Elamur Research⌕
☰
home / papers / 2026 / Analyzing Utterance Selection for Unnoticeable Topic Induction in Target-Guided Conversation Systems

Analyzing Utterance Selection for Unnoticeable Topic Induction in Target-Guided Conversation Systems

Paper · 2026 · Kai Yoshida and Koichiro Yoshino

Summary

Target-guided conversation systems steer casual chat toward a preset topic. When the steering is obvious, users can become wary and lose interest. Yoshida and Yoshino study how to choose system responses that move toward the target while keeping that intention hidden. They use SBIS-TGC, a score computed with an external language model. It favours candidates with low perplexity given the dialogue history, meaning they fit the context. It also favours candidates that leave the target hard to predict, measured as high perplexity of the target given the context plus the candidate. An LLM produces candidates that push toward the target to different degrees, and the system picks the one with the best SBIS-TGC score. The authors ran user dialogue experiments comparing an unguided "open" system, a baseline and the proposed method. The proposed method produced fewer utterances judged abrupt and more often reached the target without users guessing it. However, it reached the target less reliably than the baseline. Both score components correlated with human abruptness ratings, which makes surprisal a cheap proxy for perceived induction. The authors also flag the risk of manipulation.

Key points

  • Target achievement: the baseline reached the target in 100% of dialogues, the proposed system in 70.6% and the open system in 0%. The baseline's advantage comes from candidate levels 4–5 naming the target outright.
  • Hidden steering: the proposed system more often reached the target without users identifying it.
    • Keyword-based check: 70.6% for the proposed system (62.5% counting only successful sessions) vs. 46.1% for the baseline.
    • LLM-based check: 47.1% (44.4%) vs. 39.2%.
  • Abruptness ratings: with SBIS-TGC selection, the "Not abrupt" rating occurred more than twice as often as with the baseline. The most abrupt rating occurred less than half as often. Abrupt utterances made it roughly twice as likely that users guessed the target.
  • Correlation with human judgments: Pearson correlation with 3-level abruptness ratings was 0.418 for past (utterance surprisal) and −0.338 for future (target surprisal).
  • Progress over turns: future values fell as dialogues progressed, especially for the baseline and proposed systems. This suggests target surprisal can track progress toward the goal. It also reveals a trade-off between reaching the target and staying unnoticed.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 2The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 works

Cite

Cite as Kai Yoshida and Koichiro Yoshino. 2026. Analyzing Utterance Selection for Unnoticeable Topic Induction in Target-Guided Conversation Systems. Proceedings of the 16th International Workshop on Spoken Dialogue System Technology.

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{yoshida2026analyzing,
  title = {{Analyzing Utterance Selection for Unnoticeable Topic Induction in Target-Guided Conversation Systems}},
  author = {Kai Yoshida and Koichiro Yoshino},
  booktitle = {Proceedings of the 16th International Workshop on Spoken Dialogue System Technology},
  year = {2026},
  url = {https://aclanthology.org/2026.iwsds-1.12/}
}

