/4.0/](https://creativecommons.org/licenses/by/4.0/)).  \nthe system has its own independent dialogue target, necessitating the ability to balance the user’s preferred topics with the system’s own goal, and to proactively induce the dialogue toward achieving that objective.  \nResearch on target-guided conversation has been actively conducted not only in task-oriented dialogue domains such as recommendation (Yoshino et al. 2017; Li et al. 2018; Zhou et al. 2020) and persuasion (Hiraoka et al. 2014; Wang et al. 2019), but also in non-task-oriented dialogue settings (Tang et al. 2019; Gupta et al. 2022; Deng et al. 2023, 2025) . In particular, non-task-oriented target-guided conversations have seen increasing interest in tasks where the system induces the user toward a specific topic. Such frameworks have demonstrated their value especially in marketing and related domains (Sato et al. 2025) .  \nIn prior studies on target-guided conversation, a common approach is to infer a sequence of topic transitions from the current topic to the target topic in advance, and to induce the dialogue accordingly (Tang et al. 2019; Zhou et al. 2020; Sevegnani et al. 2021; Kishinami et al. 2022; Gupta et al. 2022; Liu et al. 2023; Deng et al. 2023; Zheng et al. 2024) . These studies emphasize the importance of coherent topic transitions between utterances, aiming to naturally reach a predefined target t","cbCaiiqFxAV4uqKr","https://ap.wps.com/l/cbCaiiqFxAV4uqKr","pdf",477035,28,"English","# Introduction\n## Target-guided conversation and user perception\n## Surprisal-based evaluation approach\n## Proposed metric and contributions","[{\"question\":\"What problem does SBIS-TGC address in target-guided conversation?\",\"answer\":\"It quantifies how much system utterances induce the user, aiming to achieve the system’s target without making the user feel induced or notice the system’s intentions.\"},{\"question\":\"How is SBIS-TGC computed?\",\"answer\":\"SBIS-TGC uses surprisal theory by measuring the surprisal between utterances via an external language model.\"},{\"question\":\"What do dialogue experiments show using SBIS-TGC?\",\"answer\":\"Selecting utterances based on SBIS-TGC reduces perceived abruptness in target-guided settings, supporting more natural induction of users.\"}]","SBIS-TGC - A Surprisal-Based Metric for Target-Guided Conversation without Perceived Induction | PDF",71]