Natural language instructions are often abstract and complex, requiring robots to execute multiple subtasks even for seemingly simple queries. For example, when a user asks a robot to prepare avocado toast, the task involves several sequential steps. Moreover, such instructions can be ambiguous or infeasible for the robot or may exceed the robot's existing knowledge. While Large Language Models (LLMs) offer strong language reasoning capabilities to handle these challenges, effectively integrating them into robotic systems remains a key challenge. To address this, we propose BT-ACTION, a test-driven approach that combines the modular structure of Behavior Trees (BT) with LLMs to generate coherent sequences of robot actions for following complex user instructions, specifically in the context of preparing recipes in a kitchen-assistance setting. We evaluated BT-ACTION in a comprehensive user study with 45 participants, comparing its performance to direct LLM prompting. Results demonstrate that the modular design of BT-ACTION helped the robot make fewer mistakes and increased user trust, and participants showed a significant preference for the robot leveraging the modular approach. The code is publicly available at https://github.com/1Eggbert7/BT_LLM.

Social robots need to be able to interact effectively with small groups. While there is a significant interest in human-robot interaction in groups, little focus has been placed on developing autonomous social robot decision-making methods that operate smoothly with small groups of any size (e.g. 2, 3, or 4 interactants). In this work, we propose a Template- and Graph-based Modeling approach for robots interacting in small groups (TGM), enabling them to interact with groups in a way that is group-size agnostic. Critically, we separate the decision about the target of their communication, or 'whom to address?' from the decision of 'what to communicate?', which allows us to use template-based actions. We further use Graph Neural Networks (GNNs) to efficiently decide on 'whom' and 'what'. We evaluated TGM using imitation learning and compared the structured reasoning achieved through GNNs to unstructured approaches for this two-part decision-making problem. On two different datasets, we show that TGM outperforms the baselines encouraging future work to invest in collecting larger datasets.

The goal of this thesis is to develop computational approaches generating autonomous social robot behaviors that can interact with multiple people and dynamically adapt to shape their interactions. Positive interactions between people impact their well-being and are essential to a fulfilled and healthy life. In this thesis, we coin the term Interaction-Shaping Robotics (ISR) as the study of robots that shape interactions between other agents, e.g., people, and capture previous efforts from the Human-Robot Interaction (HRI) community and emphasize the potential positive or negative, intended or unintended effects of these robots. Previous efforts have explored phenomena that indicate interaction-shaping capabilities of social robots, however, how to de-velop autonomous social robots that can adapt to positively shape interactions between people based on perceived human-human dynamics remains largely unexplored. In this thesis, we contribute to the technical advancement of social interaction-shaping robots by developing heuristics and machine learning methods and demonstrating their effectiveness in studies with real users. We focus on shaping behaviors, i.e., balancing people’s participation in interactions to foster inclusion among newly-arrived and already present children in a music game and support adult second language learners and native speakers in a language game. Especially when leveraging learning techniques, an effective interaction-shaping robot needs to act socially appropriately. We design heuristics that are appropriate by design and establish the feasibility of autonomy for interaction-shaping robots through minimal perception of group dynamics and simple behavior rules. Allowing for learning behaviors for more complex interactions, we provide a formal definition of the problem of interaction-shaping and show that using imitation learning (IL) or offline reinforcement learning (RL) based on previously collected HRI data is feasible without compromising the interaction. To meet the challenge of acting appropriately, we explore techniques applied prior to deployment when learning offline from data and shielding - a technique from the safe RL community - to eventually allow for learning during deployment in interaction. Overall, this thesis demonstrates the feasibility and promise of computational methods for autonomous interaction-shaping robots and demonstrates that these methods generate effective and appropriate robot behavior when balancing participation to ensure the inclusion of all human group members.

