Generating psychological counseling responses with language models relies heavily on high-quality datasets. Crowdsourced data collection methods require strict worker training, and data from real-world counseling environments may raise privacy and ethical concerns. While recent studies have explored using large language models (LLMs) to augment psychological counseling dialogue datasets, the resulting data often suffers from limited diversity and authenticity. To address these limitations, this study adopts a role-playing approach where trained counselors simulate counselor-client interactions, ensuring high-quality dialogues while mitigating privacy risks. Using this method, we construct KokoroChat, a Japanese psychological counseling dialogue dataset comprising 6,589 long-form dialogues, each accompanied by comprehensive client feedback. Experimental results demonstrate that fine-tuning open-source LLMs with KokoroChat improves both the quality of generated counseling responses and the automatic evaluation of counseling dialogues. The KokoroChat dataset is available at https://github.com/UEC-InabaLab/KokoroChat.

Conversational Recommender Systems (CRSs) aim to elicit user preferences via natural dialogue to provide suitable item recommendations.However, current CRSs often deviate from realistic human interactions by rapidly recommending items in brief sessions.This work addresses this gap by leveraging Large Language Models (LLMs) to generate dialogue summaries from dialogue history and item recommendation information from item description.This approach enables the extraction of both explicit user statements and implicit preferences inferred from the dialogue context.We introduce a method using Direct Preference Optimization (DPO) to ensure dialogue summary and item recommendation information are rich in information crucial for effective recommendations.Experiments on two public datasets validate our method's effectiveness in fostering more natural and realistic conversational recommendation processes.Our implementa

Recent advancements in AI have highlighted its application in captology, the field of using computers as persuasive technologies. We hypothesized that the “conformity effect,” where individuals align with others’ actions, also occurs with AI agents. This study verifies this hypothesis by introducing a “Persuadee Agent” that is persuaded alongside a human participant in a three-party persuasive dialogue with a Persuader Agent. We conducted a text-based dialogue experiment with human participants. We compared four conditions manipulating the Persuadee Agent’s behavior (persuasion acceptance vs. non-acceptance) and the presence of an icebreaker session. Results showed that when the Persuadee Agent accepted persuasion, both perceived persuasiveness and actual attitude change significantly improved. Attitude change was greatest when an icebreaker was also used, whereas an unpersuaded AI agent suppressed attitude change. Additionally, it was confirmed that the persuasion acceptance of participants increased at the moment the Persuadee Agent was persuaded. These results suggest that appropriately designing a Persuadee Agent can improve persuasion through the conformity effect.

Open-domain dialogue systems have been increasingly applied in various situations, with a growing need to improve user engagement. One effective approach is to generate responses based on interesting external knowledge using knowledge-grounded response generation models. However, relying solely on interestingness can lead to incoherent responses, potentially diminishing user engagement. This paper proposes a novel method for generating engaging responses while maintaining contextual coherence. Our approach leverages a pre-trained knowledge-grounded response generation model and modifies the knowledge selection process to enhance response coherence and interestingness without requiring additional training. First, knowledge candidates with high contextual relevance are retrieved. These candidates are then reranked based on their interestingness and used to generate the responses. Finally, the method detects dialogue breakdowns and regenerates responses as necessary to ensure coherence. We conducted experiments using the Wizard of Wikipedia dataset and two state-of-the-art response generation models. The results indicate that the proposed method improves both response coherence and interestingness.

Real-time collaborative dialogue tasks require dynamic, instantaneous decision-making and seamless coordination between participants, yet most existing studies on cooperative dialogues primarily focus on turn-based textual environments. This study addresses the critical gap in understanding human-human interaction patterns within dynamic, real-time collaborative scenarios. In this paper, we present a novel dataset collected from a real-time collaborative cooking game environment inspired by the popular game “Overcooked.” Our dataset comprises detailed annotations of participants’ task proficiency levels, game scores, game action logs, and transcribed voice dialogues annotated with dialogue act tags. Participants exhibited a broad range of gaming experience, from highly proficient players to those with minimal exposure to gaming controls. Through comprehensive analysis, we explore how individual differences in task proficiency influence dialogue patterns and collaborative outcomes. Our findings reveal key dialogue acts and adaptive communication strategies crucial for successful real-time collaboration. Furthermore, this study provides valuable insights into designing adaptive dialogue systems capable of dynamically adjusting interaction strategies based on user proficiency, paving the way for more effective human-AI collaborative systems. The dataset introduced in this study is publicly available at: https://github.com/UEC-InabaLab/OverCookedChat.

