Large Language Models (LLMs) as clinical agents require careful behavioral adaptation. While adept at reactive tasks (e.g., diagnosis reasoning), LLMs often struggle with proactive engagement, like unprompted identification of critical missing information or risks. We introduce **BehaviorBench**, a comprehensive dataset to evaluate agent behaviors across a clinical assistance spectrum, ranging from reactive query responses to proactive interventions (e.g., clarifying ambiguities, flagging overlooked critical data). Our BehaviorBench experiments reveal LLMs' inconsistent proactivity. To address this, we propose **BehaviorSFT**, a novel training strategy using behavioral tokens to explicitly condition LLMs for dynamic behavioral selection along this spectrum. BehaviorSFT boosts performance, achieving up to 97.3% overall Macro F1 on BehaviorBench and improving proactive task scores (e.g., from 95.0% to 96.5% for Qwen2.5-7B-Ins). Crucially, blind clinician evaluations confirmed BehaviorSFT-trained agents exhibit more realistic clinical behavior, striking a superior balance between helpful proactivity (e.g., timely, relevant suggestions) and necessary restraint (e.g., avoiding over-intervention) versus standard fine-tuning or explicit instructed agents.

BehaviorSFT: Behavioral Token Conditioning for Health Agents Across the Proactivity Spectrum

Conversational breakdowns in close relationships are deeply shaped by personal histories and emotional context, yet most NLP research treats conflict detection as a general task, overlooking the relational dynamics that influence how messages are perceived. In this work, we leverage nonviolent communication (NVC) theory to evaluate LLMs in detecting conversational breakdowns and assessing how relationship backstory influences both human and model perception of conflicts. Given the sensitivity and scarcity of real-world datasets featuring conflict between familiar social partners with rich personal backstories, we contribute the PersonaConflicts Corpus, a dataset of N=5,772 naturalistic simulated dialogues spanning diverse conflict scenarios between friends, family members, and romantic partners. Through a controlled human study, we annotate a subset of dialogues and obtain fine-grained labels of communication breakdown types on individual turns, and assess the impact of backstory on human and model perception of conflict in conversation. We find that the polarity of relationship backstories significantly shifted human perception of communication breakdowns and impressions of the social partners, yet models struggle to meaningfully leverage those backstories in the detection task. Additionally, we find that models consistently overestimate how positively a message will make a listener feel. Our findings underscore the critical role of personalization to relationship contexts in enabling LLMs to serve as effective mediators in human communication for authentic connection.

Words Like Knives: Backstory-Personalized Modeling and Detection of Violent Communication

We propose a large language model based reward decomposition framework for aligning dialogue agents using only a single session-level feedback signal. We leverage the reasoning capabilities of a frozen, pretrained large language model (LLM) to infer fine-grained local implicit rewards by decomposing global, session-level feedback. Our first text-only variant prompts the LLM to perform reward decomposition using only the dialogue transcript. The second multimodal variant incorporates additional behavioral cues, such as pitch, gaze, and facial affect, expressed as natural language descriptions. These inferred turn-level rewards are distilled into a lightweight reward model, which we utilize for RL-based fine-tuning for dialogue generation. We evaluate both text-only and multimodal variants against state-of-the-art reward decomposition methods and demonstrate notable improvements in human evaluations of conversation quality, suggesting that LLMs are strong reward decomposers that obviate the need for manual reward shaping and granular human feedback.

Aligning Dialogue Agents with Global Feedback via Large Language Model Multimodal Reward Decomposition

Part of a university initiative supporting responsible AI for social empowerment and education, the project-based RAICA (Responsible AI for Computational Action) curriculum supports middle/high school learners and novice AI literacy teachers use AI creatively for good. This paper offers a rare example of design-based implementation research (DBIR) in AI education across widely varied contexts, provides fine grain implementation data that contributes to a foundation for evaluating effectiveness and expanding access. We present a novel approach to analyzing fidelity of implementation data from RAICA’s computer vision module beta-test. Twelve educators working with ~282 students across nine pilot sites in four countries used a bespoke fidelity of implementation data collection tool (pre-made comment prompts in a Google Docs version of the teacher guide) to provide 236 qualitative responses about AI literacy and responsible design activities, plus 111 ordinal ratings of embedded teacher supports. Analyses revealed that while the curriculum was generally implemented as designed, educators frequently made modifications. Although most changes produced practical insights for improved curriculum design, others helped the design team anticipate and prevent changes that could obscure learning objectives and hinder outcomes. We discuss the pedagogical, design, and research implications of these findings for effective AI teaching/learning in diverse settings.