References

26 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Daniel Adiwardana, Minh-Thang Luong, David R. So, Jamie Hall, Noah Fiedel, Romal Thoppilan, Zi Yang, Apoorv Kulshreshtha, Gaurav Nemade, Yifeng Lu, and Quoc V. Le. 2020. Towards a human-like open-domain chatbot. arXiv:2001.09977.
  2. Huy Quang Dao, Yang Deng, Khanh-Huyen Bui, Dung D. Le, and Lizi Liao. 2024. Experience as source for anticipation and planning: Experiential policy learning for target-driven recommendation dialogues. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 14179–14198, Miami,… find
  3. Yang Deng, Lizi Liao, Liang Chen, Hongru Wang, Wenqiang Lei, and Tat-Seng Chua. 2023. Prompting and evaluating large language models for proactive dialogues: Clarification, target-guided, and non-collaboration. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 10602–106…
  4. Yang Deng, Lizi Liao, Wenqiang Lei, Grace Hui Yang, Wai Lam, and Tat-Seng Chua. 2025. Proactive conversational ai: A comprehensive survey of advancements and opportunities. ACM Trans. Inf. Syst., 43(3). find
  5. Prakhar Gupta, Harsh Jhamtani, and Jeffrey Bigham. 2022. Target-guided dialogue response generation using commonsense and data augmentation. In Findings of the Association for Computational Linguistics: NAACL 2022, pages 1301–1317, Seattle, United States. Association for Computational Linguistics. find
  6. John Hale. 2001. A probabilistic earley parser as a psycholinguistic model. In Proceedings of the second meeting of the North American Chapter of the Association for Computational Linguistics on Language technologies.
  7. Takuya Hiraoka, Graham Neubig, Sakriani Sakti, Tomoki Toda, and Satoshi Nakamura. 2014. Reinforcement learning of cooperative persuasive dialogue policies using framing. In Proceedings of coling 2014, the 25th international conference on computational linguistics: technical papers, pages 1706–1717. find
  8. Yosuke Kishinami, Reina Akama, Shiki Sato, Ryoko Tokuhisa, Jun Suzuki, and Kentaro Inui. 2022. Target-guided open-domain conversation planning. In Proceedings of the 29th International Conference on Computational Linguistics, pages 660–668, Gyeongju, Republic of Korea. International Committee on Com… find
  9. This work was partially supported by JSPS KAK- Roger Levy. 2008. Expectation-based syntactic compre-ENHI Grant Number 23K24910. hension. Cognition, 106(3):1126–1177. 120 find
  10. Raymond Li, Samira Kahou, Hannes Schulz, Vincent Michalski, Laurent Charlin, and Chris Pal. 2018. Towards deep conversational recommendations. arXiv:1812.07617. find
  11. Anqi Liu, Bo Wang, Yue Tan, Dongming Zhao, Kun Huang, Ruifang He, and Yuexian Hou. 2023. MTGP: Multi-turn target-oriented dialogue guided by generative global path with flexible turns. In Findings of the Association for Computational Linguistics: ACL 2023, pages 259–271, Toronto, Canada. Association… find
  12. Marine Riou. 2015. A methodology for the identification of topic transitions in interaction. Discours, 16. Published online on September 9, 2015. Accessed on December 4, 2024. find
  13. Shiki Sato, Jun Baba, Asahi Hentona, Shinji Iwata, Akifumi Yoshimoto, and Koichiro Yoshino. 2025. Proactive user information acquisition via chats on user-favored topics. arXiv:2504.07698. find
  14. Abigail See, Stephen Roller, Douwe Kiela, and Jason Weston. 2019. What makes a good conversation? how controllable attributes affect human judgments. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Vol…
  15. Karin Sevegnani, David M. Howcroft, Ioannis Konstas, and Verena Rieser. 2021. OTTers: One-turn topic transitions for open-domain dialogue. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Proces… find
  16. C. E. Shannon. 1948. A mathematical theory of communication. The Bell System Technical Journal, 27(3):379–423. find
  17. Jianheng Tang, Tiancheng Zhao, Chenyan Xiong, Xiaodan Liang, Eric Xing, and Zhiting Hu. 2019. Target-guided open-domain conversation. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5624–5634, Florence, Italy. Association for Computational Linguistic…
  18. Jian Wang, Yi Cheng, Dongding Lin, Chak Leong, and Wenjie Li. 2023. Target-oriented proactive dialogue systems with personalization: Problem formulation and dataset curation. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 1132–1143, Singapore. Associ… find
  19. Xuewei Wang, Weiyan Shi, Richard Kim, Yoojung Oh, Sijia Yang, Jingwen Zhang, and Zhou Yu. 2019. Persuasion for good: Towards a personalized persuasive dialogue system for social good. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5635–5649, Florenc…
  20. Wenquan Wu, Zhen Guo, Xiangyang Zhou, Hua Wu, Xiyuan Zhang, Rongzhong Lian, and Haifeng Wang. 2019. Proactive human-machine conversation with explicit conversation goal. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3794–3804, Florence, Italy. Asso…
  21. Jingxuan Yang, Si Li, and Jun Guo. 2021. Multi-turn target-guided topic prediction with Monte Carlo tree search. In Proceedings of the 18th International Conference on Natural Language Processing (ICON), pages 324–334, National Institute of Technology Silchar, Silchar, India. NLP Association of Indi… find
  22. Koichiro Yoshino, Yu Suzuki, and Satoshi Nakamura. 2017. Information navigation system with discovering user interests. In Proceedings of the 18th Annual SIGdial Meeting on Discourse and Dialogue, pages 356–359. find
  23. Steve Young, Milica Gašić, Simon Keizer, François Mairesse, Jost Schatzmann, Blaise Thomson, and Kai Yu. 2010. The hidden information state model: A practical framework for pomdp-based spoken dialogue management. Computer Speech & Language, 24(2):150–174. find
  24. Zhonghua Zheng, Lizi Liao, Yang Deng, Ee-Peng Lim, Minlie Huang, and Liqiang Nie. 2024. Thoughts to target: Enhance planning for target-driven conversation. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 21108–21124, Miami, Florida, USA. Association… find
  25. Kun Zhou, Yuanhang Zhou, Wayne Xin Zhao, Xiaoke Wang, and Ji-Rong Wen. 2020. Towards topic-guided conversational recommender system. In Proceedings of the 28th International Conference on Computational Linguistics, pages 4128–4139, Barcelona, Spain (Online). International Committee on Computational… find
  26. Jian Wang, Dongding Lin, and Wenjie Li. 2024. A target-driven planning approach for goal-directed dialog systems. IEEE Transactions on Neural Networks and Learning Systems, 35(8):10475–10487. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yoshida-2026-analyzing-utterance-selection-unnoticeable-topic/