PKSHA Technology and Tohoku University's Jointly Developed Dialogue AI Technology Paper Accepted at EMNLP 2025 Industry Track

— Paper on "Persuasive Dialogue AI," which understands human psychology and situations to encourage behavioral change, accepted—to be utilized for the advancement of "PKSHA AI Agents" —

PKSHA Technology Inc. (Head Office: Bunkyo-ku, Tokyo; Representative Director: UENOYAMA katsuya) announced today that a paper summarizing the results of its joint research with the Language AI Research Center at Tohoku University (Location: Aoba-ku, Sendai; Center Director: Jun Suzuki) has been accepted for the Industry Track of EMNLP 2025 (The 2025 Conference on Empirical Methods in Natural Language Processing), one of the world’s leading international conferences in the field of natural language processing, which will be held in China starting on November 4, 2025.

This research introduces a new "Persuasive Dialogue AI" technology that integrates insights from diverse academic fields, including social psychology and behavioral economics. Experimental results demonstrate its high effectiveness in positively influencing the attitudes of recipients who previously had low purchase or behavioral intent, a cohort that has traditionally been difficult to approach. PKSHA will apply this research to the AI and AI agents it offers to enhance communication in settings such as contact centers and sales operations, thereby contributing to the business growth of its client companies.

Key Features of the Research:

Since the advent of generative AI, its applications have been expanding rapidly, increasing the demand for the automation of more complex dialogues such as coaching, counseling, and sales. While these applications require the ability to strategically engage in goal-oriented dialogue, current dialogue AI has yet to achieve human-level performance.

This research newly incorporates persuasion strategies rooted in behavioral psychology, which are widely used in practical settings, and has demonstrated outstanding capabilities, particularly in persuading individuals with initially low purchase or behavioral intent. The practical application of this research will support the realization of more advanced problem-solving and decision-making by encouraging behavioral change through appropriate communication based on objectives and situations.

Paper Details:
Title: Enhancing Persuasive Dialogue Agents by Synthesizing Cross-Disciplinary Communication Strategies Authors : Shinnosuke Nozue (Tohoku University), Yuto Nakano (Tohoku University), Yotaro Watanabe (PKSHA Technology Inc.), Meguru Takasaki (PKSHA Technology Inc.), Shoji Moriya (Tohoku University), Reina Akama (Tohoku University / NINJAL), Jun Suzuki (Tohoku University) This paper explores effective strategies for persuasive dialogue agents. Current approaches often rely on a limited set of predefined strategies, failing to capture the complexity of real-world persuasive interactions. To create more practically relevant strategies, we propose a cross-disciplinary approach drawing on social psychology, behavioral economics, and communication theory. We validate our method through experiments on two distinct datasets: P4G, representing a specific, in-domain scenario, and DailyPersuasion, covering a wide variety of scenarios. Our approach achieves strong results across both of these datasets, demonstrating a significant improvement in persuasion success rates and suggesting promising generalizability. Notably, our method also excels in persuading individuals with initially low intent, addressing a critical challenge in persuasive AI.

About EMNLP 2025:

The EMNLP 2025 (Conference on Empirical Methods in Natural Language Processing) is a top international conference in the field of Natural Language Processing, held in Suzhou, China, from November 4 to November 9, 2025. This conference is organized by the Association for Computational Linguistics (ACL) and is an authoritative venue that attracts the latest research in NLP based on empirical methods from around the world. The special theme this year is "Interdisciplinary Recontextualization of NLP," and cutting-edge topics such as the safety, efficiency, ethics, and interdisciplinary application of Large Language Models (LLMs) will be discussed. The accepted Industry Track aims to delve into language technology methods in industrial applications, and submissions related to technological innovation and implementation relevant to real-world applications are highly valued.

Future Outlook:

Based on the results of this research, we will accelerate research and development toward the practical application of this "Persuasive Dialogue AI" technology and expand the communication capabilities of PKSHA AI Agents (※). Specifically, we envision its application in advanced operator support in contact centers and assistance in sales domains. Through this technology, PKSHA is committed to contributing to the business growth of our client companies and, under our mission of "Shaping the Future of Software," aims to realize a sustainable society in which humans and software evolve together to solve social challenges.

(※) About "PKSHA AI Agents": https://prtimes.jp/main/html/rd/p/000000192.000022705.html

About Tohoku University Language AI Research Center:
Tohoku University boasts one of the largest research groups in the field of Natural Language Processing in Japan, with internationally recognized achievements. Numerous papers have been accepted at the highest-level international conferences in the fields of natural language processing, machine learning, and artificial intelligence (including ACL, EMNLP, NAACL, NeurIPS, ICLR, ICML, AAAI, IJCAI), underscoring the high caliber of its research capabilities.
The Language AI Research Center, established in October 2023, is tasked with consolidating these outstanding researchers to drive the development and application of advanced AI technologies, including Generative AI. Furthermore, it also aims to contribute to the advancement of AI technology across a wide range of areas, such as strengthening AI alignment and safety research and promoting AI-driven research.
Center Name:Tohoku University Language AI Research Center Center Address:6-3-09 Aoba, Aramaki, Aoba-ku, Sendai, Miyagi, Japan Representative:Center Director Jun Suzuki URL:https://langai.tohoku.ac.jp/

About PKSHA Technology Inc.
PKSHA, with the mission of "shaping the software of the future," offers a range of AI and AI Agents addressing social issues. These are deployed as tailored for industries such as finance, manufacturing, and education, as well as versatile products including "PKSHA AI Help Desk," and "PKSHA Chat Agent," The company supports future work styles and aims to foster a society where humans and software evolve together.
Company Name:PKSHA Technology Inc.
Location:Hongo Segawa Building 4F, 2-35-10 Hongo, Bunkyo-ku, Tokyo
Representative:Representative Director UENOYAMA Katsuya
URL:https://www.pkshatech.com/