Målet med denna avhandling är att utveckla beräkningsbaserade meto-der för att generera autonoma sociala robotbeteenden som kan interagera med flera människor och dynamiskt anpassa sig för att forma deras interak-tioner. Positiva interaktioner mellan människor påverkar deras välbefinnande och är avgörande för ett meningsfullt och hälsosamt liv. I denna avhandling myntar vi termen "Interaction-Shaping Robotics"(ISR) som studerandet av robotar som formar interaktioner mellan andra aktörer, t.ex. människor, och sammanställer tidigare studier inom människ-robot-interaktion (eng. Human-Robot Interaction, HRI) samt betonar den potentiella positiva eller negativa, avsiktliga eller oavsiktliga, inverkan av dessa robotar. Tidigare studier har utforskat fenomen som indikerar på interaktionsformande förmågor hos sociala robotar, men utvecklandet av autonoma sociala robotar som kan anpassa sig för att positivt forma interaktioner mellan människor baserat på observerad människa-till-människa dynamik är fortfarande till stor del outforskat. I denna avhandling bidrar vi till den tekniska utvecklingen av sociala interaktionsformande robotar genom att utveckla heuristiker och maskininlärningsmetoder och demonstrera deras effektivitet i studier med användare. Vi fokuserar på att forma beteenden, d.v.s. balansera människors deltagande i interaktioner för att främja inkludering bland nyanlända och redan närvarande barn i ett musikspel och stödja vuxna andraspråksinlärare och modersmålstalare i ett språkspel. Särskilt när man utnyttjar maskininlärningsmetoder, behöver en effektiv interaktionsformande robot agera socialt korrekt. Vi designar heuristiker som är lämpliga by design” och fastställer genomförbarheten av autonomi för interaktionsformande robotar genom minimal perception av gruppdynamik och enkla beteenderegler. Genom att tillåta inlärning av beteenden för mer komplexa interaktioner, tillhandahåller vi en formell definition av problemet av interaktionsformande och visar att användning av imitationsinlärning (eng. imitation learning, IL) off-line förstärkningsinlärning (eng. reinforcement learning, RL), baserat på tidigare insamlad HRI-data är genomförbart utan att kompromissa med interaktionen. För att möta utmaningen att agera korrekt, utforskar vi tekniker som tillämpas innan implementering när man lär sig off-line från data och ”shielding” - en teknik inom säker RL - för att så småningom möjliggöra inlärning under implementering vid interaktion. Sammanfattningsvis visar denna avhandling genomförbarheten och utsikten av beräkningsbaserade metoder för autonoma interaktionsformande robotar och demonstrerar att dessa metoder genererar effektiva och lämpliga robotbeteenden när de balanserar deltagande för att säkerställa inkludering av alla mänskliga gruppmedlemmar.

Work in Human–Robot Interaction (HRI) has investigated interactions between one human and one robot as well as human–robot group interactions. Yet the field lacks a clear definition and understanding of the influence a robot can exert on interactions between other group members (e.g., human-to-human). In this article, we define Interaction-Shaping Robotics (ISR), a subfield of HRI that investigates robots that influence the behaviors and attitudes exchanged between two (or more) other agents. We highlight key factors of interaction-shaping robots that include the role of the robot, the robot-shaping outcome, the form of robot influence, the type of robot communication, and the timeline of the robot’s influence. We also describe three distinct structures of human–robot groups to highlight the potential of ISR in different group compositions and discuss targets for a robot’s interaction-shaping behavior. Finally, we propose areas of opportunity and challenges for future research in ISR.

Robot-moderated group discussions have the potential to facilitate engaging and productive interactions among human participants. Previous work on topic management in conversational agents has predominantly focused on human engagement and topic personalization, with the agent having an active role in the discussion. Also, studies have shown the usefulness of including robots in groups, yet further exploration is still needed for robots to learn when to change the topic while facilitating discussions. Accordingly, our work investigates the suitability of machine-learning models and audiovisual non-verbal features in predicting appropriate topic changes. We utilized interactions between a robot moderator and human participants, which we annotated and used for extracting acoustic and body language-related features. We provide a detailed analysis of the performance of machine learning approaches using sequential and non-sequential data with different sets of features. The results indicate promising performance in classifying inappropriate topic changes, outperforming rule-based approaches. Additionally, acoustic features exhibited comparable performance and robustness compared to the complete set of multimodal features. Our annotated data is publicly available at https://github.com/ghadj/topic-change-robot-discussions-data-2024.