Challenges in multimodal task-oriented dialogue between humans and systems, particularly those involving audio and visual interactions, have not been sufficiently explored or shared, forcing researchers to define improvement directions individually without a clearly shared roadmap. To address these challenges, we organized a competition for multimodal task-oriented dialogue systems and constructed a large competition-based dataset of 1,865 minutes of Japanese task-oriented dialogues. This dataset includes audio and visual interactions between diverse systems and human participants. After analyzing system behaviors identified as problematic by the human participants in questionnaire surveys and notable methods employed by the participating teams, we identified key challenges in multimodal task-oriented dialogue systems and discussed potential directions for overcoming these challenges.

In human-human conversation, interpersonal consideration for the interlocutor is essential, and similar expectations are increasingly placed on dialogue systems. This study examines the behavior of dialogue systems in a specific interpersonal scenario where a user vents frustrations and seeks emotional support from a long-time friend represented by a dialogue system. We conducted a human evaluation and qualitative analysis of 15 dialogue systems under this setting. These systems implemented diverse strategies, such as structuring dialogue into distinct phases, modeling interpersonal relationships, and incorporating cognitive behavioral therapy techniques. Our analysis reveals that these approaches contributed to improved perceived empathy, coherence, and appropriateness, highlighting the importance of design choices in socially sensitive dialogue.

We developed an interface to explain news articles through dialogue by considering the user's comprehension level.The interface generates several pertinent questions based on the ongoing dialogue and news article, and users advance the conversation by selecting a question.Based on the user's selected questions, the interface estimates their comprehension level of the news article and adjusts the difficulty of the generated questions accordingly.This enables a personalized dialogue tailored to each user's comprehension needs.The results of the baseline comparison experiments confirmed the usefulness of the interface. Candidate questions generationUser's comprehension level

Persona-aware dialogue systems can improve the consistency of the system's responses, users' trust and user enjoyment.Filtering nonpersona-like utterances is important for constructing persona-aware dialogue systems.This paper presents the PersonaCLR model for capturing a given utterance's intensity of persona characteristics.We trained the model with contrastive learning based on the sameness of the utterances' speaker.Contrastive learning enables PersonaCLR to evaluate the persona characteristics of a given utterance, even if the target persona is not included in training data.For training and evaluating our model, we also constructed a new dataset of 2,155 character utterances from 100 Japanese online novels.Experimental results indicated that our model outperforms existing methods and a strong baseline using a large language model.Our source code, pre-trained model, and dataset are available at https://github.com/1never/PersonaCLR.

This study addresses the interaction challenges encountered by spoken dialogue systems (SDSs) when engaging with users who exhibit distinct conversational behaviors, particularly minors, in scenarios where data are scarce.We propose a novel data augmentation framework to enhance SDS performance for user groups with limited resources.Our approach leverages a large language model (LLM) to extract speaker styles and a pre-trained language model (PLM) to simulate dialogue act history.This method generates enriched and personalized dialogue data, facilitating improved interactions with unique user demographics.Extensive experiments validate the efficacy of our methodology, highlighting its potential to foster the development of more adaptive and inclusive dialogue systems.

User reviews on e-commerce and review sites are crucial for making purchase decisions, although creating detailed reviews is timeconsuming and labor-intensive.In this study, we propose a novel use of dialogue systems to facilitate user review creation by generating reviews from information gathered during interview dialogues with users.To validate our approach, we implemented our system using GPT-4 and conducted comparative experiments from the perspectives of system users and review readers.The results indicate that participants who used our system rated their interactions positively.Additionally, reviews generated by our system required less editing to achieve user satisfaction compared to those by the baseline.We also evaluated the reviews from the readers' perspective and found that our system-generated reviews are more helpful than those written by humans.Despite challenges with the fluency of the generated reviews, our method offers a promising new approach to review writing.

When individuals communicate, they use different vocabularies, speaking speeds, facial expressions, and gestural languages, depending on those with whom they are speaking. This study focuses on the age of the speaker as a factor that affects the style of communication. We collected a multimodal dialogue corpus with various speaker ages. We used travel as the topic, as it interests people of all ages, and we set up a task based on a tourism consultation between an operator and a customer at a travel agency. This article presents the details of the dialogue task, collection procedures and annotations, and analysis of the characteristics of the dialogues and facial expressions, focusing on the age of the speakers. The results of the analysis suggest that the adult speakers have more independent opinions, the older speakers express their opinions more frequently than other age groups, and those in the operator role smile more frequently at minors.