Advancing Research on Equitable AI Education Through a Focus on Implementation: Insights from a Middle School Computer Vision Module Beta-Test

Empathy serves as a cornerstone in enabling prosocial behaviors, and can be evoked through sharing of personal experiences in stories. While empathy is influenced by narrative content, intuitively, people respond to the way a story is told as well, through narrative style. Yet the relationship between empathy and narrative style is not fully understood. In this work, we empirically examine and quantify this relationship between style and empathy using LLMs and large-scale crowdsourcing studies. We introduce a novel, theory-based taxonomy, HEART (Human Empathy and Narrative Taxonomy) that delineates elements of narrative style that can lead to empathy with the narrator of a story. We establish the performance of LLMs in extracting narrative elements from HEART, showing that prompting with our taxonomy leads to reasonable, human-level annotations beyond what prior lexicon-based methods can do. To show empirical use of our taxonomy, we collect a dataset of empathy judgments of stories via a large-scale crowdsourcing study with $N=2,624$ participants. We show that narrative elements extracted via LLMs, in particular, vividness of emotions and plot volume, can elucidate the pathways by which narrative style cultivates empathy towards personal stories. Our work suggests that such models can be used for narrative analyses that lead to human-centered social and behavioral insights.

HEART-felt Narratives: Tracing Empathy and Narrative Style in Personal Stories with LLMs

We describe an approach for aligning an LLM based dialogue agent for long-term social dialogue, where there is only a single global score given by the user at the end of the session. In this paper, we propose the usage of denser naturally-occurring multimodal communicative signals as local implicit feedback to improve the turn-level utterance generation. Therefore, our approach (dubbed GELI) learns a local, turn-level reward model by decomposing the human-provided Global Explicit (GE) session level reward, using Local Implicit (LI) multimodal reward signals to crossmodally shape the reward decomposition step. This decomposed reward model is then used as part of the RLHF pipeline to improve an LLM-based dialog agent. We run quantitative and qualitative human studies on two large-scale datasets to evaluate the performance of our GELI approach, and find that it shows consistent improvements across various conversational metrics compared to baseline methods.

Global Reward to Local Rewards: Multimodal-Guided Decomposition for Improving Dialogue Agents

Social-emotional learning (SEL) skills are essential for children to develop to provide a foundation for future relational and academic success. Using art as a medium for creation or as a topic to provoke conversation is a well-known method of SEL education. Similarly, social robots have been used to teach SEL competencies like empathy, but the combination of art and social robotics has been minimally explored. In this paper, we present a novel child-robot interaction designed to foster empathy and promote SEL competencies via a conversation about art scaffolded by a social robot. Participants (N=11, age range: 7-11) conversed with a social robot about emotional and neutral art. Analysis of video and speech data demonstrated that this interaction design successfully engaged children in the practice of SEL skills, like emotion recognition and self-awareness, and greater rates of empathetic reasoning were observed when children engaged with the robot about emotional art. This study demonstrated that art-based reflection with a social robot, particularly on emotional art, can foster empathy in children, and interactions with a social robot may help alleviate discomfort when sharing deep or vulnerable emotions.

A HeARTfelt Robot: Social Robot-Driven Deep Emotional Art Reflection with Children

Modeling empathy is a complex endeavor that is rooted in interpersonal and experiential dimensions of human interaction, and remains an open problem within AI. Existing empathy datasets fall short in capturing the richness of empathy responses, often being confined to in-lab or acted scenarios, lacking longitudinal data, and missing self-reported labels. We introduce a new multimodal dataset for empathy during personal experience sharing: the EmpathicStories++ dataset containing 53 hours of video, audio, and text data of 41 participants sharing vulnerable experiences and reading empathically resonant stories with an AI agent. EmpathicStories++ is the first longitudinal dataset on empathy, collected over a month-long deployment of social robots in participants' homes, as participants engage in natural, empathic storytelling interactions with AI agents. We then introduce a novel task of predicting individuals' empathy toward others' stories based on their personal experiences, evaluated in two contexts: participants' own personal shared story context and their reflections on stories they read. We benchmark this task using state-of-the-art models to pave the way for future improvements in contextualized and longitudinal empathy modeling. Our work provides a valuable resource for further research in developing empathetic AI systems and understanding the intricacies of human empathy within genuine, real-world settings.

EmpathicStories++: A Multimodal Dataset for Empathy Towards Personal Experiences

Despite the many use cases for large language models (LLMs) in creating personalized chatbots, there has been limited research on evaluating the extent to which the behaviors of personalized LLMs accurately and consistently reflect specific personality traits. We consider studying the behavior of LLM-based agents which we refer to as LLM personas and present a case study with GPT-3.5 and GPT-4 to investigate whether LLMs can generate content that aligns with their assigned personality profiles. To this end, we simulate distinct LLM personas based on the Big Five personality model, have them complete the 44-item Big Five Inventory (BFI) personality test and a story writing task, and then assess their essays with automatic and human evaluations. Results show that LLM personas' self-reported BFI scores are consistent with their designated personality types, with large effect sizes observed across five traits. Additionally, LLM personas' writings have emerging representative linguistic patterns for personality traits when compared with a human writing corpus. Furthermore, human evaluation shows that humans can perceive some personality traits with an accuracy of up to 80\%. Interestingly, the accuracy drops significantly when the annotators were informed of AI authorship.