A wide range of studies in Human-Robot Interaction (HRI) has shown that robots can influence the social behavior of humans. This phenomenon is commonly explained by the Media Equation. Fundamental to this theory is the idea that when faced with technology (like robots), people perceive it as a social agent with thoughts and intentions similar to those of humans. This perception guides the interaction with the technology and its predicted impact. However, HRI studies have also reported examples in which the Media Equation has been violated, that is when people treat the influence of robots differently from the influence of humans. To address this gap, we propose a model of Robot Social Influence (RoSI) with two contributing factors. The first factor is a robot's violation of a person's expectations, whether the robot exceeds expectations or fails to meet expectations. The second factor is a person's social belonging with the robot, whether the person belongs to the same group as the robot or a different group. These factors are primary predictors of robots' social influence and commonly mediate the influence of other factors. We review HRI literature and show how RoSI can explain robots' social influence in concrete HRI scenarios.

A crucial part of traditional reinforcement learning (RL) is the initial exploration phase, in which trying available actions randomly is a critical element. As random behavior might be detrimental to a social interaction, this work proposes a novel paradigm for learning social robot behavior-the use of shielding to ensure socially appropriate behavior during exploration and learning. We explore how a data-driven approach for shielding could be used to generate listening behavior. In a video-based user study (N=110), we compare shielded exploration to two other exploration methods. We show that the shielded exploration is perceived as more comforting and appropriate than a straightforward random approach. Based on our findings, we discuss the potential for future work using shielded and socially guided approaches for learning idiosyncratic social robot behaviors through RL.

A crucial part of traditional reinforcement learning (RL) is the initial exploration phase, in which trying available actions randomly is a critical element. As random behavior might be detrimental to a social interaction, this work proposes a novel paradigm for learning social robot behavior--the use of shielding to ensure socially appropriate behavior during exploration and learning. We explore how a data-driven approach for shielding could be used to generate listening behavior. In a video-based user study (N=110), we compare shielded exploration to two other exploration methods. We show that the shielded exploration is perceived as more comforting and appropriate than a straightforward random approach. Based on our findings, we discuss the potential for future work using shielded and socially guided approaches for learning idiosyncratic social robot behaviors through RL.   

With societies growing more and more conscious of human social biases that are implicit in most of our interactions, the development of automated robot social behavior is failing to address these issues as more than just an afterthought. In the present work, we describe how we unintentionally implemented robot listener behavior that was biased toward the gender of the participants, while following typical design procedures in the field. In a post-hoc analysis of data collected in a between-subject user study (n=60), we find that both a rule-based and a deep learning-based listener behavior models produced a higher number of backchannels (listener feedback, through nodding or vocal utterances) if the participant identified as a male. We investigate the cause of this bias in both models and discuss the implications of our findings. Further, we provide approaches that may be taken to address the issue of algorithmic fairness, and preventative measures to avoid the development of biased social robot behavior.

While thinking aloud has been reported to positively affect problem-solving, the effects of the presence of an embodied entity (e.g., a social robot) to whom words can be directed remain mostly unexplored. In this work, we investigated the role of a robot in a "rubber duck debugging"setting, by analyzing how a robot's listening behaviors could support a thinking-aloud problem-solving session. Participants completed two different tasks while speaking their thoughts aloud to either a robot or an inanimate object (a giant rubber duck). We implemented and tested two types of listener behavior in the robot: a rule-based heuristic and a deep-learning-based model. In a between-subject user study with 101 participants, we evaluated how the presence of a robot affected users' engagement in thinking aloud, behavior during the task, and self-reported user experience. In addition, we explored the impact of the two robot listening behaviors on those measures. In contrast to prior work, our results indicate that neither the rule-based heuristic nor the deep learning robot conditions improved performance or perception of the task, compared to an inanimate object. We discuss potential explanations and shed light on the feasibility of designing social robots as assistive tools in thinking-aloud problem-solving tasks.