When individuals communicate with each other, they use different vocabulary, speaking speed, facial expressions, and body language depending on the people they talk to. This paper focuses on the speaker’s age as a factor that affects the change in communication. We collected a multimodal dialogue corpus with a wide range of speaker ages. As a dialogue task, we focus on travel, which interests people of all ages, and we set up a task based on a tourism consultation between an operator and a customer at a travel agency. This paper provides details of the dialogue task, the collection procedure and annotations, and the analysis on the characteristics of the dialogues and facial expressions focusing on the age of the speakers. Results of the analysis suggest that the adult speakers have more independent opinions, the older speakers more frequently express their opinions frequently compared with other age groups, and the operators expressed a smile more frequently to the minor speakers.

Dialogue personalization is an important issue in the field of open-domain chat-oriented dialogue systems. If these systems could consider their users' interests, user engagement and satisfaction would be greatly improved. This paper proposes a neural network-based method for estimating users' interests from their utterances in chat dialogues to personalize dialogue systems' responses. We introduce a method for effectively extracting topics and user interests from utterances and also propose a pre-training approach that increases learning efficiency. Our experimental results indicate that the proposed model can estimate user's interest more accurately than baseline approaches.

“Are you a werewolf?” is one of the most popular communication games and is played globally. The AIWolf Project developed an agent, named “the AIWolf,” that can play “Are you a werewolf?”. An AIWolf utters its thoughts using an AIWolf Protocol. As it is difficult for humans to understand the AIWolf Protocol, translation into natural language is required when human players are involved. However, the conventional method of translation uses a word-to-word method, creating the impression that the utterances have been generated by a machine. This study aimed make the utterances of AIWolf sound more human. The authors set the target that a human player would be unable to distinguish human speech from that generated by AIWolves (the Turing test). The authors define the situation as the maximum value of humanity. The output of translated AIWolf Protocol was paraphrased using data from Werewolf BBS Logs. This study considers making the utterances of AIWolf sound more human using Werewolf BBS Logs and a possibility assignment equation with fuzzy sets. In this paper, an experiment was conducted to confirm whether paraphrasing the utterances of AIWolf using Werewolf BBS Logs for human-like speech is useful or not. It was shown that the experimental method produced slightly more human-like speech than the conventional method.

In this paper, we performed an experimental investigation aimed at developing an agent to support the formation of human relationships by supporting the user’s daily communication “casually”, “anytime” and “anywhere”. First, we collected conversations between men and women meeting for the first time, then analyzed what type of support would be effective for the formation of human relationships. Based on the results of this analysis, we performed experiments supporting communication. In the experiment, we provided not only topics to the user during conversation, but also comprehensive presentation of instructions such as expressions, eye contact and gestures. The results confirmed that this significantly improved human relationships after conversation and showed the validity of this support.

This paper proposes a method to improve genetic programming with multiple trees (GP CN ). An individual in GP CN comprises multiple trees, and each tree has a number P that indicates the number of repetitive actions based on the tree. In previous work, a method for updating the number P has been proposed to obtain P suitable to the tree in evolution. However, in the method efficiency becomes worse as the range of P becomes wider. In order to solve the problem, in this study, two methods are proposed: inheriting the number P of a tree from an excellent individual and using mutation for preventing the number P from being into a local optimum. Additionally, a method to eliminate trees consisting of a single terminal node is proposed.

Among many reinforcement learning methods, FALCON is a machine learning method which is an extend fuzzy ART(Adaptive Resonance Theory), and can appropriately discretize a state space. FALCON is an on-line method proposed by Ah-Hwee Tan. It can discretize a state space and learn action rules simultaneously by learning relations among percepts, actions, and rewards. In this study, a learning agent using FALCON is interactively trained, and the learning effect is measured through experiments. In experiments, the learning agent learns by playing 50,000 card games of “Hearts” against three rule-based agents. Then, the interface that agents can interactively play the game with human cooperators is made so that human cooperators can play the game against the learning agent to strengthen it. It continues learning during games. The effectiveness of interactive learning is ascertained through the experiments.

The 3-channel fuzzy ART network FALCON (Fusion Architecture for Learning, COgnition, and Navigation) is known as an effective method for combining reinforcement learning with state segmentation. It has been shown that FALCON is effective in making a player agent for the card game Hearts, although the agent was unable to beat an agent using the UCT algorithm developed for Monte-Carlo simulation. This study proposes an ensemble method for FALCON to make an agent stronger. The method uses nine types of learners and combines them to decide an action. Experiments demonstrate that our approach is superior to an agent using a single learner.

In this study, we present our neural utterance ranking (NUR) model, an utterance selection model for conversational dialogue agents.The NUR model ranks candidate utterances with respect to their suitability in relation to a given context using neural networks; in addition, a dialogue system based on the model converses with humans using highly ranked utterances.Specifically, the model processes word sequences in utterances and utterance sequences in context via recurrent neural networks.Experimental results show that the proposed model ranks utterances with higher precision relative to deep learning and other existing methods.Furthermore, we construct a conversational dialogue system based on the proposed method and conduct experiments on human subjects to evaluate performance.The experimental result indicates that our system can offer a response that does not provoke a critical dialogue breakdown with a probability of 92% and a very natural response with a probability of 58%.

In this research, we target at the interactive communication game “werewolf” with a subject of research. Werewolf is a popular party game all over the world, and the relevance studies have been advanced in recent years. However, the life-like agent who does werewolf has not been developed. Therefore the purpose of this research is to analyze non-verbal information from movies which play the werewolf with face-to-face communication and to make clear the impression for others by the movements of players in the game. Moreover, we verify whether the life-like agent gives an impression like human in werewolf game by mounting the movements on a life-like agent.

Werewolf is one of the popular communication games all over the world. It treats ambiguity of human discussion including the utterances, gestures and facial expressions in a broad sense. In this research, we pay attention to this imperfect information game werewolf. The purpose of the research is to develop an intelligent agent “AI werewolf” which is enabled to naturally play werewolf with human. This paper aims to investigate how behavior contribute to victory of own-side players by using machine learning as a first step. As the results of investigation and analysis of the playing movie, we found that nonverbal information in the game of werewolf has importance to winning or losing the game.

"Are You a Werewolf?" is a conversation type game. We construct a Werewolf match system for humans with lifelike agents. We evaluate whether it is possible to realize conversation space of "Are You a Werewolf?" with the system.

This paper provides a novel method for building non-task-oriented dialogue agents such as chatbots. The dialogue agent constructed using our method automatically selects a suitable utterance depending on a context from a set of candidate utterances prepared in advance. To realize automatic utterance selection, we rank the candidate utterances in order of suitability by application of a machine learning algorithm. We employed both right and wrong dialogue data to learn relative suitability to rank the utterances. Additionally, we provide a low-cost and quality-assured learning data acquisition environment using crowdsourcing and gamification. The results of an experiment using learning data obtained via the environment demonstrate that the appropriate utterance is ranked on the top in 82.6% of cases and within the top 3 at 95.0% of cases. Results show that using context information that is not used in most existing agents is necessary for appropriate responses.

This paper proposes Genetic Programming(GP) with control nodes using the conditional probability and the island model for efficient learning of agent behavior. In the methods, each individual has a chromosome representing agent behavior as several trees. In GP using the conditional probability, individuals with high fitness values are used to produce conditional probability tables to generate individuals in the next generation. In GP using the island model, the population is divided into two islands of individuals: one island keeps diversity of individuals and the other puts emphasis on the accuracy of the solution. The methods are applied to a garbage collection problem and Santa Fe Trail problem. The proposed methods are compared with traditional GP, GP with control nodes, and Genetic Network Programming(GNP) with control nodes. Experimental results show that the proposed methods are efficient.

This paper discusses the group control of elevators for improving efficiency, an efficient control method for multi-car elevator using reinforcement learning is proposed. In the method, the control agent selects the best strategy among four strategies, namely Transportation strategy, Passenger strategy, Zone strategy, and Difference strategy according to traffic flow. The control agent takes the number of total passengers and the distance from the departure floor to the destination floor of a call into account. Through experiments, the performance of the proposed method is shown, the average service time of the proposed method is compared with the average service time obtained for the cases where the car assignment is made by each of the three or four strategies.

E-learning systems that use computers and the internet have become popular and ubiquitous. E-learning systems offer many benefits. However, users often lose their motivation for learning during the process of studying, and the frequency with which they use e-learning systems sometimes decreases. Results show that a function to praise or scold users by displaying images of a teacher is effective to improve or maintain their motivation. In this study, that function is modified to display images that users like or dislike, respectively, for praising and scolding learners. Through experimentation, this study assesses the effects of displayed images on maintaining the will to learn.

E-learning systems that use computers and the Internet have become popular. E-learning systems have many advantages. However, the users often lose their motivation for learning in the process of studying, and the frequency that they use e-learning systems sometimes decreases. In order to improve or keep their motivation, we add a function that the users are praised or scolded by displaying images of a teacher, actors, and friends. We check to see if the function is effective or not through experiments.