PersonaLLM: Investigating the Ability of Large Language Models to Express Personality Traits

Text-to-image generation (TTIG) technologies are Artificial Intelligence (AI) algorithms that use natural language algorithms in combination with visual generative algorithms. Consumer TTIG tools have gained popularity in recent months, garnering interest from non-AI experts, including educators and K-12 students. While they have exciting creative potential when used by K-12 learners and educators for creative learning, they are also accompanied by serious ethical implications, such as data privacy, spreading misinformation, and algorithmic bias. Given the potential learning applications, social implications, and ethical concerns, we designed learning materials to teach K-12 teachers from diverse subject expertise about the technical implementation, classroom applications, and ethical implications of TTIG algorithms. We piloted the learning materials titled “Demystify text-to-image generative tools for K-12 educators" with 51 teachers across two 6-hour workshops with the goal of preparing them to teach about and use TTIG tools in their classrooms. We found that teachers demonstrated a technical, applied and ethical understanding of TTIG algorithms and successfully designed early prototypes of teaching materials for their K-12 classrooms.

A Picture Is Worth a Thousand Words: Co-designing Text-to-Image Generation Learning Materials for K-12 with Educators | VIDEO

The most meaningful connections between people are often fostered through expression of shared vulnerability and emotional experiences in personal narratives. We introduce a new task of identifying similarity in personal stories based on empathic resonance, i.e., the extent to which two people empathize with each others' experiences, as opposed to raw semantic or lexical similarity, as has predominantly been studied in NLP. Using insights from social psychology, we craft a framework that operationalizes empathic similarity in terms of three key features of stories: main events, emotional trajectories, and overall morals or takeaways. We create EmpathicStories, a dataset of 1,500 personal stories annotated with our empathic similarity features, and 2,000 pairs of stories annotated with empathic similarity scores. Using our dataset, we fine-tune a model to compute empathic similarity of story pairs, and show that this outperforms semantic similarity models on automated correlation and retrieval metrics. Through a user study with 150 participants, we also assess the effect our model has on retrieving stories that users empathize with, compared to naive semantic similarity-based retrieval, and find that participants empathized significantly more with stories retrieved by our model. Our work has strong implications for the use of empathy-aware models to foster human connection and empathy between people.

Modeling Empathic Similarity in Personal Narratives

AI Audit: A card game to reflect on everyday AI systems

Analyzing the COMPAS Recidivism Algorithm

As social robots rapidly become mainstream tech- nologies, it is critical for HRI researchers and practitioners to consider their societal and ethical impacts as well as their ability to perpetuate or mitigate intersectional social inequities and hierarchies relating to race, class, gender, disability, and other social axes. Through an equity, ethics, and justice- centered audit of human-robot interaction (HRI) scholarship, we reveal how the HRI community has engaged with these topics over the past two decades. We use the five senses ethical framework that has been proposed specifically for use in HRI contexts to perform the review paired with an analysis of equity and justice. We then expand the Design Justice framework (a framework for analyzing how design impacts society and distributes benefits and burdens to society through the lenses of equity, values, scope, ownership, and accountability) to HRI contexts through the inclusion of HRI-specific topics such as autonomy, transparency, deception, and policies. We invite researchers and practitioners to explore the HRI Equitable Design framework to work towards designing equitable and inclusive HRI research studies and technologies.

Ethics, Equity, & Justice in Human-Robot Interaction: A Review and Future Directions

The main contribution of this work is in the development of a long-term 12-month older adult co-design methodology for home social robot design with long-term co-design guidelines for how other researchers can adopt the methodology. Through the process, we promote co-design as a human-centered approach for designing robots, empowering older adults in the design of these technologies.

Long-Term Co-Design Guidelines: Empowering Older Adults As Co-Designers of Social Robots

Retrofitting is a technique used to move word vectors closer together or further apart in their space to reflect their relationships in a Knowledge Base (KB). However, retrofitting only works on concepts that are present in that KB. RetroGAN uses a pair of Generative Adversarial Networks (GANs) to learn a one-to-one mapping between concepts and their
retrofitted counterparts. It applies that mapping (post-specializes) to handle concepts that
do not appear in the original KB in a manner similar to how some natural language systems handle out-of-vocabulary entries. We test our system on three word-similarity benchmarks and a downstream sentence simplification task, and achieve the state of the art (CARD-660). Altogether, our results demonstrate our system’s effectiveness for out-of-knowledge and rare word generalization.

Cynthia Breazeal

18

51

2

SHORT BIO

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES